Keywords
Computer Science and Digital Science
 A1.2. Networks
 A1.3.5. Cloud
 A1.3.6. Fog, Edge
 A1.6. Green Computing
 A3.4. Machine learning and statistics
 A3.5.2. Recommendation systems
 A5.2. Data visualization
 A6. Modeling, simulation and control
 A6.2.3. Probabilistic methods
 A6.2.4. Statistical methods
 A6.2.6. Optimization
 A6.2.7. High performance computing
 A8.2. Optimization
 A8.9. Performance evaluation
 A8.11. Game Theory
 A9.2. Machine learning
 A9.9. Distributed AI, Multiagent
Other Research Topics and Application Domains
 B4.4. Energy delivery
 B4.4.1. Smart grids
 B4.5.1. Green computing
 B6.2. Network technologies
 B6.2.1. Wired technologies
 B6.2.2. Radio technology
 B6.4. Internet of things
 B8.3. Urbanism and urban planning
 B9.6.7. Geography
 B9.7.2. Open data
 B9.8. Reproducibility
1 Team members, visitors, external collaborators
Research Scientists
 Arnaud Legrand [Team leader, CNRS, Researcher, until Jun 2022, HDR]
 Jonatha Anselmi [INRIA, Researcher]
 Nicolas Gast [INRIA, Researcher, HDR]
 Bruno Gaujal [INRIA, Senior Researcher, HDR]
 Panayotis Mertikopoulos [CNRS, Researcher, HDR]
 Bary Pradelski [CNRS, Researcher]
Faculty Members
 Romain Couillet [GRENOBLE INP, Professor, HDR]
 Vincent Danjean [UGA, Associate Professor]
 Guillaume Huard [UGA, Associate Professor]
 Florence Perronnin [UGA, Associate Professor]
 JeanMarc Vincent [UGA, Associate Professor]
 Philippe Waille [UGA, Associate Professor]
PostDoctoral Fellows
 HenryJoseph Audéoud [INPGSA Entreprise, until Jun 2022]
 Simon Finster [CNRS, from Nov 2022]
 Sacha Hodencq [FLORALIS, from Oct 2022]
 Dong Quan Vu [CNRS, until Mar 2022]
PhD Students
 Sebastian Allmeier [INRIA]
 Achille Baucher [FLORALIS, from Oct 2022]
 Victor Boone [ENS DE LYON]
 Rémi Castera [UGA]
 Romain Cravic [INRIA]
 YuGuan Hsieh [UGA]
 Simon Jantschgi [UNIV ZURICH, from Aug 2022]
 Simon Jantschgi [CNRS, until Jul 2022]
 Kimang Khun [INRIA]
 Till Kletti [NAVER LABS]
 Lucas Leandro Nesi [GOUV BRESIL]
 Hugo Lebeau [UGA]
 Davide Legacci [UGA, from Nov 2022]
 Victor Leger [UGA]
 Mathieu Molina [INRIA]
 LouisSebastien Rebuffi [UGA]
 Chen Yan [UGA]
Technical Staff
 Achille Baucher [FLORALIS, Engineer, from Apr 2022 until Jun 2022]
Interns and Apprentices
 Achille Baucher [FLORALIS, until Mar 2022]
 Ran Cheng [UGA, Intern, from May 2022 until Jul 2022]
 Armand Grenier [UGA, Intern, from Jun 2022 until Aug 2022]
Administrative Assistant
 Annie Simon [INRIA]
2 Overall objectives
2.1 Context
Large distributed infrastructures are rampant in our society. Numerical simulations form the basis of computational sciences and high performance computing infrastructures have become scientific instruments with similar roles as those of test tubes or telescopes. Cloud infrastructures are used by companies in such an intense way that even the shortest outage quickly incurs the loss of several millions of dollars. But every citizen also relies on (and interacts with) such infrastructures via complex wireless mobile embedded devices whose nature is constantly evolving. In this way, the advent of digital miniaturization and interconnection has enabled our homes, power stations, cars and bikes to evolve into smart grids and smart transportation systems that should be optimized to fulfill societal expectations.
Our dependence and intense usage of such gigantic systems obviously leads to very high expectations in terms of performance. Indeed, we strive for lowcost and energyefficient systems that seamlessly adapt to changing environments that can only be accessed through uncertain measurements. Such digital systems also have to take into account both the users' profile and expectations to efficiently and fairly share resources in an online way. Analyzing, designing and provisioning such systems has thus become a real challenge.
Such systems are characterized by their evergrowing size, intrinsic heterogeneity and distributedness, userdriven requirements, and an unpredictable variability that renders them essentially stochastic. In such contexts, many of the former design and analysis hypotheses (homogeneity, limited hierarchy, omniscient view, optimization carried out by a single entity, openloop optimization, user outside of the picture) have become obsolete, which calls for radically new approaches. Properly studying such systems requires a drastic rethinking of fundamental aspects regarding the system's observation (measure, trace, methodology, design of experiments), analysis (modeling, simulation, trace analysis and visualization), and optimization (distributed, online, stochastic).
2.2 Objectives
The goal of the POLARIS project is to contribute to the understanding of the performance of very large scale distributed systems by applying ideas from diverse research fields and application domains. We believe that studying all these different aspects at once without restricting to specific systems is the key to push forward our understanding of such challenges and to propose innovative solutions. This is why we intend to investigate problems arising from application domains as varied as large computing systems, wireless networks, smart grids and transportation systems.
The members of the POLARIS project cover a very wide spectrum of expertise in performance evaluation and models, distributed optimization, and analysis of HPC middleware. Specifically, POLARIS' members have worked extensively on:

Experiment design:
Experimental methodology, measuring/monitoring/tracing tools, experiment control, design of experiments, and reproducible research, especially in the context of large computing infrastructures (such as computing grids, HPC, volunteer computing and embedded systems).

Trace Analysis:
Parallel application visualization (paje, triva/viva, framesoc/ocelotl, ...), characterization of failures in large distributed systems, visualization and analysis for geographical information systems, spatiotemporal analysis of media events in RSS flows from newspapers, and others.

Modeling and Simulation:
Emulation, discrete event simulation, perfect sampling, Markov chains, Monte Carlo methods, and others.

Optimization:
Stochastic approximation, mean field limits, game theory, discrete and continuous optimization, learning and information theory.
2.3 Contribution to AI/Learning
AI and Learning is everywhere now. Let us clarify how our research activities are positioned with respect to this trend.
A first line of research in POLARIS is devoted to the use of statistical learning techniques (Bayesian inference) to model the expected performance of distributed systems, to build aggregated performance views, to feed simulators of such systems, or to detect anomalous behaviours.
In a distributed context it is also essential to design systems that can seamlessly adapt to the workload and to the evolving behaviour of its components (users, resources, network). Obtaining faithful information on the dynamic of the system can be particularly difficult, which is why it is generally more efficient to design systems that dynamically learn the best actions to play through trial and errors. A key characteristic of the work in the POLARIS project is to leverage regularly gametheoretic modeling to handle situations where the resources or the decision is distributed among several agents or even situations where a centralised decision maker has to adapt to strategic users.
An important research direction in POLARIS is thus centered on reinforcement learning (Multiarmed bandits, Qlearning, online learning) and active learning in environments with one or several of the following features:
 Feedback is limited (e.g., gradient or even stochastic gradients are not available, which requires for example to resort to stochastic approximations);
 Multiagent setting where each agent learns, possibly not in a synchronised way (i.e., decisions may be taken asynchronously, which raises convergence issues);
 Delayed feedback (avoid oscillations and quantify convergence degradation);
 Non stochastic (e.g., adversarial) or non stationary workloads (e.g., in presence of shocks);
 Systems composed of a very large number of entities, that we study through mean field approximation (meanfield games and mean field control).
As a side effect, many of the gained insights can often be used to dramatically improve the scalability and the performance of the implementation of more standard machine or deep learning techniques over supercomputers.
The POLARIS members are thus particularly interested in the design and analysis of adaptive learning algorithms for multiagent systems, i.e. agents that seek to progressively improve their performance on a specific task. The resulting algorithms should not only learn an efficient (Nash) equilibrium but they should also be able of doing so quickly (low regret), even when facing the difficulties associated to a distributed context (lack of coordination, uncertain world, information delay, limited feedback, …)
In the rest of this document, we describe in detail our new results in the above areas.
3 Research program
3.1 Performance Evaluation
Participants: Jonatha Anselmi, Vincent Danjean, Nicolas Gast, Guillaume Huard, Arnaud Legrand, Florence Perronnin, JeanMarc Vincent.
Projectteam positioning
Evaluating the scalability, robustness, energy consumption and performance of large infrastructures such as exascale platforms and clouds raises severe methodological challenges. The complexity of such platforms mandates empirical evaluation but direct experimentation via an application deployment on a realworld testbed is often limited by the few platforms available at hand and is even sometimes impossible (cost, access, early stages of the infrastructure design, etc.). Furthermore, such experiments are costly, difficult to control and therefore difficult to reproduce. Although many of these digital systems have been built by human, they have reached such a complexity level that we are no longer able to study them like artificial systems and have to deal with the same kind of experimental issues as natural sciences. The development of a sound experimental methodology for the evaluation of resource management solutions is among the most important ways to cope with the growing complexity of computing environments. Although computing environments come with their own specific challenges, we believe such general observation problems should be addressed by borrowing good practices and techniques developed in many other domains of science, in particular (1) Predictive Simulation, (2) Trace Analysis and Visualization, and (3) the Design of Experiments.
Scientific achievements
Large computing systems are particularly complex to understand because of the interplay between their discrete nature (originating from deterministic computer programs) and their stochastic nature (emerging from the physical world, long distance interactions, and complex hardware and software stacks). A first line of research in POLARIS is devoted to the design of relatively simple statistical models of key components of distributed systems and their exploitation to feed simulators of such systems, to build aggregated performance views, and to detect anomalous behaviors.
Predictive Simulation
Unlike direct experimentation via an application deployment on a realworld testbed, simulation enables fully repeatable and configurable experiments that can often be conducted quickly for arbitrary hypothetical scenarios. In spite of these promises, current simulation practice is often not conducive to obtaining scientifically sound results. To date, most simulation results in the parallel and distributed computing literature are obtained with simulators that are ad hoc, unavailable, undocumented, and/or no longer maintained. As a result, most published simulation results build on throwaway (shortlived and non validated) simulators that are specifically designed for a particular study, which prevents other researchers from building upon it. There is thus a strong need for recognized simulation frameworks by which simulation results can be reproduced, further analyzed and improved.
Many simulators of MPI applications have been developed by renowned HPC groups (e.g., at SDSC 115, BSC 53, UIUC 123, Sandia Nat. Lab. 121, ORNL 54 or ETH Zürich 85) but most of them build on restrictive network and application modeling assumptions that generally prevent to faithfully predict execution times, which limits the use of simulation to indication of gross trends at best.
The SimGrid simulation toolkit, whose development started more than 20 years ago in UCSD, is a renowned project which gathers more than 1,700 citations and has supported the research of at least 550 articles. The most important contribution of POLARIS to this project in the last years has been to improve the quality of SimGrid to the point where it can be used effectively on a daily basis by practitioners to accurately reproduce the dynamic of real HPC systems. In particular, SMPI 61, a simulator based on SimGrid that simulates unmodified MPI applications written in C/C++ or FORTRAN, has now become a very unique tool allowing to faithfully study particularly complex scenario such as legacy Geophysics application that suffers from spatial and temporal load balancing problem 89, 88 or the HPL benchmark 603. We have shown that the performance (both for time and energy consumption 84) predicted through our simulations was systematically within a few percents of real experiments, which allows to reliably tune the applications at very low cost. This capacity has also been leveraged to study (through StarPUSimGrid) complex and modern taskbased applications running on heterogeneous sets of hybrid (CPUs + GPUs) nodes 102. The phenomenon studied through this approach would be particularly difficult to study through real experiments but yet allow to address real problems of these applications. Finally, SimGrid is also heavily used through BatSim, a batch simulator developed in the DATAMOVE team and which leverages SimGrid, to investigate the performance of machine learning strategies in a batch scheduling context 92, 16.
Trace Analysis and Visualization
Many monolithic visualization tools have been developed by renowned HPC groups since decades (e.g., BSC 106, Jülich and TU Dresden 101, 56, UIUC 83, 110, 87 and ANL 122) but most of these tools build on the classical information visualization 112 that consists in always first presenting an overview of the data, possibly by plotting everything if computing power allows, and then to allow users to zoom and filter, providing details on demand. However in our context, the amount of data comprised in such traces is several orders of magnitude larger than the number of pixels on a screen and displaying even a small fraction of the trace leads to harmful visualization artifacts. Such traces are typically made of events that occur at very different time and space scales and originate from different sources, which hinders classical approaches, especially when the application structure departs from classical MPI programs with a BSP/SPMD structure. In particular, modern HPC applications that build on a taskbased runtime and run on hybrid nodes are particularly challenging to analyze. Indeed, the underlying taskgraph is dynamically scheduled to avoid spurious synchronizations, which prevents classical visualizations to exploit and reveal the application structure.
In 68, we explain how modern data analytics tools can be used to build, from heterogeneous information sources, custom, reproducible and insightful visualizations of taskbased HPC applications at a very low development cost in the StarVZ framework. By specifying and validating statistical models of the performance of HPC applications/systems, we manage to identify when their behavior departs from what is expected and detect performance anomalies. This approach has first been applied to stateofthe art linear algebra libraries in 68 and more recently to a sparse direct solver 13. In both cases, we have been able to identify and fix several nontrivial anomalies that had not been noticed even by the application and runtime developers. Finally, these models not only allow to reveal when applications depart from what is expected but also to summarize the execution by focusing on the most important features, which is particularly useful when comparing two executions.
Design of Experiments and Reproducibility
Part of our work is devoted to the control of experiments on both classical (HPC) and novel (IoT/Fog in a smart home context) infrastructures. To this end, we heavily rely on experimental testbeds such as Grid5000 and FITIoTLab that can be wellcontrolled but real experiments are nonetheless quite resourceconsuming. Design of experiments has been successfully applied in many fields (e.g., agriculture, chemistry, industrial processes) where experiments are considered expensive. Building on concrete use cases, we explore how Design of Experiments and Reproducible Research techniques can be used to (1) design transparent autotuning strategies of scientific computation kernels 55, 111 (2) set up systematic performance non regression tests on Grid5000 (450 nodes for 1.5 year) and detect many abnormal events (related to bios and system upgrades, cooling, faulty memory and power instability) that had a significant effect on the nodes, from subtle performance changes of 1% to much more severe degradation of more than 10%, and had yet been unnoticed by both Grid’5000 technical team and Grid’5000 users (3) design and evaluate the performance of service provisioning strategies 462 in Fog infrastructures.
3.2 Asymptotic Methods
Participants: Jonatha Anselmi, Romain Couillet, Nicolas Gast, Bruno Gaujal, Florence Perronnin, JeanMarc Vincent.
Projectteam positioning
Stochastic models often suffer from the curse of dimensionality: their complexity grows exponentially with the number of dimensions of the system. At the same time, very large stochastic systems are sometimes easier to analyze: it can be shown that some classes of stochastic systems simplify as their dimension goes to infinity because of averaging effects such as the law of large numbers, or the central limit theorem. This forms the basis of what is called an asymptotic method, which consists in studying what happens when a system gets large in order to build an approximation that is easier to study or to simulate.
Within the team, the research that we conduct in this axis is to foster the applicability of these asymptotic methods to new application areas. This leads us to work on the application of classical methods to new problems, but also to develop new approximation methods that take into account special features of the systems we study (i.e., moderate number of dimensions, transient behavior, random matrices). Typical applications are mean field method for performance evaluation, application to distributed optimization, and more recently statistical learning. One originality of our work is to quantify precisely what is the error made by such approximations. This allows us to define refinement terms that lead to more accurate approximations.
Scientific achievements
Refined mean field approximation
Mean field approximation is a wellknown technique in statistical physics, that was originally introduced to study systems composed of a very large number of particles (say $n>{10}^{20}$). The idea of this approximation is to assume that objects are independent and only interact between them through an average environment (the mean field). Nowadays, variants of this technique are widely applied in many domains: in game theory for instance (with the example of mean field games), but also to quantify the performance of distributed algorithms. Mean field approximation is often justified by showing that a system of $n$ wellmixed interacting objects converges to its deterministic mean field approximation as $n$ goes to infinity. Yet, this does not explain why mean field approximation provides a very accurate approximation of the behavior of systems composed by a few hundreds of objects or less. Until recently, this was essentially an open question.
In 70, we give a partial answer to this question. We show that, for most of the mean field models used for performance evaluation, the error made when using a mean field approximation is a $\Theta (1/n)$. This results greatly improved compared to previous work that showed that the error made by mean field approximation was smaller than $O(1/\sqrt{n})$. On the contrary, we obtain the exact rate of accuracy. This result came from the use of Stein's method that allows one to quantify precisely the distance between two stochastic processes. Subsequently, in 72, we show that the constant in the $\Theta (1/n)$ can be computed numerically by a very efficient algorithm. By using this, we define the notion of refined approximation which consists in adding the $1/n$correction term. This methods can also be generalize to higher order extensions or 74, 69.
Design and analysis of distributed control algorithms
Mean field approximation is widely used in the performance evaluation community to analyze and design distributed control algorithms. Our contribution in this domain has covered mainly two applications: cache replacement algorithms and load balancing algorithms.
Cache replacement algorithms are widely used in content delivery networks. In 58, 76, 75, we show how mean field and refined mean field approximation can be used to evaluate the performance of listbased cache replacement algorithms. In particular, we show that such policies can outperform the classically used LRU algorithm. A methodological contribution of our work is that, when evaluating precisely the behavior of such a policy, the refined mean field approximation is both faster and more accurate than what could be obtained with a stochastic simulator.
Computing resources are often spread across many machines. An efficient use of such resources requires the design of a good load balancing strategy, to distribute the load among the available machines. In 51, 52, 50, we study two paradigms that we use to design asymptotically optimal load balancing policies where a central broker sends tasks to a set of parallel servers. We show in 51, 50 that combining the classical roundrobin allocation plus an evaluation of the tasks sizes can yield a policy that has a zero delay in the large system limit. This policy is interesting because the broker does not need any feedback from the servers. At the same time, this policy needs to estimate or know job durations, which is not always possible. A different approach is used in 52 where we consider a policy that does not need to estimate job durations but that uses some feedback from the servers plus a memory of where jobs where send. We show that this paradigm can also be used to design zerodelay load balancing policies as the system size grows to infinity.
Mean field games
Various notions of mean field games have been introduced in the years 20002010 in theoretical economics, engineering or game theory. A mean field game is a game in which an individual tries to maximize its utility while evolving in a population of other individuals whose behavior are not directly affected by the individual. An equilibrium is a population dynamics for which a selfish individual would behave as the population. In 64, we develop the notion of discrete space mean field games, that is more amenable to study than the previously introduced notions of mean field games. This leads to two interesting contributions: mean field games are not always the limits of stochastic games as the number of players grow 63, mean field games can be used to study how much vaccination should be subsidized to encourage people to adapt a socially optimal behaviour 77.
3.3 Distributed Online Optimization and Learning in Games
Participants: Nicolas Gast, Romain Couillet, Bruno Gaujal, Arnaud Legrand, Patrick Loiseau, Panayotis Mertikopoulos, Bary Pradelski.
Projectteam positioning
Online learning concerns the study of repeated decisionmaking in changing environments. Of course, depending on the context, the words “learning” and “decisionmaking” may refer to very different things: in economics, this could mean predicting how rational agents react to market drifts; in data networks, it could mean adapting the way packets are routed based on changing traffic conditions; in machine learning and AI applications, it could mean training a neural network or the guidance system of a selfdriving car; etc. In particular, the changes in the learner's environment could be either exogenous (that is, independent of the learner's decisions, such as the weather affecting the time of travel), or endogenous (i.e., they could depend on the learner's decisions, as in a game of poker), or any combination thereof. However, the goal for the learner(s) is always the same: to make more informed decisions that lead to better rewards over time.
The study of online learning models and algorithms dates back to the seminal work of Robbins, Nash and Bellman in the 50's, and it has since given rise to a vigorous research field at the interface of game theory, control and optimization, with numerous applications in operations research, machine learning, and data science. In this general context, our team focuses on the asymptotic behavior of online learning and optimization algorithms, both single and multiagent: whether they converge, at what speed, and/or what type of nonstationary, offequilibrium behaviors may arise when they do not.
The focus of POLARIS on gametheoretic and Markovian models of learning covers a set of specific challenges that dovetail in a highly synergistic manner with the work of other learningoriented teams within Inria (like SCOOL in Lille, SIERRA in Paris, and THOTH in Grenoble), and it is an important component of Inria's activities and contributions in the field (which includes major industrial stakeholders like Google / DeepMind, Facebook, Microsoft, Amazon, and many others).
Scientific achievements
Our team's work on online learning covers both single and multiagent models; in the sequel, we present some highlights of our work structured along these basic axes.
In the singleagent setting, an important problem in the theory of Markov decision processes – i.e., discretetime control processes with decisiondependent randomness – is the socalled “restless bandit” problem. Here, the learner chooses an action – or “arm” – from a finite set, and the mechanism determining the action's reward changes depending on whether the action was chosen or not (in contrast to standard Markov problems where the activation of an arm does not have this effect). In this general setting, Whittle conjectured – and Weber and Weiss proved – that Whittle's eponymous index policy is asymptotically optimal. However, the result of Weber and Weiss is purely asymptotic, and the rate of this convergence remained elusive for several decades. This gap was finally settled in a series of POLARIS papers 7142, where we showed that Whittle indices (as well as other index policies) become optimal at a geometric rate under the same technical conditions used by Weber and Weiss to prove Whittle's conjecture, plus a technical requirement on the nonsingularity of the fixed point of the meanfield dynamics. We also propose the first subcubic algorithm to compute Whittle and Gittins indexes. As for reinforcement learning in Markovian bandits, we have shown that Bayesian and optimistic approaches do not use the structure of Markovian bandits similarly: While Bayesian learning has both a regret and a computational complexity that scales linearly with the number of arms, optimistic approaches all incur an exponential computation time, at least in their current versions 40.
In the multiagent setting, our work has focused on the following fundamental question:
Does the concurrent use of (possibly optimal) singleagent learning algorithms
ensure convergence to Nash equilibrium in multiagent, gametheoretic environments?
Conventional wisdom might suggest a positive answer to this question because of the following “folk theorem”: under noregret learning, the agents' empirical frequency of play converges to the game's set of coarse correlated equilibria. However, the actual implications of this result are quite weak: First, it concerns the empirical frequency of play and not the daytoday sequence of actions employed by the players. Second, it concerns coarse correlated equilibria which may be supported on strictly dominated strategies – and are thus unacceptable in terms of rationalizability. These realizations prompted us to make a clean break with conventional wisdom on this topic, ultimately showing that the answer to the above question is, in general, “no”: specifically, 97, 95 showed that the (optimal) class of “followtheregularizedleader” (FTRL) learning algorithms leads to Poincaré recurrence even in simple, $2\times 2$ minmax games, thus precluding convergence to Nash equilibrium in this context.
This negative result generated significant interest in the literature as it contributed in shifting the focus towards identifying which Nash equilibria may arise as stable limit points of FTRL algorithms and dynamics. Earlier work by POLARIS on the topic 57, 98, 99 suggested that strict Nash equilibria play an important role in this question. This suspicion was recently confirmed in a series of papers 67, 82 where we established a sweeping negative result to the effect that mixed Nash equilibria are incompatible with noregret learning. Specifically, we showed that any Nash equilibrium which is not strict cannot be stable and attracting under the dynamics of FTRL, especially in the presence of randomness and uncertainty. This result has significant implications for predicting the outcome of a multiagent learning process because, combined with 98, it establishes the following farreaching equivalence: a state is asymptotically stable under noregret learning if and only if it is a strict Nash equilibrium.
Going beyond finite games, this further raised the question of what type of nonconvergent behaviors can be observed in continuous games – such as the class of stochastic minmax problems that are typically associated to generative adversarial networks (GANs) in machine learning. This question was one of our primary collaboration axes with EPFL, and led to a joint research project focused on the characterization of the convergence properties of zeroth, first, and (scalable) secondorder methods in nonconvex/nonconcave problems. In particular, we showed in 86 that these stateoftheart minmax optimization algorithms may converge with arbitrarily high probability to attractors that are in no way minmax optimal or even stationary – and, in fact, may not even contain a single stationary point (let alone a Nash equilibrium). Spurious convergence phenomena of this type can arise even in twodimensional problems, a fact which corroborates the empirical evidence surrounding the formidable difficulty of training GANs.
3.4 Responsible Computer Science
Participants: Nicolas Gast, Romain Couillet, Bruno Gaujal, Arnaud Legrand, Patrick Loiseau, Panayotis Mertikopoulos, Bary Pradelski.
Projectteam positioning
The topics in this axis emerge from current social and economic questions rather than from a fixed set of mathematical methods. To this end we have identified large trends such as energy efficiency, fairness, privacy, and the growing number of new market places. In addition, COVID has posed new questions that opened new paths of research with strong links to policy making.
Throughout these works, the focus of the team is on modeling aspects of the aforementioned problems, and obtaining strong theoretical results that can give highlevel guidelines on the design of markets or of decisionmaking procedures. Where relevant, we complement those works by measurement studies and audits of existing systems that allow identifying key issues. As this work is driven by topics, rather than methods, it allows for a wide range of collaborations, including with enterprises (e.g., Naverlabs), policy makers, and academics from various fields (economics, policy, epidemiology, etc.).
Other teams at Inria cover some of the societal challenges listed here (e.g., PRIVATICS, COMETE) but rather in isolation. The specificity of POLARIS resides in the breadth of societal topics covered and of the collaborations with nonCS researchers and nonresearch bodies; as well as in the application of methods such as game theory to those topics.
Scientific achievements
Algorithmic fairness
As algorithmic decisionmaking became increasingly omnipresent in our daily lives (in domains ranging from credits to advertising, hiring, or medicine); it also became increasingly apparent that the outcome of algorithms can be discriminatory for various reasons. Since 2016, the scientific community working on the problem of algorithmic fairness has been exponentially increasing. In this context, in the early days, we worked on better understanding the extent of the problem through measurement in the case of social networks 114. In particular, in this work, we showed that in advertising platforms, discrimination can occur from multiple different internal processes that cannot be controlled, and we advocate for measuring discrimination on the outcome directly. Then we worked on proposing solutions to guarantee fair representation in online public recommendations (aka trending topics on Twitter) 59. This is an example of an application in which it was observed that recommendations are typically biased towards some demographic groups. In this work, our proposed solution draws an analogy between recommendation and voting and builds on existing works on fair representation in voting. Finally, in most recent times, we worked on better understanding the sources of discrimination, in the particular simple case of selection problems, and the consequences of fixing it. While most works attribute discrimination to implicit bias of the decision maker 91, we identified a fundamentally different source of discrimination: Even in the absence of implicit bias in a decision maker’s estimate of candidates’ quality, the estimates may differ between the different groups in their variance—that is, the decision maker’s ability to precisely estimate a candidate’s quality may depend on the candidate’s group 66. We show that this differential variance leads to discrimination for two reasonable baseline decision makers (groupoblivious and Bayesian optimal). Then we analyze the consequence on the selection utility of imposing fairness mechanisms such as demographic parity or its generalization; in particular we identify some cases for which imposing fairness can improve utility. In 65, we also study similar questions in the twostage setting, and derive the optimal selector and the “price of local fairness’’ one pays in utility by imposing that the interim stage be fair.
Privacy and transparency in social computing system
Online services in general, and social networks in particular, collect massive amounts of data about their users (both online and offline). It is critical that (i) the users’ data is protected so that it cannot leak and (ii) users can know what data the service has about them and understand how it is used—this is the transparency requirement. In this context, we did two kinds of work. First, we studied social networks through measurement, in particular using the use case of Facebook. We showed that their advertising platform, through the PII1based targeting option, allowed attackers to discover some personal data of users 116. We also proposed an alternative design—valid for any system that proposed PIIbased targeting—and proved that it fixes the problem. We then audited the transparency mechanisms of the Facebook ad platform, specifically the “Ad Preferences’’ page that shows what interests the platform inferred about a user, and the “Why am I seeing this’’ button that gives some reasons why the user saw a particular ad. In both cases, we laid the foundation for defining the quality of explanations and we showed that the explanations given were lacking key desirable properties (they were incomplete and misleading, they have since been changed) 49. A followup work shed further light on the typical uses of the platform 48. In another work, we proposed an innovative protocol based on randomized withdrawal to protect public posts deletion privacy 100. Finally, in 73, we study an alternative data sharing ecosystem where users can choose the precision of the data they give. We model it as a game and show that, if users are motivated to reveal data by a public good component of the outcome’s precision, then certain basic statistical properties (the optimality of generalized least squares in particular) no longer hold.
Online markets
Market design operates at the intersection of computer science and economics and has become increasingly important as many markets are redesigned on digital platforms. Studying markets for commodities, in an ongoing project we evaluate how different fee models alter strategic incentives for both buyers and sellers. We identify two general classes of fees: for one, strategic manipulation becomes infeasible as the market grows large and agents therefore have no incentive to misreport their true valuation. On the other hand, strategic manipulation is possible and we show that in this case agents aim to maximally shade their bids. This has immediate implications for the design of such markets. By contrast, 96 considers a matching market where buyers and sellers have heterogeneous preferences over each other. Traders arrive at random to the market and the market maker, having limited information, aims to optimize when to open the market for a clearing event to take place. There is a tradeoff between thickening the market (to achieve better matches) and matching quickly (to reduce waiting time of traders in the market). The tradeoff is made explicit for a wide range of underlying preferences. These works are adding to an ongoing effort to better understand and design markets 10793.
COVID
The COVID19 pandemic has put humanity to one of the defining challenges of its generation and as such naturally transdisciplinary efforts have been necessary to support decision making. In a series of articles 109105 we proposed Green Zoning. `Green zones’–areas where the virus is under control based on a uniform set of conditions–can progressively return to normal economic and social activity levels, and mobility between them is permitted. By contrast, stricter public health measures are in place in ‘red zones’, and mobility between red and green zones is restricted. France and Spain were among the first countries to introduce green zoning in April 2020. The initial success of this proposal opened up the way to a large amount of followup work analyzing and proposing various tools to effectively deploy different tools to combat the pandemic (e.g., focusmass testing 108 and a vaccination policy 103). In a joint work with a group of leading economists, public health researchers and sociologists it was found that countries that opted to aim to eliminate the virus fared better not only for public health, but also for the economy and civil liberties 104. Overall this work has been characterized by close interactions with policy makers in France, Spain and the European Commission as well as substantial activity in public discourse (via TV, newspapers and radio).
Energy efficiency
Our work on energy efficiency spanned multiple different areas and applications such as embedded systems and smart grids. Minimizing the energy consumption of embedded systems with realtime constraints is becoming more important for ecological as well as practical reasons since batteries are becoming standard power supplies. Dynamically changing the speed of the processor is the most common and efficient way to reduce energy consumption 113. In fact, this is the reason why modern processors are equipped with Dynamic Voltage and Frequency Scaling (DVFS) technology 120. In a stochastic environment, with random job sizes and arrival times, combining hard deadlines and energy minimization via DVFSbased techniques is difficult because forcing hard deadlines requires considering the worst cases, hardly compatible with random dynamics. Nevertheless, progress have been made to solve these types of problems in a series of papers using constrained Markov decision processes, both on the theoretical side (proving existence of optimal policies and showing their structure 80, 78, 79) as well as on the experimental side (showing the gains of optimal policies over classical solutions 81).
In the context of a collaboration with Enedis and Schneider Electric (via the Smart Grid chair of GrenobleINP), we also study the problem of using smart meters to optimize the behavior of electrical distribution networks. We made three kinds of contributions on this subject: (1) how to design efficient control strategies in such a system 117, 119, 118, (2) how to cosimulate an electrical network and a communication network 90, and (3) what is the performance of the communication protocol (PLC G3) used by the Linky smart meters 94.
4 Application domains
4.1 Large Computing Infrastructures
Supercomputers typically comprise thousands to millions of multicore CPUs with GPU accelerators interconnected by complex interconnection networks that are typically structured as an intricate hierarchy of network switches. Capacity planning and management of such systems not only raises challenges in term of computing efficiency but also in term of energy consumption. Most legacy (SPMD) applications struggle to benefit from such infrastructure since the slightest failure or load imbalance immediately causes the whole program to stop or at best to waste resources. To scale and handle the stochastic nature of resources, these applications have to rely on dynamic runtimes that schedule computations and communications in an opportunistic way. Such evolution raises challenges not only in terms of programming but also in terms of observation (complexity and dynamicity prevents experiment reproducibility, intrusiveness hinders large scale data collection, ...) and analysis (dynamic and flexible application structures make classical visualization and simulation techniques totally ineffective and require to build on ad hoc information on the application structure).
4.2 NextGeneration Wireless Networks
Considerable interest has arisen from the seminal prediction that the use of multipleinput, multipleoutput (MIMO) technologies can lead to substantial gains in information throughput in wireless communications, especially when used at a massive level. In particular, by employing multiple inexpensive service antennas, it is possible to exploit spatial multiplexing in the transmission and reception of radio signals, the only physical limit being the number of antennas that can be deployed on a portable device. As a result, the wireless medium can accommodate greater volumes of data traffic without requiring the reallocation (and subsequent reregulation) of additional frequency bands. In this context, throughput maximization in the presence of interference by neighboring transmitters leads to games with convex action sets (covariance matrices with trace constraints) and individually concave utility functions (each user's Shannon throughput); developing efficient and distributed optimization protocols for such systems is one of the core objectives of the research theme presented in Section 3.3.
Another major challenge that occurs here is due to the fact that the efficient physical layer optimization of wireless networks relies on perfect (or close to perfect) channel state information (CSI), on both the uplink and the downlink. Due to the vastly increased computational overhead of this feedback – especially in decentralized, smallcell environments – the continued transition to fifth generation (5G) wireless networks is expected to go handinhand with distributed learning and optimization methods that can operate reliably in feedbackstarved environments. Accordingly, one of POLARIS' applicationdriven goals will be to leverage the algorithmic output of Theme 5 into a highly adaptive resource allocation framework for nextgéneration wireless systems that can effectively "learn in the dark", without requiring crippling amounts of feedback.
4.3 Energy and Transportation
Smart urban transport systems and smart grids are two examples of collective adaptive systems. They consist of a large number of heterogeneous entities with decentralised control and varying degrees of complex autonomous behaviour. We develop an analysis tool to help to reason about such systems. Our work relies on tools from fluid and meanfield approximation to build decentralized algorithms that solve complex optimization problems. We focus on two problems: decentralized control of electric grids and capacity planning in vehiclesharing systems to improve load balancing.
4.4 Social Computing Systems
Social computing systems are online digital systems that use personal data of their users at their core to deliver personalized services directly to the users. They are omnipresent and include for instance recommendation systems, social networks, online medias, daily apps, etc. Despite their interest and utility for users, these systems pose critical challenges of privacy, security, transparency, and respect of certain ethical constraints such as fairness. Solving these challenges involves a mix of measurement and/or audit to understand and assess issues, and modeling and optimization to propose and calibrate solutions.
5 Social and environmental responsibility
5.1 Footprint of research activities
The carbon footprint of the team has been quite minimal in 2021 since there has been no travel allowed with most of us working from home. Our team does not train heavy ML models requiring important processing power although some of us perform computer science experiments, mostly using the Grid5000 platforms. We keep this usage very reasonable and rely on cheaper alternatives (e.g., simulations) as much as possible.
5.2 Raising awareness on the climate crisis
Romain Couillet has organized several introductory seminars on the Anthropocene, which he has presented to students at the UGA and GrenobleINP, as well as to associations in Grenoble (FNE, AgirAlternatif). He is also coresponsible of the Digital Transformation DU. He has published three articles on the issue of "usability" of artificial intelligence, and is the organizer of a special session on "Signal processing and resilience" for the GRETSI 2022 conference. He is also cocreator of the sustainable AI transversal axis of the MIAI project in Grenoble. He connects his professionnal activity with public action (Lowtechlab de Grenoble, Université Autogérée, Arche des Innovateurs, etc). Finally, he is a trainer for the "Fresque du Climat" and a member of Adrastia and FNE Isère. See section 11.2 and 11.3.3 for more details.
5.3 Impact of research results
The efforts of Barry Pradelski on COVID policy have received lots of media coverage and he was appointed to share his expertise to the Centre européen de prévention et de contrôle des maladies (ECDC).
JeanMarc Vincent is heavily engaged since several years in the training of computer science teachers at the elementary/middle/high school levels 47. Among one of his many activities, we can mention his involvement in the design of the Numérique et Sciences Informatiques, NSI : les fondamentaux MOOC. See section 11.2 and 11.3.3 for more details.
6 Highlights of the year
6.1 Awards
 The paper "Energy optimal activation of processors for the execution of a single task with unknown size" 17, authored by J. Anselmi and B. Gaujal, won the best paper award at the international conference IEEE MASCOTS 2022.
 The paper "Robust power management via learning and game design" 124, by Z. Zhou, P. Mertikopoulos, A. L. Moustakas, N. Bambos, and P. W. Glynn, published in Operations Research, 69(1):331–345, in January 2021 won the INFORMS best paper award in the network analytics section.
 The paper "UnderGrad: A universal blackbox optimization method with almost dimensionfree convergence rate guarantees" 19, authored by K. Antonakopoulos, D. Q. Vu, V. Cevher, K. Levy, and P. Mertikopoulos, won the best paper award at the ICML'22 conference.
 Panayotis Mertikopoulos was recognized as an Outstanding reviewer at ICLR 2022.
 The book "Random matrix methods for machine learning" 35 by R. Couillet and Z. Liao has been published at Cambridge University Press in June 2022.
7 New software and platforms
7.1 New software
7.1.1 SimGrid

Keywords:
Largescale Emulators, Grid Computing, Distributed Applications

Scientific Description:
SimGrid is a toolkit that provides core functionalities for the simulation of distributed applications in heterogeneous distributed environments. The simulation engine uses algorithmic and implementation techniques toward the fast simulation of large systems on a single machine. The models are theoretically grounded and experimentally validated. The results are reproducible, enabling better scientific practices.
Its models of networks, cpus and disks are adapted to (Data)Grids, P2P, Clouds, Clusters and HPC, allowing multidomain studies. It can be used either to simulate algorithms and prototypes of applications, or to emulate real MPI applications through the virtualization of their communication, or to formally assess algorithms and applications that can run in the framework.
The formal verification module explores all possible message interleavings in the application, searching for states violating the provided properties. We recently added the ability to assess liveness properties over arbitrary and legacy codes, thanks to a systemlevel introspection tool that provides a finely detailed view of the running application to the model checker. This can for example be leveraged to verify both safety or liveness properties, on arbitrary MPI code written in C/C++/Fortran.

Functional Description:
SimGrid is a simulation toolkit that provides core functionalities for the simulation of distributed applications in large scale heterogeneous distributed environments.

News of the Year:
There were 3 major releases in 2022. The SimDag API for the simulation of the scheduling of Directed Acyclic Graphs has been dropped and replaced by the SimDag++ API which provides the different features of SimDag directly on top of the S4U API. We also dropped the old and clumsy Lua bindings to create platforms in a programmatic way. It can be done in C++ in a much cleaner way now, which motivates this suppression. The C++ platform description has been improved to reject forbidden topologies, improve exporting for visualization, and allow users to dynamically change injected costs for MPI_* operations. The Python API to S4U has been extended. A new solver for parallel task (BMF) has been introduced and provides with more realistic sharing of heterogeneous resources compared to the fair bottleneck solver used by ptask_L07. Although this is still ongoing work, this paves the way to efficient macroscopic modeling of streaming activities and parallel applications. The internals of the Model Checker have been heavily reworked and new test suites from the MPI Bugs Initiative (MBI) are now used. The documentation was thoroughly overhauled to ease the use of the framework. We also pursued our efforts to improve the overall framework, through bug fixes, code refactoring and other software quality improvements.
 URL:

Contact:
Martin Quinson

Participants:
Adrien Lebre, AnneCécile Orgerie, Arnaud Legrand, Augustin Degomme, Arnaud Giersch, Emmanuelle Saillard, Frédéric Suter, Jonathan Pastor, Martin Quinson, Samuel Thibault

Partners:
CNRS, ENS Rennes
7.1.2 PSI

Name:
Perfect Simulator

Keywords:
Markov model, Simulation

Functional Description:
Perfect simulator is a simulation software of markovian models. It is able to simulate discrete and continuous time models to provide a perfect sampling of the stationary distribution or directly a sampling a functional of this distribution by using coupling from the past. The simulation kernel is based on the CFTP algorithm, and the internal simulation of transitions on the Aliasing method.

News of the Year:
No active development. Maintenance is ensured by the POLARIS team. The next generation of PSI lies in the MARTO project.
 URL:

Contact:
JeanMarc Vincent
7.1.3 marmoteCore

Name:
Markov Modeling Tools and Environments  the Core

Keywords:
Modeling, Stochastic models, Markov model

Functional Description:
marmoteCore is a C++ environment for modeling with Markov chains. It consists in a reduced set of highlevel abstractions for constructing state spaces, transition structures and Markov chains (discretetime and continuoustime). It provides the ability of constructing hierarchies of Markov models, from the most general to the particular, and equip each level with specifically optimized solution methods.
This software was started within the ANR MARMOTE project: ANR12MONU00019.

News of the Year:
No active development. Current development lies now in the MARTO project (next generations of PSI and marmoteCore) and in the forthcoming Marmote project.
 URL:
 Publications:

Contact:
Alain JeanMarie

Participants:
Alain JeanMarie, Hlib Mykhailenko, Benjamin Briot, Franck Quessette, Issam Rabhi, JeanMarc Vincent, JeanMichel Fourneau

Partners:
Université de Versailles StQuentinenYvelines, Université Paris Nanterre
7.1.4 MarTO

Name:
Markov Toolkit for Markov models simulation: perfect sampling and Monte Carlo simulation

Keywords:
Perfect sampling, Markov model

Functional Description:
MarTO is a simulation software of markovian models. It is able to simulate discrete and continuous time models to provide a perfect sampling of the stationary distribution or directly a sampling of functional of this distribution by using coupling from the past. The simulation kernel is based on the CFTP algorithm, and the internal simulation of transitions on the Aliasing method. This software is a rewrite, more efficient and flexible, of PSI

News of the Year:
No official release yet. The code development is in progress.
 URL:

Contact:
Vincent Danjean
7.1.5 GameSeer

Keyword:
Game theory

Functional Description:
GameSeer is a tool for students and researchers in game theory that uses Mathematica to generate phase portraits for normal form games under a variety of (usercustomizable) evolutionary dynamics. The whole point behind GameSeer is to provide a dynamic graphical interface that allows the user to employ Mathematica's vast numerical capabilities from a simple and intuitive frontend. So, even if you've never used Mathematica before, you should be able to generate fully editable and customizable portraits quickly and painlessly.

News of the Year:
No new release but the development is still active.
 URL:

Contact:
Panayotis Mertikopoulos
7.1.6 rmf_tool

Name:
A library to Compute (Refined) Mean Field Approximations

Keyword:
Mean Field

Functional Description:
The tool accepts three model types:
 homogeneous population processes (HomPP)
 density dependent population processes (DDPPs)
 heterogeneous population models (HetPP)
In particular, it provides a numerical algorithm to compute the constant of the refined mean field approximation provided in the paper "A Refined Mean Field Approximation" by N. Gast and B. Van Houdt, SIGMETRICS 2018, and a framework to compute heterogeneous mean field approximations as proposed in "Mean Field and Refined Mean Field Approximations for Heterogeneous Systems: It Works!" by N. Gast and S. Allmeier, SIGMETRICS 2022.
 URL:
 Publications:

Contact:
Nicolas Gast
8 New results
The new results produced by the team in 2022 can be grouped into the following categories.
8.1 Performance evaluation of large systems
Participants: Sebastian Allmeier, Vincent Danjean, Arnaud Legrand, Nicolas Gast, Guillaume Huard, JeanMarc Vincent.
Finely tuning applications and understanding the influence of key parameters (number of processes, granularity, collective operation algorithms, virtual topology, and process placement) is critical to obtain good performance on supercomputers. With the high consumption of running applications at scale, doing so solely to optimize their performance is particularly costly. We have shown in 3 that SimGrid and SMPI simgrid.org could be used to obtain inexpensive but faithful predictions of expected performance. The methodology we propose decouples the complexity of the platform, which is captured through statistical models of the performance of its main components (MPI communications, BLAS operations), from the complexity of adaptive applications by emulating the application and skipping regular nonMPI parts of the code. We demonstrate the capability of our method with HighPerformance Linpack (HPL), the benchmark used to rank supercomputers in the TOP500, which requires careful tuning. This work presents an extensive (in)validation study that compares simulation with real experiments and demonstrates our ability to predict the performance of HPL within a few percent consistently. This study allows us to identify the main modeling pitfalls (e.g., spatial and temporal node variability or network heterogeneity and irregular behavior) that need to be considered. Our “surrogate” also allows studying several subtle HPL parameter optimization problems while accounting for uncertainty on the platform.
We have also shown in 13 how the structure of complex applications such as a multifrontal sparse linear solvers could be exploited to detect and correct nontrivial performance problems. Efficiently exploiting computational resources in heterogeneous platforms is a real challenge which has motivated the adoption of the taskbased programming paradigm where resource usage is dynamic and adaptive. Unfortunately, classical performance visualization techniques used in routine performance analysis often fail to provide any insight in this new context, especially when the application structure is irregular. We propose and implement in StarVZ several performance visualization techniques tailored for the analysis of taskbased multifrontal sparse linear solvers and show that by building on both a performance model of irregular tasks and on structure of the application (in particular the elimination tree), we can detect and highlight anomalies and understand resource utilization from the application pointofview in a very insightful way. We validate these novel performance analysis techniques with the QR_mumps sparse parallel solver by describing a series of case studies where we identify and address non trivial performance issues thanks to our visualization methodology.
Large systems can be particularly difficult to analyze because of inherent statespace explosion and most computations become untractable. Mean field approximation is a powerful technique to study the performance of very large stochastic systems represented as systems of interacting objects. Applications include load balancing models, epidemic spreading, cache replacement policies, or largescale data centers, for which mean field approximation gives very accurate estimates of the transient or steadystate behaviors. In a series of recent papers, a new and more accurate approximation, called the refined mean field approximation has been presented. A key strength of this technique lies in its applicability to notsolarge systems. Yet, computing this new approximation can be cumbersome, which is why develop a tool, called rmf tool and available at github.com/ngast/rmf_tool, that takes the description of a mean field model, and can numerically compute its mean field approximations and refinement.
Mean field approximation is asymptotically exact for systems composed of $n$ homogeneous objects under mild conditions. In 1, we study what happens when objects are heterogeneous. This can represent servers with different speeds or contents with different popularities. We define an interaction model that allows obtaining asymptotic convergence results for stochastic systems with heterogeneous object behavior, and show that the error of the mean field approximation is of order $O(1/n)$. More importantly, we show how to adapt the refined mean field approximation and show that the error of this approximation is reduced to $O(1/{n}^{2})$. To illustrate the applicability of our result, we present two examples. The first addresses a listbased cache replacement model, $RANDOM\left(m\right)$, which is an extension of the $RANDOM$ policy. The second is a heterogeneous supermarket model. These examples show that the proposed approximations are computationally tractable and very accurate. They also show that for moderate system sizes (30) the refined mean field approximation tends to be more accurate than simulations for any reasonable simulation time.
8.2 Energy optimization
Participants: Jonatha Anselmi, Bruno Gaujal, LouisSébastien Rebuffi.
A key objective in the management of modern computer systems consists in minimizing the electrical energy consumed by processing resources while satisfying certain target performance criteria. In 17, we consider the execution of a single task with unknown size on top of a service system that offers a limited number of processing speeds, say $N$, and investigate the problem of finding a speed profile that minimizes the resulting energy consumption subject to a deadline constraint. Existing works mainly investigated this problem when speed profiles are continuous functions. In contrast, the novelty of our work is to consider discontinuous speed profiles, i.e., a case that arises naturally when the underlying computational platform offers a finite number of speeds. In our main result, we show that the computation of an optimal speed profile boils down to solving a convex optimization problem. Under mild assumptions, for such convex optimization we prove some structural results that yield the formulation of an extremely efficient solution algorithm. Specifically, we show that the optimal speed profile can be computed by solving $O(logN)$ onedimensional equations. Our results hold when the task size follows a known probability distribution function and the set of available speeds, if listed in increasing order, forms a sublinear concave sequence.
More generally, energy optimization should be performed at a global scale and requires to revisit load balancing strategies. Techniques like replication and speculation are doubleedged weapons that must be handled with caution as the resource overhead may be detrimental when used too aggressively. We have studied such strategies in previous work and presented an overview 2 in the special issue of Queueing Systems, 100 views on queues.
8.3 Large system performance optimization through learning techniques
Participants: Arnaud Legrand, Lucas Leandro Nesi, Panayotis Mertikopoulos.
Large infrastructures and computing applications typically exhibit some form of regularity which should be exploited but their stochastic nature makes their optimization difficult. In this series of work, we demonstrate that simple machine and reinforcement learning techniques can be tailored to optimize these systems.
Parallel applications performance strongly depends on the number of resources. Although adding new nodes usually reduces execution time, excessive amounts are often detrimental as they incur substantial communication overhead, which is difficult to anticipate. Characteristics like network contention, data distribution methods, synchronizations, and how communications and computations overlap generally impact the performance. Finding the correct number of resources can thus be particularly tricky for multiphase applications as each phase may have very different needs, and the popularization of hybrid (CPU+GPU) machines and heterogeneous partitions makes it even more difficult. In 32, we study and propose, in the context of a taskbased GeoStatistic application, strategies for the application to actively learn and adapt to the best set of heterogeneous nodes it has access to. We propose strategies that use the Gaussian Process method with trends, bound mechanisms for reducing the search space, and heterogeneous behavior modeling. We compare these methods with traditional exploration strategies in 16 different machines scenarios. In the end, the proposed strategies are able to gain up to $\approx 51\%$ compared to the standard case of using all the nodes while having low overhead.
At the scale of the whole highperformance computing platform, job scheduling is also a hard problem that involves uncertainties on both the job arrival process and their execution times. Users typically provide only loose upper bounds for job execution times, which are not so useful for scheduling heuristics based on processing times. Previous studies focused on applying regression techniques to obtain better execution time estimates, which worked reasonably well and improved scheduling metrics. However, these approaches require a long period of training data. In 16, we propose a simpler approach by classifying jobs as small or large and prioritizing the execution of small jobs over large ones. Indeed, small jobs are the most impacted by queuing delays, but they typically represent a light load and incur a small burden on the other jobs. The classifier operates online and learns by using data collected over the previous weeks, facilitating its deployment and enabling a fast adaptation to changes in the workload characteristics. We evaluate our approach using four scheduling policies on seven HPC platform workload traces. We show that: first, incorporating such classification reduces the average bounded slowdown of jobs in all scenarios, second, in most considered scenarios, the improvements are comparable to the ideal hypothetical situation where the scheduler would know in advance the exact running time of jobs.
In 4, we evaluate the relevance of banditlike strategies for the Fog computing context and explore the informationcoordination tradeoff. Fog computing emerges as a potential solution to handle the growth of traffic and processing demands, providing nearby resources to run IoT applications. In this paper, we consider the reconfiguration problem, i.e., how to dynamically adapt the placement of IoT applications running in the Fog, depending on application needs and evolution of resource usage. We propose and evaluate a series of reconfiguration algorithms, based on both online scheduling (dynamic packing) and online learning (bandit) approaches. Through an extensive set of experiments in a realistic testbed built on Grid5000 and FITIoT lab, we demonstrate that the performance strongly and mainly depends on the quality and availability of information from both Fog infrastructure and IoT applications. We show that a reactive and greedy strategy can overcome the performance of stateoftheart online learning algorithms, as long as the strategy has access to a little extra information.
Finally, the high degree of variability present in current and emerging mobile wireless networks calls for mathematical tools and techniques that transcend classical (convex) optimization paradigms. In 20, we provide a gentle introduction to online learning and optimization algorithms that are able to provably cope with this variability and provide policies that are asymptotically optimal in hindsighta property known as no regret. The focal point of this survey is to delineate the tradeoff between the information available as feedback to the learner, and the achievable regret guarantees starting with the case of gradientbased (firstorder) feedback, then moving on to valuebased (zerothorder) feedback, and, ultimately, pushing the envelope to the extreme case of a single bit of feedback. We illustrate our theoretical analysis with a series of practical wireless network examples that highlight the potential of this elegant toolbox.
8.4 Exploiting specific structures in reinforcement learning
Participants: Jonatha Anselmi, Yan Chen, Kimang Khun, Nicolas Gast, Bruno Gaujal, LouisSébastin Rebuffi, Panayotis Mertikopoulos.
The Multiarmed Stochastic Bandit framework is a classic reinforcement learning problem to study the exploration exploitation tradeoff dilemma and for which several optimal algorithms like UCB 2 and Thompson sampling3, whose optimality has only recently been proved by Kaufmann et al.4, have been proposed. Although the first strategy is an optimistic strategy which systematically chooses the "most promising" arm, the second one build on a Bayesian perspective and samples the posterior to decide which arm to select. The Markovian Bandit allows to model situations where the reward distribution is modeled as a Markov chain and may thus exhibit temporal changes. A key challenge in this context is the curse of dimensionality, which basically says that the state size of the Markov process is exponential in the number of the system components so that the complexity of computing an optimal policy and its value are exponential. This is why specific algorithms should be designed to exploit the very specific structures that some state space exhibit. Based on earlier results, we have presented these lines of thought in two articles in special issue of Queueing Systems, 100 views on queues: Learning in Queues7, and Why (and When) do Asymptotic Methods Work so well?6.
Restless bandits are a specific kind of bandits in which the state of each arm evolves according to a Markov process independently of the learner's actions. Most restless Markovian bandits problems in infinite horizon can be solved quasioptimally using the famous Whittle index, which is a generalization of the Gittins index. In 41, we develop an algorithm to test the indexability and compute the Whittle indices of any finitestate restless bandit arm. This algorithm works in the discounted and nondiscounted cases, and can compute Gittins index. Our algorithm builds on three tools: (1) a careful characterization of Whittle index that allows one to compute recursively the $k$th smallest index from the (k  1)th smallest, and to test indexability, (2) the use of the ShermanMorrison formula to make this recursive computation efficient, and (3) a sporadic use of the fastest matrix inversion and multiplication methods to obtain a subcubic complexity. We show that an efficient use of the ShermanMorrison formula leads to an algorithm that computes Whittle index in $2/3{n}^{3}+o\left({n}^{3}\right)$ arithmetic operations, where $n$ is the number of states of the arm. The careful use of fast matrix multiplication leads to the first subcubic algorithm to compute Whittle or Gittins index: By using the current fastest matrix multiplication, the theoretical complexity of our algorithm is $O\left({n}^{2.5286}\right)$. We also develop an efficient implementation of our algorithm that can compute indices of Markov chains with several thousands of states in less than a few seconds.
In 42, we provide a framework to analyse control policies for the restless Markovian bandit model, under both finite and infinite time horizon. We show that when the population of arms goes to infinity, the value of the optimal control policy converges to the solution of a linear program (LP). We provide necessary and sufficient conditions for a generic control policy to be: i) asymptotically optimal; ii) asymptotically optimal with square root convergence rate; iii) asymptotically optimal with exponential rate. We then construct the LPindex policy that is asymptotically optimal with square root convergence rate on all models, and with exponential rate if the model is nondegenerate in finite horizon, and satisfies a uniform global attractor property in infinite horizon. We next define the LPupdate policy, which is essentially a repeated LPindex policy that solves a new linear program at each decision epoch. We provide numerical experiments to compare the efficiency of LPbased policies. We compare the performance of the LPindex policy and the LPupdate policy with other heuristics. Our result demonstrates that the LPupdate policy outperforms the LPindex policy in general, and can have a significant advantage when the transition matrices are wrongly estimated.
In 33, we revisit the regret of undiscounted reinforcement learning in MDPs with a birth and death structure, which are typical of queuing systems. Specifically, we consider a controlled queue with impatient jobs and the main objective is to optimize a tradeoff between energy consumption and userperceived performance. Within this setting, the diameter $D$ of the MDP is $\Omega \left({S}^{S}\right)$, where $S$ is the number of states. Therefore, the existing lower and upper bounds on the regret at time $T$, of order $O\left(\sqrt{DSAT}\right)$ for MDPs with $S$ states and $A$ actions, may suggest that reinforcement learning is inefficient here. In our main result however, we exploit the structure of our MDPs to show that the regret of a slightlytweaked version of the classical learning algorithm UCRL2 is in fact upper bounded by $\tilde{\mathcal{O}}\left(\sqrt{{E}_{2}AT}\right)$ where ${E}_{2}$ is related to the weighted second moment of the stationary measure of a reference policy. Importantly, ${E}_{2}$ is bounded independently of S. Thus, our bound is asymptotically independent of the number of states and of the diameter. This result is based on a careful study of the number of visits performed by the learning algorithm to the states of the MDP, which is highly nonuniform.
Finally, in many online decision processes, the optimizing agent is called to choose between large numbers of alternatives with many inherent similarities; in turn, these similarities imply closely correlated losses that may confound standard discrete choice models and bandit algorithms. We study this question in the context of nested bandits, a class of adversarial multiarmed bandit problems where the learner seeks to minimize their regret in the presence of a large number of distinct alternatives with a hierarchy of embedded (noncombinatorial) similarities. In this setting, optimal algorithms based on the exponential weights blueprint (like Hedge, EXP3, and their variants) may incur significant regret because they tend to spend excessive amounts of time exploring irrelevant alternatives with similar, suboptimal costs. To account for this, we propose in 30 a nested exponential weights (NEW) algorithm that performs a layered exploration of the learner's set of alternatives based on a nested, stepbystep selection method. In so doing, we obtain a series of tight bounds for the learner's regret showing that online learning problems with a high degree of similarity between alternatives can be resolved efficiently, without a red bus / blue bus paradox occurring.
8.5 Distributed learning and optimization
Participants: YuGuan Hsieh, Panayotis Mertikopoulos.
Many learning algorithms operate in centralized way, which raises many practical issues in terms of scalability, privacy, hence a high interest for designing efficient distributed and federated machine learning algorithms. In such context it is essential to design systems that can seamlessly adapt to the workload and to the evolving behaviour of its components (users, resources, network).
In 43 we consider decentralized optimization problems in which a number of agents collaborate to minimize the average of their local functions by exchanging over an underlying communication graph. Specifically, we place ourselves in an asynchronous model where only a random portion of nodes perform computation at each iteration, while the information exchange can be conducted between all the nodes and in an asymmetric fashion. For this setting, we propose an algorithm that combines gradient tracking and variance reduction over the entire network. This enables each node to track the average of the gradients of the objective functions. Our theoretical analysis shows that the algorithm converges linearly, when the local objective functions are strongly convex, under mild connectivity conditions on the expected mixing matrices. In particular, our result does not require the mixing matrices to be doubly stochastic. In the experiments, we investigate a broadcast mechanism that transmits information from computing nodes to their neighbors, and confirm the linear convergence of our method on both synthetic and realworld datasets.
One of the most widely used methods for solving largescale stochastic optimization problems is distributed asynchronous stochastic gradient descent (DASGD), a family of algorithms that result from parallelizing stochastic gradient descent on distributed computing architectures (possibly) asychronously. However, a key obstacle in the efficient implementation of DASGD is the issue of delays: when a computing node contributes a gradient update, the global model parameter may have already been updated by other nodes several times over, thereby rendering this gradient information stale. These delays can quickly add up if the computational throughput of a node is saturated, so the convergence of DASGD may be compromised in the presence of large delays. In 15, we show that, by carefully tuning the algorithm's stepsize, convergence to the critical set is still achieved in mean square, even if the delays grow unbounded at a polynomial rate. We also establish finer results in a broad class of structured optimization problems (called variationally coherent), where we show that DASGD converges to a global optimum with probability 1 under the same delay assumptions. Together, these results contribute to the broad landscape of largescale nonconvex stochastic optimization by offering stateoftheart theoretical guarantees and providing insights for algorithm design.
In 10, we provide a general framework for studying multiagent online learning problems in the presence of delays and asynchronicities. Specifically, we propose and analyze a class of adaptive dual averaging schemes in which agents only need to accumulate gradient feedback received from the whole system, without requiring any betweenagent coordination. In the singleagent case, the adaptivity of the proposed method allows us to extend a range of existing results to problems with potentially unbounded delays between playing an action and receiving the corresponding feedback. In the multiagent case, the situation is significantly more complicated because agents may not have access to a global clock to use as a reference point; to overcome this, we focus on the information that is available for producing each prediction rather than the actual delay associated with each feedback. This allows us to derive adaptive learning strategies with optimal regret bounds, even in a fully decentralized, asynchronous environment. Finally, we also analyze an "optimistic" variant of the proposed algorithm which is capable of exploiting the predictability of problems with a slower variation and leads to improved regret bounds.
In decentralized optimization environments, each agent $i$ in a network of $n$ nodes has its own private function ${f}_{i}$, and nodes communicate with their neighbors to cooperatively minimize the aggregate objective ${\sum}_{i=1}^{n}{f}_{i}$. In this setting, synchronizing the nodes' updates incurs significant communication overhead and computational costs, so much of the recent literature has focused on the analysis and design of asynchronous optimization algorithms, where agents activate and communicate at arbitrary times without needing a global synchronization enforcer. However, most works assume that when a node activates, it selects the neighbor to contact based on a fixed probability (e.g., uniformly at random), a choice that ignores the optimization landscape at the moment of activation. Instead, in 21 we introduce an optimizationaware selection rule that chooses the neighbor providing the highest dual cost improvement (a quantity related to a dualization of the problem based on consensus). This scheme is related to the coordinate descent (CD) method with the GaussSouthwell (GS) rule for coordinate updates; in our setting however, only a subset of coordinates is accessible at each iteration (because each node can communicate only with its neighbors), so the existing literature on GS methods does not apply. To overcome this difficulty, we develop a new analytical framework for smooth and strongly convex ${f}_{i}$ that covers the class of setwise CD algorithms – a class that directly applies to decentralized scenarios, but is not limited to them – and we show that the proposed setwise GS rule achieves a speedup factor of up to the maximum degree in the network (which is in the order of $\Theta \left(n\right)$ for highly connected graphs). The speedup predicted by our analysis is validated in numerical experiments with synthetic data.
8.6 Learning in games
Participants: Kimon Antonakopoulos, Yu Guan Hsieh, Panayotis Mertikopoulos, Bary Pradelski, Patrick Loiseau.
Learning in games naturally occurs in situations where the resources or the decision is distributed among several agents or even in situations where a centralised decision maker has to adapt to strategic users. Yet, it is considerably more difficult than in classical minimization games as the resulting equilibria may be attractive or not and the dynamic often exhibit cyclic behaviors.
In 24, we examine the problem of regret minimization when the learner is involved in a continuous game with other optimizing agents: in this case, if all players follow a noregret algorithm, it is possible to achieve significantly lower regret relative to fully adversarial environments. We study this problem in the context of variationally stable games (a class of continuous games which includes all convexconcave and monotone games), and when the players only have access to noisy estimates of their individual payoff gradients. If the noise is additive, the gametheoretic and purely adversarial settings enjoy similar regret guarantees; however, if the noise is multiplicative, we show that the learners can, in fact, achieve constant regret. We achieve this faster rate via an optimistic gradient scheme with learning rate separation that is, the method's extrapolation and update steps are tuned to different schedules, depending on the noise profile. Subsequently, to eliminate the need for delicate hyperparameter tuning, we propose a fully adaptive method that smoothly interpolates between worstand bestcase regret guarantees.
In 9, we examine the longrun behavior of a wide range of dynamics for learning in nonatomic games, in both discrete and continuous time. The class of dynamics under consideration includes fictitious play and its regularized variants, the best reply dynamics (again, possibly regularized), as well as the dynamics of dual averaging / "follow the regularized leader" (which themselves include as special cases the replicator dynamics and Friedman's projection dynamics). Our analysis concerns both the actual trajectory of play and its timeaverage, and we cover potential and monotone games, as well as games with an evolutionarily stable state (global or otherwise). We focus exclusively on games with finite action spaces; nonatomic games with continuous action spaces are treated in detail in Part II of this work.
In 45, we develop a unified stochastic approximation framework for analyzing the longrun behavior of multiagent online learning in games. Our framework is based on a "primaldual", mirrored RobbinsMonro (MRM) template which encompasses a wide array of popular gametheoretic learning algorithms (gradient methods, their optimistic variants, the EXP3 algorithm for learning with payoffbased feedback in finite games, etc.). In addition to providing an integrated view of these algorithms, the proposed MRM blueprint allows us to obtain a broad range of new convergence results, both asymptotic and in finite time, in both continuous and finite games.
Learning in stochastic games is a notoriously difficult problem because, in addition to each other's strategic decisions, the players must also contend with the fact that the game itself evolves over time, possibly in a very complicated manner. Because of this, the convergence properties of popular learning algorithms  like policy gradient and its variants  are poorly understood, except in specific classes of games (such as potential or twoplayer, zerosum games). In view of this, we examine in 23 the longrun behavior of policy gradient methods with respect to Nash equilibrium policies that are secondorder stationary (SOS) in a sense similar to the type of sufficiency conditions used in optimization. Our first result is that SOS policies are locally attracting with high probability, and we show that policy gradient trajectories with gradient estimates provided by the REINFORCE algorithm achieve an $O(1/\sqrt{n})$ distancesquared convergence rate if the method's stepsize is chosen appropriately. Subsequently, specializing to the class of deterministic Nash policies, we show that this rate can be improved dramatically and, in fact, policy gradient methods converge within a finite number of iterations in that case.
In 5, we examine the longrun behavior of multiagent online learning in games that evolve over time. Specifically, we focus on a wide class of policies based on mirror descent, and we show that the induced sequence of play (a) converges to Nash equilibrium in timevarying games that stabilize in the long run to a strictly monotone limit; and (b) it stays asymptotically close to the evolving equilibrium of the sequence of stage games (assuming they are strongly monotone). Our results apply to both gradientbased and payoffbased feedback  i.e., when players only get to observe the payoffs of their chosen actions.
In 29, we also investigate the impact of feedback quantization on multiagent learning. In particular, we analyze the equilibrium convergence properties of the wellknown "follow the regularized leader" (FTRL) class of algorithms when players can only observe a quantized (and possibly noisy) version of their payoffs. In this informationconstrained setting, we show that coarser quantization triggers a qualitative shift in the convergence behavior of FTRL schemes. Specifically, if the quantization error lies below a threshold value (which depends only on the underlying game and not on the level of uncertainty entering the process or the specific FTRL variant under study), then (i) FTRL is attracted to the game's strict Nash equilibria with arbitrarily high probability; and (ii) the algorithm's asymptotic rate of convergence remains the same as in the nonquantized case. Otherwise, for larger quantization levels, these convergence properties are lost altogether: players may fail to learn anything beyond their initial state, even with full information on their payoff vectors. This is in contrast to the impact of quantization in continuous optimization problems, where the quality of the obtained solution degrades smoothly with the quantization level.
Finally, the literature on evolutionary game theory suggests that pure strategies that are strictly dominated by other pure strategies always become extinct under imitative game dynamics, but they can survive under innovative dynamics. As we explain in 12, this is because innovative dynamics favour rare strategies while standard imitative dynamics do not. However, as we also show, there are reasonable imitation protocols that favour rare or frequent strategies, thus allowing strictly dominated strategies to survive in large classes of imitation dynamics. Dominated strategies can persist at nontrivial frequencies even when the level of domination is not small.
8.7 Advanced learning and optimization methods
Participants: Kimon Antonakopoulos, Yu Guan Hsieh, Panayotis Mertikopoulos.
Variational inequalities  and, in particular, stochastic variational inequalities  have recently attracted considerable attention in machine learning and learning theory as a flexible paradigm for "optimization beyond minimization", i.e., for problems where finding an optimal solution does not necessarily involve minimizing a loss function.
In 39, we examine the lastiterate convergence rate of Bregman proximal methods  from mirror descent to mirrorprox  in constrained variational inequalities. Our analysis shows that the convergence speed of a given method depends sharply on the Legendre exponent of the underlying Bregman regularizer (Euclidean, entropic, or other), a notion that measures the growth rate of said regularizer near a solution. In particular, we show that boundary solutions exhibit a clear separation of regimes between methods with a zero and nonzero Legendre exponent respectively, with linear convergence for the former versus sublinear for the latter. This dichotomy becomes even more pronounced in linearly constrained problems where, specifically, Euclidean methods converge along sharp directions in a finite number of steps, compared to a linear rate for entropic methods.
Universal methods for optimization are designed to achieve theoretically optimal convergence rates without any prior knowledge of the problem's regularity parameters or the accuracy of the gradient oracle employed by the optimizer. In this regard, existing stateoftheart algorithms achieve an $O(1/{T}^{2})$ value convergence rate in Lipschitz smooth problems with a perfect gradient oracle, and an $O(1/\sqrt{T})$ convergence rate when the underlying problem is nonsmooth and/or the gradient oracle is stochastic. On the downside, these methods do not take into account the problem's dimensionality, and this can have a catastrophic impact on the achieved convergence rate, in both theory and practice. In 19 we aim to bridge this gap by providing a scalable universal gradient method  dubbed UnderGrad  whose oracle complexity is almost dimensionfree in problems with a favorable geometry (like the simplex, linearly constrained semidefinite programs and combinatorial bandits), while retaining the orderoptimal dependence on $T$ described above. These "bestofbothworlds" results are achieved via a primaldual update scheme inspired by the dual exploration method for variational inequalities.
Adaptive firstorder methods in optimization are prominent in machine learning and data science owing to their ability to automatically adapt to the landscape of the function being optimized. However, their convergence guarantees are typically stated in terms of vanishing gradient norms, which leaves open the issue of converging to undesirable saddle points (or even local maximizers). In 18, we focus on the AdaGrad family of algorithmswith scalar, diagonal or fullmatrix preconditioningand we examine the question of whether the method's trajectories avoid saddle points. A major challenge that arises here is that AdaGrad's stepsize (or, more accurately, the method's preconditioner) evolves over time in a filtrationdependent way, i.e., as a function of all gradients observed in earlier iterations; as a result, avoidance results for methods with a constant or vanishing stepsize do not apply. We resolve this challenge by combining a series of stepsize stabilization arguments with a recursive representation of the AdaGrad preconditioner that allows us to employ stable manifold techniques and ultimately show that the induced trajectories avoid saddle points from almost any initial condition.
Many important learning algorithms, such as stochastic gradient methods, are often deployed to solve nonlinear problems on Riemannian manifolds. Motivated by these applications, we propose in 25 a family of Riemannian algorithms generalizing and extending the seminal stochastic approximation framework of Robbins and Monro. Compared to their Euclidean counterparts, Riemannian iterative algorithms are much less understood due to the lack of a global linear structure on the manifold. We overcome this difficulty by introducing an extended Fermi coordinate frame which allows us to map the asymptotic behavior of the proposed Riemannian RobbinsMonro (RRM) class of algorithms to that of an associated deterministic dynamical system under very mild assumptions on the underlying manifold. In so doing, we provide a general template of almost sure convergence results that mirrors and extends the existing theory for Euclidean RobbinsMonro schemes, albeit with a significantly more involved analysis that requires a number of new geometric ingredients. We showcase the flexibility of the proposed RRM framework by using it to establish the convergence of a retractionbased analogue of the popular optimistic / extragradient methods for solving minimization problems and games, and we provide a unified treatment for their convergence.
Last, in 44, we propose and analyze exact and inexact regularized Newtontype methods for finding a global saddle point of a convexconcave unconstrained minmax optimization problem. Compared to their firstorder counterparts, investigations of secondorder methods for minmax optimization are relatively limited, as obtaining global rates of convergence with secondorder information is much more involved. In this paper, we highlight how secondorder information can be used to speed up the dynamics of dual extrapolation methods despite inexactness. Specifically, we show that the proposed algorithms generate iterates that remain within a bounded set and the averaged iterates converge to an $\u03f5$saddle point within $O\left({\u03f5}^{2/3}\right)$ iterations in terms of a gap function. Our algorithms match the theoretically established lower bound in this context and our analysis provides a simple and intuitive convergence analysis for secondorder methods without requiring any compactness assumptions. Finally, we present a series of numerical experiments on synthetic and real data that demonstrate the efficiency of the proposed algorithms.
8.8 Random matrix analysis and Machine Learning
Participants: Romain Couillet, Hugo Lebeau.
Random matrix theory has recently proven to be a very effective tool to understand Machine Learning challenges. In particular, concentration results can be used to derive more efficient and frugal algorithms.
Several machine learning problems such as latent variable model learning and community detection can be addressed by estimating a lowrank signal from a noisy tensor. Despite recent substantial progress on the fundamental limits of the corresponding estimators in the largedimensional setting, some of the most significant results are based on spin glass theory, which is not easily accessible to nonexperts. In 8, we propose a sharply distinct and more elementary approach, relying on tools from random matrix theory. The key idea is to study random matrices arising from contractions of a random tensor, which give access to its spectral properties. In particular, for a symmetric $d$th order rankone model with Gaussian noise, our approach yields a novel characterization of maximum likelihood (ML) estimation performance in terms of a fixedpoint equation valid in the regime where weak recovery is possible. For $d=3$, the solution to this equation matches the existing results. We conjecture that the same holds for any order $d$, based on numerical evidence for $d\in 4,5$. Moreover, our analysis illuminates certain properties of the largedimensional ML landscape. Our approach can be extended to other models, including asymmetric and nonGaussian ones.
In 11, we introduce a random matrix framework for the analysis of clustering on highdimensional data streams, a particularly relevant setting for a more sober processing of large amounts of data with limited memory and energy resources. Assuming data ${x}_{1},{x}_{2},...$ arrives as a continuous flow and a small number $L$ of them can be kept in the learning pipeline, one has only access to the diagonal elements of the Gram kernel matrix: ${\left[{K}_{L}\right]}_{i,j}=\frac{1}{p}{x}_{i}^{T}{x}_{j}{1}_{ij<L}$. Under a largedimensional data regime, we derive the limiting spectral distribution of the banded kernel matrix ${K}_{L}$ and study its isolated eigenvalues and eigenvectors, which behave in an unfamiliar way. We detail how these results can be used to perform efficient online kernel spectral clustering and provide theoretical performance guarantees. Our findings are empirically confirmed on image clustering tasks. Leveraging on optimality results of spectral methods for clustering, this work offers insights on efficient online clustering techniques for highdimensional data. This work has also been presented at the GRETSI 28.
8.9 Fairness and equity in digital (recommendation, advertising, persistent storage) systems
Participants: Nicolas Gast, Patrick Loiseau, Rémi Molina, Bary Pradelski, Benjamin Roussillon, Till Kletti.
The general deployment of machinelearning systems in many domains ranging from security to recommendation and advertising to guide strategic decisions leads to an interesting line of research from a game theory perspective. In this context, fairness, discrimination, and privacy are particularly important issues.
Discrimination in machine learning often arises along multiple dimensions (a.k.a. protected attributes); it is then desirable to ensure intersectional fairnessi.e., that no subgroup is discriminated against. It is known that ensuring marginal fairness for every dimension independently is not sufficient in general. Due to the exponential number of subgroups, however, directly measuring intersectional fairness from data is impossible. In 31, our primary goal is to understand in detail the relationship between marginal and intersectional fairness through statistical analysis. We first identify a set of sufficient conditions under which an exact relationship can be obtained. Then, we prove bounds (easily computable through marginal fairness and other meaningful statistical quantities) in high probability on intersectional fairness in the general case. Beyond their descriptive value, we show that these theoretical bounds can be leveraged to derive a heuristic improving the approximation and bounds of intersectional fairness by choosing, in a relevant manner, protected attributes for which we describe intersectional subgroups. Finally, we test the performance of our approximations and bounds on real and synthetic datasets.
To better understand discriminations and the effect of affirmative actions in selection problems (e.g., college admission or hiring), a recent line of research proposed a model based on differential variance. This model assumes that the decisionmaker has a noisy estimate of each candidate’s quality and puts forward the difference in the noise variances between different demographic groups as a key factor to explain discrimination. The literature on differential variance, however, does not consider the strategic behavior of candidates who can react to the selection procedure to improve their outcome, which is wellknown to happen in many domains. In 22, we study how the strategic aspect affects fairness in selection problems. We propose to model selection problems with strategic candidates as a contest game: A population of rational candidates compete by choosing an effort level to increase their quality. They incur a costofeffort but get a (random) quality whose expectation equals the chosen effort. A Bayesian decisionmaker observes a noisy estimate of the quality of each candidate (with differential variance) and selects the fraction α of best candidates based on their posterior expected quality; each selected candidate receives a reward S. We characterize the (unique) equilibrium of this game in the different parameters’ regimes, both when the decisionmaker is unconstrained and when they are constrained to respect the fairness notion of demographic parity. Our results reveal important impacts of the strategic behavior on the discrimination observed at equilibrium and allow us to understand the effect of imposing demographic parity in this context. In particular, we find that, in many cases, the results contrast with the nonstrategic setting. We also find that, when the costofeffort depends on the demographic group (which is reasonable in many cases), then it entirely governs the observed discrimination (i.e., the noise becomes a secondorder effect that does not have any impact on discrimination). Finally we find that imposing demographic parity can sometimes increase the quality of the selection at equilibrium; which surprisingly contrasts with the optimality of the Bayesian decisionmaker in the nonstrategic case. Our results give a new perspective on fairness in selection problems, relevant in many domains where strategic behavior is a reality.
In 34, we consider the problem of linear regression from strategic data sources with a public good component, i.e., when data is provided by strategic agents who seek to minimize an individual provision cost for increasing their data's precision while benefiting from the model's overall precision. In contrast to previous works, our model tackles the case where there is uncertainty on the attributes characterizing the agents' data – a critical aspect of the problem when the number of agents is large. We provide a characterization of the game's equilibrium, which reveals an interesting connection with optimal design. Subsequently, we focus on the asymptotic behavior of the covariance of the linear regression parameters estimated via generalized least squares as the number of data sources becomes large. We provide upper and lower bounds for this covariance matrix and we show that, when the agents' provision costs are superlinear, the model's covariance converges to zero but at a slower rate relative to virtually all learning problems with exogenous data. On the other hand, if the agents' provision costs are linear, this covariance fails to converge. This shows that even the basic property of consistency of generalized least squares estimators is compromised when the data sources are strategic.
In 26, we consider the problem of computing a sequence of rankings that maximizes consumerside utility while minimizing producerside individual unfairness of exposure. While prior work has addressed this problem using linear or quadratic programs on bistochastic matrices, such approaches, relying on Birkhoffvon Neumann (BvN) decompositions, are too slow to be implemented at large scale. In this paper we introduce a geometrical object, a polytope that we call expohedron, whose points represent all achievable exposures of items for a Position Based Model (PBM). We exhibit some of its properties and lay out a Carathéodory decomposition algorithm with complexity $O({n}^{2}log\left(n\right))$ able to express any point inside the expohedron as a convex sum of at most $n$ vertices, where $n$ is the number of items to rank. Such a decomposition makes it possible to express any feasible target exposure as a distribution over at most n rankings. Furthermore we show that we can use this polytope to recover the whole Pareto frontier of the multiobjective fairnessutility optimization problem, using a simple geometrical procedure with complexity $O({n}^{2}log\left(n\right))$. Our approach compares favorably to linear or quadratic programming baselines in terms of algorithmic complexity and empirical runtime and is applicable to any merit that is a nondecreasing function of item relevance. Furthermore our solution can be expressed as a distribution over only $n$ permutations, instead of the ${(n1)}^{2}+1$ achieved with BvN decompositions. We perform experiments on synthetic and realworld datasets, confirming our theoretical results.
In recent years, it has become clear that rankings delivered in many areas need not only be useful to the users but also respect fairness of exposure for the item producers. We consider the problem of finding ranking policies that achieve a Paretooptimal tradeoff between these two aspects. Several methods were proposed to solve it; for instance a popular one is to use linear programming with a Birkhoffvon Neumann decomposition. These methods, however, are based on a classical Position Based exposure Model (PBM), which assumes independence between the items (hence the exposure only depends on the rank). In many applications, this assumption is unrealistic and the community increasingly moves towards considering other models that include dependencies, such as the Dynamic Bayesian Network (DBN) exposure model. For such models, computing (exact) optimal fair ranking policies remains an open question. In 27, we answer this question by leveraging a new geometrical method based on the socalled expohedron proposed recently for the PBM. We lay out the structure of a new geometrical object (the DBNexpohedron), and propose for it a Carathéodory decomposition algorithm of complexity $O\left({n}^{3}\right)$, where $n$ is the number of documents to rank. Such an algorithm enables expressing any feasible expected exposure vector as a distribution over at most n rankings; furthermore we show that we can compute the whole set of Paretooptimal expected exposure vectors with the same complexity $O\left({n}^{3}\right)$. Our work constitutes the first exact algorithm able to efficiently find a Paretooptimal distribution of rankings. It is applicable to a broad range of fairness notions, including classical notions of meritocratic and demographic fairness. We empirically evaluate our method on the TREC2020 and MSLR datasets and compare it to several baselines in terms of Paretooptimality and speed.
9 Bilateral contracts and grants with industry
Participants: HenryJoseph Audéoud, Till Kletti, Nicolas Gast, Patrick Loiseau.
Patrick Loiseau has a Cifre contract with Naver labs (20202023) on "Fairness in multistakeholder recommendation platforms”, which supports the PhD student Till Kletti.
Nicolas Gast obtained a grant from Enedis to evaluate the performance of the PLCG3 protocol. This grant supported the postdoc of HenryJoseph Audeoud.
10 Partnerships and cooperations
10.1 International initiatives
10.1.1 Inria associate team not involved in an IIL or an international program
Participants: Vincent Danjean, Guillaume Huard, Arnaud Legrand, Lucas Leandro Nesi, JeanMarc Vincent.
ReDaS

Title:
Reproducible Data Science

Duration:
2019  2022

Coordinator:
Lucas Mello Schnorr (schnorr@inf.ufrgs.br)

Partners:
 Universidade Federal do Rio Grande do Sul Porto Alegre (Brésil)

Inria contact:
Guillaume Huard

Summary:
Data science builds on a variety of techniques and tools that makes analysis often difficult to follow and reproduce. The goal of this project is to develop interactive, reproducible and scalable analysis workflows that provide uncertainty and quality estimators about the analysis.
10.2 International research visitors
10.2.1 Visits of international scientists
Anshul Gandhi

Status
Full Professor

Institution of origin:
Stony Brook University

Country:
USA

Dates:
20  29 Sep.

Context of the visit:
Collaboration with Nicolas Gast

Mobility program/type of mobility:
Research stay
Nicolas Maillard

Status
Full Professor

Institution of origin:
Univ. Federale do Rio Grande do Sul

Country:
Brazil

Dates:
16  26 Oct.

Context of the visit:
ReDAS associated team

Mobility program/type of mobility:
Research stay
Bryce Ferguson

Status
PhD Student

Institution of origin:
Univ. of California, Santa Barbara

Country:
USA

Dates:
9 Oct 23 Nov.

Context of the visit:
Collaboration with Panayotis Mertikopoulos

Mobility program/type of mobility:
Research stay
Swann Perarnau

Status
Research scientist

Institution of origin:
Argonne National Laboratory

Country:
USA

Dates:
1 May  10 July

Context of the visit:
Joint Laboratory on Extreme Scale Computing

Mobility program/type of mobility:
Research stay
10.2.2 Visits to international teams
Research stays abroad
Panayotis Mertikopoulos

Visited institution:
Simons Institute for the Theory of Computing, Berkeley, CA

Country:
USA

Dates:
MarchApril 2022

Context of the visit:
Visiting scientist

Mobility program/type of mobility:
Research stay
10.3 European initiatives
10.3.1 Other european programs/initiatives
Participants: Arnaud Legrand.

Unite!
Arnaud Legrand is involved in the WP6 (Open Science) of the Unite! (University Network for Innovation, Technology and Engineering) project , which aims to create a large European campus from Finland to Portugal. Unite! brings together 7 partners, recognized for the quality of their education and research: Technische Universität Darmstadt (Germany), Aalto University (Finland), Kunglia Tekniska Hoegskolan (Sweden), Politecnico di Torino (Italy), Universitat Politecnica de Catalunya (Spain), Universidade de Lisboa (Portugal) and Grenoble INP, Graduate Schools of Engineering and Management, Université Grenoble Alpes.
10.4 National initiatives
Projects indicated with a $\u2606$ are projects coordinated by members of the POLARIS team.
ANR

ANR ALIAS (PRCI 20202023)$\u2606$
Adaptive Learning for Interactive Agents and Systems
[284K€] Partners: Singapore University of Technology and Design (SUTD).
ALIAS is a bilateral PRCI (collaboration internationale) project joint with Singapore University of Technology and Design (SUTD), coordinated by Bary Pradelski (PI) and involving P. Mertikopoulos and P. Loiseau. The Singapore team consists of G. Piliouras and G. Panageas. The goal of the project is to provide a unified answer to the question of stability in multiagent systems: for systems that can be controlled (such as programmable machine learning models), prescriptive learning algorithms can steer the system towards an optimum configuration; for systems that cannot (e.g., online assignment markets), a predictive learning analysis can determine whether stability can arise in the long run. We aim at identifying the fundamental limits of learning in multiagent systems and design novel, robust algorithms that achieve convergence in cases where conventional online learning methods fail.

ANR REFINO (JCJC 20202024)$\u2606$
Refined Mean Field Optimization
[250K€] REFINO is an ANR starting grant (JCJC) coordinated by Nicolas Gast. The main objective on this project is to provide an innovative framework for optimal control of stochastic distributed agents. Restless bandit allocation is one particular example where the control that can be sent to each arm is restricted to an on/off signal. The originality of this framework is the use of refined mean field approximation to develop control heuristics that are asymptotically optimal as the number of arms goes to infinity and that also have a better performance than existing heuristics for a moderate number of arms. As an example, we will use this framework in the context of smart grids, to develop control policies for distributed electric appliances.

ANR FAIRPLAY (JCJC 20212025)$\u2606$
Fair algorithms via game theory and sequential learning
[245K€] FAIRPLAY is an ANR starting grant (JCJC) coordinated by Patrick Loiseau. Machine learning algorithms are increasingly used to optimize decision making in various areas, but this can result in unacceptable discrimination. The main objective of this project is to propose an innovative framework for the development of learning algorithms that respect fairness constraints. While the literature mostly focuses on idealized settings, the originality of this framework and central focus of this project is the use of game theory and sequential learning methods to account for constraints that appear in practical applications: strategic and decentralized aspects of the decisions and the data provided and absence of knowledge of certain parameters key to the fairness definition.
IRS/UGA

UGA MIAI Chaire (20192023)$\u2606$
[365K€] Patrick Loiseau is in charge of the Explainable and Responsible AI chaire of the MIAI institute. To build more trustworthy AI systems, we investigate how to produce explanations for results returned by AI systems and how to build AI algorithms with guarantees of fairness and privacy, in the setting of varied tasks such as classification, recommendation, resource allocation or matching.

IRS DISCMAN (IRS 20202022)$\u2606$
(Distributed Control for MultiAgent systems and Networks) is a joint IRS project funded by IDEX Université GrenobleAlpes. Its main objectives is to develop distributed equilibrium convergence algorithms for largescale control and optimization problems, both offline and online. It is being coordinated by Panayotis Mertikopoulos (POLARIS), and it involves a joint team of researchers from the LIG and LJK laboratories in Grenoble.
11 Dissemination
11.1 Promoting scientific activities
11.1.1 Scientific events: organisation
 Bruno Gaujal , Nicolas Gast , and Annie Simon have organized the 12th Atelier d'Évaluation de Performance in Grenoble, France, July 2022, which has attracted about 70 researchers of the domain.
 Arnaud Legrand has organized the 3rd Workshop of the LIG SRCPR Axis, Grenoble, June 2022, which is an internal event to the LIG laboratory.
 Panayotis Mertikopoulos has been coorganizer of the 7th workshop on "Stochastic Methods in Game Theory" in Erice, Italy, May 2022.
General chair, scientific chair
Panayotis Mertikopoulos has been area Chair for ICLR 2022 and NeurIPS 2022.
Member of the conference program committees
 Jonatha Anselmi has been a member of the IEEE MASCOTS 2022 Program Committee.
 Nicolas Gast has been a member of the Sigmetrics 2022 and ICLR 2023 Program Committees
 Bruno Gaujal has been a member of the NeurIPS 2022 Program Committee.
 Arnaud Legrand has been a member of the Europar 2022 Program Committee.
Member of the editorial boards
 Nicolas Gast is member of the editorial boards of the journals "Performance Evaluation" and "Stochastic Models”.
 Panayotis Mertikopoulos has been Guest editor of the special issue on "Optimization Challenges in Data Science" for the EURO Journal on Computational Optimization (EJCO).
 Panayotis Mertikopoulos has been guest editor of special issue on "Population Games and Evolutionary Dynamics in Memory of William H. Sandholm" for the Journal of Dynamics and Games (JDG).
11.1.2 Invited talks
 Arnaud Legrand has been invited to present his latest results at the "15th Scheduling for Large Scale Systems" workshop in Fréjus, June 2022 and at the Scheduling workshop in Aussois, June 2022.
 Romain Couillet has been invited to present "Pourquoi et comment démanteler l'IA et le numérique?" at many seminars (with up to 120 participants) and round tables.
 Nicolas Gast has been invited to present his latest results to the GDR COSMOS, at Séminaire Inria Paris, at the DDQC workshop, and at the Stochastic Networks conference.
 Nicolas Gast has been invited to present his latest results at the Workshop scheduling, Fréjus, June 2022.
 Nicolas Gast has been invited to present his latest results at the Gipsa Lab seminar, Grenoble, March 2022.
 Nicolas Gast has been invited to present his latest results at the MASCOTS conference, Nice, October 2022.
 Panayotis Mertikopoulos has given "A crash course in optimization for machine learning" at the Archimedes research center for Artificial Intelligence and Data Science, Athens, JuneJuly 2022.
 Panayotis Mertikopoulos has been invited to present his latest results “On the limits – and limitations – of learning in games” at the Eccellenza workshop on algorithmic game theory, mechanism design, and learning, Turin, IT, Nov. 2022
 Panayotis Mertikopoulos has been invited to present his latest results "Equilibrium and optimality under uncertainty" at the Amazon Science Summit, Barcelona, ES, September 2022
 Panayotis Mertikopoulos has been invited to present his latest results "The dynamics of artificial intelligence" at the Second Congress of Greek Mathematicians, Athens, GR, July 2022
 Panayotis Mertikopoulos has been invited to present his latest results "The limits of gametheoretic learning" at the Controversies in Game Theory Symposium, ETH Zürich, CH, June 2022
 Panayotis Mertikopoulos has been invited to present his latest results "Regularized learning in games" at the Learning with Strategic Agents Keynote at AAMAS 2022, Auckland, NZ, June 2022
 Panayotis Mertikopoulos has been invited to present his latest results "The limits of regularized learning", at the Learning in the Presence of Strategic Behavior workshop, Berkeley, CA, April 2022
 Panayotis Mertikopoulos has been invited to present his latest results "Minmax optimization from a dynamical systems viewpoint" at the Adversarial Approaches in Machine Learning workshop, Berkeley, CA, March 2022
 Panayotis Mertikopoulos has been invited to present his latest results "The longrun limit of online learning in games" at the Purdue University, Lafayette, IN, September 2022
 Panayotis Mertikopoulos has been invited to present his latest results "The evolution of learning in games" at the University of Athens, Athens, GR, March 2022
 Bary Pradelski has been invited to present his experience on COVID policy at Bruegel, at the OECD Economics Brown Bag Seminar, at Terra Nova, at the Institute for Interdisciplinary Innovation in healthcare (Université Libre de Bruxelles) and at the Institut Pasteur.
 Bary Pradelski has been invited to present his latest research results at the Multidisciplinary Institute in Artificial Intelligence (Univ. Grenoble Alpes) and at the Séminaire parisien de Théorie des Jeux.
11.1.3 Scientific expertise
 Barry Pradelski was appointed to share his expertise on COVID policy to the Centre européen de prévention et de contrôle des maladies (ECDC).

JeanMarc Vincent
is vicehead of the SIF, adjunct on teaching.
 He participated to the coordination of the trophées NSI and has been a member of the national committee.
 He also coorganised the teaching days of the SIF.
 He participated to the mediation school organized by the SIF and the Blaise Pascal foundation.
 JeanMarc Vincent has been a member of the NSI (high school computer science teachers) CAPES committee.
11.1.4 Research administration
 Arnaud Legrand is a member of the Section 6 of the CoNRS.
 Arnaud Legrand is head of the SRCPR axis of the LIG and a member of LIG bureau.
 Arnaud Legrand is a member of Comité Scientique of the Inria Grenoble.
 Nicolas Gast has been a member of the hiring committee of a Maître de Conférences at Centrale/Supelec.
 Arnaud Legrand has been a member of the hiring committee of a Professeur at University of Bordeaux.
11.2 Teaching  Supervision  Juries
11.2.1 Teaching
 Jonatha Anselmi teaches the Probability and Simulation and the Performance Evaluation lectures in M1, Polytech Grenoble.
 Arnaud Legrand and JeanMarc Vincent teach the transversal Scientific Methodology and Empirical Evaluation lecture (36h) at the M2 MOSIG.
 Bruno Gaujal teaches the M2 course on Optimization under Uncertainties in M2 ORCO Grenoble
 Bruno Gaujal taught the Reinforcement Learning course at the special week of the computer science department a ENS Lyon
 Nicolas Gast is responsible of the course Reinforcement Learning in Master MOSIG/MSIAM (Grenoble) and of the course « Introduction to machine learning » (License 3, Grenoble).
 Guillaume Huard is responsible of the courses UNIX & C programming in the L1 and L3 INFO, of Object Oriented and EventDriven Programming in the L3 INFO, and of the Objet Oriented Design in M1 INFO.
 Vincent Danjean teaches the Operating Systems, Programming Languages, Algorithms, Computer Science and Mediation lectures in L3, M1 and Polytech Grenoble.
 Romain Couillet is the initiator of a new lecture on the Introduction to Artificial Intelligence in the L1 INFO.
 Romain Couillet has created an Écosophie workshop with Yoan Svejcar (écopsychologue), which has been taught 4 times, in particular in the context of the Penser la crise écologique transversal lecture (Aurélien Barrau, Éléonore Cartelier) and in the PISTE program of GrenobleINP.
 Romain Couillet has created and managed the continuing education program Transformation numérique in collaboration with Denis Trystram (5 students this year, including 2 academics)
 Romain Couillet has created the Incarner le changement: le numérique à l'ère des lowtechs lecture in M1 MoSIG (practical sessions world dynamics, “fresque du numérique”, ecosophy workshop, surveying Illich's "conviviality", etc.)
 Panayotis Mertikopoulos teaches the Online Optimization and Learning in Games M2 lecture at Univ. Limoges
 Panayotis Mertikopoulos teaches the Reinforcement and Online Learning in M2 MOSIG/MSIAM (Grenoble)
 The 3rd edition of the MOOC of Arnaud Legrand , K. Hinsen and C. Pouzat on Reproducible Research: Methodological Principles for a Transparent Science is still running. Over the 3 editions (Oct.Dec. 2018, Apr.June 2019, March 2020  end of 2023), about 18,800 persons have followed this MOOC and 1900 certificates of achievement have been delivered. 54% of participants are PhD students and 12% are undergraduates.
 JeanMarc Vincent teaches Algorithms and Probabilities at the L3, UGA.
 JeanMarc Vincent participates to the Histoire de l’informatique lecture at the ENSIMAG.
11.2.2 Supervision
 Arnaud Legrand has been a member of the Comité de Suivi Individuel of Amaury Maillé (ENS Lyon)
 Arnaud Legrand has been a member of the Comité de Suivi Individuel of Yishu Du (ENS Lyon)
 Arnaud Legrand has been a member of the Comité de Suivi Individuel of Adeyemi Adetula (Univ. Grenoble Alpes)
 Arnaud Legrand has been a member of the Comité de Suivi Individuel of Julien Emmanuel (ENS Lyon)
 Panayotis Mertikopoulos is cosupervising the PhD thesis of Waïss Azizian with Jérôme Malick.
11.2.3 Juries
 Nicolas Gast has been reviewer for the PhD thesis of Thomas Tournaire (Telecom Sud Paris): Modelbased reinforcement learning for dynamic resource allocation in cloud environments
 Bruno Gaujal has been reviewer for the PhD thesis of Leonardo Cianfanelli (Polytechnico di Torino): Learning dynamics in congestion games, intervention in traffic networks, and epidemics control
 Bruno Gaujal has been reviewer for the PhD thesis of Leonardo Massai (Univ. degli studi di Torino): Bankruptcy cascades in interbank market
 Bruno Gaujal has been reviewer for the PhD thesis of Marin Boyet (école polytechnique): Systèmes dynamiques affines par morceaux appliqués à l'évaluation de performance de centres d'appels d'urgence
 Bruno Gaujal has been reviewer for the PhD thesis of Guilherme EspindolaWinck (Univ. Angers): On the stochastic filtering of maxplus linear systems
 Arnaud Legrand has been reviewer for the PhD thesis of Philippe Swartvagher (Univ. Bordeaux): On the Interactions between HPC Taskbased Runtime Systems and Communication Libraries
 Arnaud Legrand has been member of the PhD thesis committee of Mathieu Vérité (Univ. Bordeaux): Algorithmes d’allocation statique pour la planification d’applications haute performance
 Arnaud Legrand has been member of the PhD thesis committee of Alexis Colin (Télécom Sud Paris): De la collecte de trace à la prédiction du comportement d’applications parallèles
 Nicolas Gast has been a member of the committee of the prix de thèse Paul Casseau
 Panayotis Mertikopoulos has been reviewer for the PhD thesis of Geovani Rizk (Univ. ParisDauphine): Stochastic graphical bilinear bandits
 Panayotis Mertikopoulos has been reviewer for the PhD thesis of Laurent Meunier (Univ. ParisDauphine): Adversarial attacks: A theoretical journey
 Panayotis Mertikopoulos has been reviewer for the PhD thesis of Juliette Achddou (École Normale Supérieure): Zerothorder optimization for realtime bidding: A mathematical perspective
11.3 Popularization
11.3.1 Articles and contents
 Romain Couillet and his PhD student Achille Baucher have designed an introduction to “World dynamics and limits to the growth” as a Python practical session which has been taught in several occasions (in the PISTE program of Grenoble INP and in several awarenessraising groups).
11.3.2 Education

JeanMarc Vincent
is particularly involved in the teaching of computer science at the national level.
 He coanimates the national group infosansordi (unplugged computer science).
 He participates to the training of high school computer science teachers (NSI) and he is preparing a new training on programming methods.
 He also coorganizes series of seminars for high school computer science teachers.
 He is a member of the national interIREM group on computer science and writes leaflets for high school computer science teachers.
 He contributes to the MOOC « Enseigner l’Informatique au Lycée » and is responsible of the algorithms part.
 JeanMarc Vincent has participated to the Module d’Initiative Nationale: École inclusive et outils numériques au service des apprentissages, in November 2022.
11.3.3 Interventions
 Arnaud Legrand has presented Open Science and Reproducible Research Challenges at the Université Inter Ages du Dauphiné, May 2022, at the séminaire des élèves de l'ENS Lyon, October 2022, and at the M1 MOSIG students, Univ. Grenoble Alpes, December 2022.
 JeanMarc Vincent has organized 4 conferences at the Kateb Yacine library for general audience: Le numérique en questions, with Inria, UGA and the Grenoble city.
 Romain Couillet has organized a "Ingénieur lowtech" workshop at the forum des métiers of the Lycée du Grésivaudan à Meylan.
 Romain Couillet connects his professionnal activity with public action: Lowtechlab de Grenoble, Université Autogérée, Arche des Innovateurs, etc.

12 Scientific production
12.1 Publications of the year
International journals
 1 articleMean Field and Refined Mean Field Approximations for Heterogeneous Systems: It Works!Proceedings of the ACM on Measurement and Analysis of Computing Systems 61February 2022, 143
 2 articleReplication vs speculation for load balancing.Queueing Systems1003April 2022, 389391
 3 articleSimulationbased Optimization and Sensibility Analysis of MPI Applications: Variability Matters.Journal of Parallel and Distributed ComputingApril 2022
 4 articleOnline Reconfiguration of IoT Applications in the Fog: The InformationCoordination Tradeoff.IEEE Transactions on Parallel and Distributed Systems3352022, 11561172
 5 articleMultiagent online learning in timevarying games.Mathematics of Operations Research2022
 6 articleWhy (and When) do Asymptotic Methods Work so well?Queueing SystemsMay 2022, 13
 7 articleLearning in Queues.Queueing Systems100April 2022, 521–523
 8 articleA Random Matrix Perspective on Random Tensors.Journal of Machine Learning Research23264September 2022, 136
 9 articleLearning in nonatomic games, Part I: Finite action spaces and population games.Journal of Dynamics and Games94October 2022, 433460
 10 articleMultiagent online optimization with delays: Asynchronicity, adaptivity, and optimism.Journal of Machine Learning Research23782022, 149
 11 articleA Random Matrix Analysis of Data Stream Clustering: Coping With Limited Memory Resources.Proceedings of Machine Learning ResearchJune 2022, 129
 12 articleSurvival of dominated strategies under imitation dynamics.Journal of Dynamics and Games94October 2022, 499528
 13 articlePerformance Analysis of Irregular TaskBased Applications on Hybrid Platforms: Structure Matters.Future Generation Computer Systems135October 2022
 14 articleElimination versus mitigation of SARSCoV2 in the presence of effective vaccines.The Lancet global health101January 2022, e142e147
 15 articleDistributed stochastic optimization with large delays.Mathematics of Operations Research473August 2022, 20822111
 16 articleImproving the Performance of Batch Schedulers Using Online Job Runtime Classification.Journal of Parallel and Distributed Computing164February 2022, 8395
International peerreviewed conferences
 17 inproceedingsEnergy Optimal Activation of Processors for the Execution of a Single Task with Unknown Size.30th International Symposium on the Modeling, Analysis, and Simulation of Computer and Telecommunication SystemsNice, FranceOctober 2022
 18 inproceedingsAdaGrad avoids saddle points.ICML 2022  39th International Conference on Machine LearningBaltimore, United States2022, 141
 19 inproceedingsUnderGrad: A universal blackbox optimization method with almost dimensionfree convergence rate guarantees.ICML 2022  39th International Conference on Machine LearningBaltimore, United States2022, 131
 20 inproceedingsOnline convex optimization in wireless networks and beyond: The feedback performance tradeoff.RAWNET 2022  International Workshop on Resource Allocation and Cooperation in Wireless NetworksTurin, Italy2022, 18
 21 inproceedingsPick your neighbor: Local GaussSouthwell rule for fast asynchronous decentralized optimization.CDC 2022  61st IEEE Annual Conference on Decision and ControlCancun, Mexico2022
 22 inproceedingsFairness in Selection Problems with Strategic Candidates.EC 2022  ACM Conference on Economics and ComputationBoulder, Colorado, United StatesACMJuly 2022, 129
 23 inproceedingsOn the convergence of policy gradient methods to Nash equilibria in general stochastic games.NeurIPS 2022  36th International Conference on Neural Information Processing SystemsNew Orleans, United States2022, 143
 24 inproceedingsNoRegret Learning in Games with Noisy Feedback: Faster Rates and Adaptivity via Learning Rate Separation.NeurIPS 2022  36th International Conference on Neural Information Processing SystemsNew Orleans, United States2022
 25 inproceedingsThe dynamics of Riemannian RobbinsMonro algorithms.COLT 2022  35th Annual Conference on Learning TheoryLondon, United Kingdom2022, 131
 26 inproceedingsIntroducing the Expohedron for Efficient Paretooptimal FairnessUtility Amortizations in Repeated Rankings.WSDM 2022  15th ACM International Conference on Web Search and Data MiningPhoenix (virtual), United StatesACMFebruary 2022, 110
 27 inproceedingsParetoOptimal FairnessUtility Amortizations in Rankings with a DBN Exposure Model.SIGIR 2022  45th International ACM SIGIR Conference on Research and Development in Information RetrievalMadrid, SpainACMJuly 2022, 112
 28 inproceedingsUne analyse par matrices aléatoires du clustering en ligne : comprendre l'impact des limitations en mémoire.GRETSI 2022  XXVIIIème Colloque Francophone de Traitement du Signal et des ImagesNancy, FranceSeptember 2022, 14
 29 inproceedingsLearning in games with quantized payoff observations.CDC 2022  61st IEEE Annual Conference on Decision and ControlCancun, MexicoIEEE2022, 18
 30 inproceedingsNested bandits.ICML 2022  39th International Conference on Machine LearningBaltimore, United StatesJuly 2022
 31 inproceedingsBounding and Approximating Intersectional Fairness through Marginal Fairness.NeurIPS 2022  36th Conference on Neural Information Processing SystemsNew Orleans, United StatesNovember 2022, 132
 32 inproceedingsMultiPhase TaskBased HPC Applications: Quickly Learning how to Run Fast.IPDPS 2022  36th IEEE International Parallel & Distributed Processing SymposiumLyon, FranceIEEEMay 2022, 111
Conferences without proceedings
 33 inproceedingsReinforcement Learning in a Birth and Death Process: Breaking the Dependence on the State Space.NeurIPS 2022  36th Conference on Neural Information Processing SystemsLa Nouvelle Orléans, United StatesNovember 2022
 34 inproceedingsAsymptotic Degradation of Linear Regression Estimates with Strategic Data Sources.ALT 2022  33rd International Conference on Algorithmic Learning TheoryParis, FranceMarch 2022, 131
Scientific books
 35 bookRandom Matrix Methods for Machine Learning.1Cambridge University PressJune 2022
Doctoral dissertations and habilitation theses
 36 thesisAdaptive Algorithms for Optimization Beyond Lipschitz Requirements.Université Grenoble Alpes [2020....]January 2022
 37 thesisFairness in selection problems.Université Grenoble Alpes [2020....]June 2022
 38 thesisClosetoopimal policies for Markovian bandits.Université Grenoble Alpes (UGA)December 2022
Reports & preprints
 39 miscOn the rate of convergence of Bregman proximal methods in constrained variational inequalities.November 2022
 40 misc Reinforcement Learning for Markovian Bandits: Is Posterior Sampling more Scalable than Optimism? 2022
 41 miscTesting Indexability and Computing Whittle and Gittins Index in Subcubic Time.December 2022
 42 miscLPbased policies for restless bandits: necessary and sufficient conditions for (exponentially fast) asymptotic optimality.September 2022
 43 miscPushPull with Device Sampling.June 2022
 44 miscExplicit secondorder minmax optimization methods with optimal convergence guarantees.October 2022
 45 miscLearning in games from a stochastic approximation viewpoint.June 2022
Other scientific publications
 46 articleA multinational Delphi consensus to end the COVID19 public health threat.Nature6117935November 2022, 332345
12.2 Other
Scientific popularization
 47 miscDes professionnel·le·s de l'informatique deviennent enseignant·e·s.October 2022
12.3 Cited publications
 48 inproceedingsMeasuring the Facebook Advertising Ecosystem.NDSS 2019  Proceedings of the Network and Distributed System Security SymposiumSan Diego, United StatesFebruary 2019, 115
 49 inproceedingsInvestigating Ad Transparency Mechanisms in Social Media: A Case Study of Facebook's Explanations.NDSS 2018  Network and Distributed System Security SymposiumSan Diego, United StatesFebruary 2018, 115
 50 articleCombining SizeBased Load Balancing with RoundRobin for Scalable Low Latency.IEEE Transactions on Parallel and Distributed Systems2019, 13
 51 articleAsymptotically Optimal SizeInterval Task Assignments.IEEE Transactions on Parallel and Distributed Systems30112019, 24222433
 52 articlePowerofdChoices with Memory: Fluid Limit and Optimality.Mathematics of Operations Research4532020, 862888
 53 inproceedingsDimemas: Predicting MPI Applications Behaviour in Grid Environments.Proc. of the Workshop on Grid Applications and Programming Tools2003
 54 conferencexSim: The ExtremeScale Simulator.HPCSIstanbul, Turkey2011
 55 inproceedingsAutotuning under Tight Budget Constraints: A Transparent Design of Experiments Approach.CCGrid 2019  International Symposium in Cluster, Cloud, and Grid ComputingLarcana, CyprusIEEEMay 2019, 110
 56 incollectionComprehensive Performance Tracking with VAMPIR 7.Tools for High Performance Computing 2009The paper details the latest improvements in the Vampir visualization tool.Springer Berlin Heidelberg2010
 57 articlePenaltyRegulated Dynamics and Robust Learning Procedures in Games.Mathematics of Operations Research4032015, 611633
 58 articlePerformance analysis methods for listbased caches with nonuniform access.IEEE/ACM Transactions on NetworkingDecember 2020, 118
 59 inproceedingsEquality of Voice: Towards Fair Representation in Crowdsourced TopK Recommendations.FAT* 2019  ACM Conference on Fairness, Accountability, and TransparencyProceedings of the ACM Conference on Fairness, Accountability, and Transparency (FAT*)Atlanta, United StatesACMJanuary 2019, 129138
 60 inproceedingsFast and Faithful Performance Prediction of MPI Applications: the HPL Case Study.2019 IEEE International Conference on Cluster Computing (CLUSTER)Albuquerque, United StatesSeptember 2019
 61 articleSimulating MPI applications: the SMPI approach.IEEE Transactions on Parallel and Distributed Systems288August 2017, 14
 62 inproceedingsLoad Aware Provisioning of IoT Services on Fog Computing Platform.IEEE International Conference on Communications (ICC)Shanghai, ChinaIEEEMay 2019
 63 inproceedings Are meanfield games the limits of finite stochastic games? The 18th Workshop on MAthematical performance Modeling and Analysis Nice, France June 2016
 64 articleDiscrete Mean Field Games: Existence of Equilibria and Convergence.Journal of Dynamics and Games632019, 119
 65 inproceedingsThe Price of Local Fairness in Multistage Selection.IJCAI2019  TwentyEighth International Joint Conference on Artificial IntelligenceMacao, FranceInternational Joint Conferences on Artificial Intelligence OrganizationAugust 2019, 58365842
 66 inproceedingsOn Fair Selection in the Presence of Implicit Variance.EC 2020  TwentyFirst ACM Conference on Economics and ComputationBudapest, HungaryACMJuly 2020, 649–675
 67 inproceedingsNoregret learning and mixed Nash equilibria: They do not mix.NeurIPS 2020  34th International Conference on Neural Information Processing SystemsVancouver, CanadaDecember 2020, 124
 68 articleA Visual Performance Analysis Framework for Taskbased Parallel Applications running on Hybrid Clusters.Concurrency and Computation: Practice and Experience3018April 2018, 131
 69 articleSize Expansions of Mean Field Approximation: Transient and SteadyState Analysis.Performance Evaluation2018, 115

70
inproceedingsExpected Values Estimated via MeanField Approximation are
$1/N$ Accurate.ACM SIGMETRICS / International Conference on Measurement and Modeling of Computer Systems , SIGMETRICS '171ACM SIGMETRICS / International Conference on Measurement and Modeling of Computer Systems , SIGMETRICS '17UrbanaChampaign, United StatesJune 2017, 26  71 unpublishedExponential Convergence Rate for the Asymptotic Optimality of Whittle Index Policy.December 2020,
 72 inproceedingsA Refined Mean Field Approximation.ACM SIGMETRICS 2018Irvine, United StatesJune 2018, 1
 73 articleLinear Regression from Strategic Data Sources.ACM Transactions on Economics and Computation82May 2020, 124
 74 inproceedingsA Refined Mean Field Approximation for Synchronous Population Processes.MAMA 2018Workshop on MAthematical performance Modeling and AnalysisIrvine, United StatesJune 2018, 13
 75 inproceedingsAsymptotically Exact TTLApproximations of the Cache Replacement Algorithms LRU(m) and hLRU.28th International Teletraffic Congress (ITC 28)Würzburg, GermanySeptember 2016
 76 articleTTL Approximations of the Cache Replacement Algorithms LRU(m) and hLRU.Performance EvaluationSeptember 2017
 77 inproceedingsVaccination in a Large Population: Mean Field Equilibrium versus Social Optimum.NETGCOOP 2020  10th International Conference on NETwork Games, COntrol and OPtimizationCargèse, FranceSeptember 2021, 19
 78 inproceedingsA Linear Time Algorithm for Computing Offline Speed Schedules Minimizing Energy Consumption.MSR 2019  12ème Colloque sur la Modélisation des Systèmes RéactifsAngers, FranceNovember 2019, 114
 79 inproceedingsDiscrete and Continuous Optimal Control for Energy Minimization in RealTime Systems.EBCCSP 2020  6th International Conference on EventBased Control, Communication, and Signal ProcessingKrakow, PolandIEEESeptember 2020, 18
 80 articleDynamic Speed Scaling Minimizing Expected Energy Consumption for RealTime Tasks.Journal of SchedulingJuly 2020, 125
 81 techreportExploiting Job Variability to Minimize Energy Consumption under RealTime Constraints.RR9300Inria Grenoble RhôneAlpes, Université de Grenoble ; Université Grenoble  AlpesNovember 2019, 23
 82 inproceedingsSurvival of the strictest: Stable and unstable equilibria under regularized learning with partial information.COLT 2021  34th Annual Conference on Learning TheoryBoulder, United StatesAugust 2021, 130
 83 articleVisualizing the performance of parallel programs.IEEE software85The paper presents Paragraph.1991
 84 inproceedingsPredicting the Energy Consumption of MPI Applications at Scale Using a Single Node.Cluster 2017IEEEHawaii, United StatesSeptember 2017
 85 inproceedingsLogGOPSim  Simulating LargeScale Applications in the LogGOPS Model.ACM Workshop on LargeScale System and Application Performance2010
 86 inproceedingsThe limits of minmax optimization algorithms: Convergence to spurious noncritical sets.ICML 2021  38th International Conference on Machine LearningVienna, AustriaJuly 2021
 87 articleScaling applications to massively parallel machines using Projections performance analysis tool.Future Generation Comp. Syst.2232006
 88 inproceedingsUsing Simulation to Evaluate and Tune the Performance of Dynamic Load Balancing of an Overdecomposed Geophysics Application.EuroPar 2017: 23rd International European Conference on Parallel and Distributed ComputingSantiago de Compostela, SpainAugust 2017, 15
 89 articlePerformance Modeling of a Geophysics Application to Accelerate the Tuning of Overdecomposition Parameters through Simulation.Concurrency and Computation: Practice and Experience2018, 121
 90 inproceedingsASGriDS: Asynchronous SmartGrids Distributed Simulator.APPEEC 2019  11th IEEE PES AsiaPacific Power and Energy Engineering ConferenceMacao, Macau SAR ChinaIEEEDecember 2019, 15
 91 inproceedingsSelection Problems in the Presence of Implicit Bias.Proceedings of the 9th Innovations in Theoretical Computer Science Conference (ITCS)2018, 33:133:17
 92 inproceedingsAdapting Batch Scheduling to Workload Characteristics: What can we expect From Online Learning?IPDPS 2019  33rd IEEE International Parallel & Distributed Processing SymposiumRio de Janeiro, BrazilIEEEMay 2019, 686695
 93 articleThe importance of memory for price discovery in decentralized markets.Games and Economic Behavior125January 2021, 6278
 94 inproceedingsCollisions groupées lors du mécanisme d'évitement de collisions de CPLG3.CoRes 2020  Rencontres Francophones sur la Conception de Protocoles, l’Évaluation de Performance et l’Expérimentation des Réseaux de CommunicationLyon, FranceSeptember 2020, 14
 95 inproceedingsOptimistic Mirror Descent in SaddlePoint Problems: Going the Extra (Gradient) Mile.ICLR 2019  7th International Conference on Learning RepresentationsNew Orleans, United StatesMay 2019, 123
 96 inproceedingsQuick or cheap? Breaking points in dynamic markets.EC 2020  21st ACM Conference on Economics and ComputationBudapest, HungaryJuly 2020, 132
 97 inproceedingsCycles in adversarial regularized learning.SODA '18  TwentyNinth Annual ACMSIAM Symposium on Discrete AlgorithmsNew Orleans, United StatesJanuary 2018, 27032717
 98 articleLearning in games via reinforcement learning and regularization.Mathematics of Operations Research414November 2016, 12971324
 99 articleRiemannian game dynamics.Journal of Economic Theory177September 2018, 315364
 100 inproceedingsForgetting the Forgotten with Lethe: Conceal Content Deletion from Persistent Observers.PETS 2019  19th Privacy Enhancing Technologies SymposiumStockholm, SwedenJuly 2019, 121
 101 articleVAMPIR: Visualization and Analysis of MPI Resources.Supercomputer1211996
 102 inproceedingsExploiting system level heterogeneity to improve the performance of a GeoStatistics multiphase taskbased application.ICPP 2021  50th International Conference on Parallel ProcessingLemont, United StatesAugust 2021, 110
 103 techreportA vaccination policy by zones.Think tank Terra NovaOctober 2020
 104 articleSARSCoV2 elimination, not mitigation, creates best outcomes for health, the economy, and civil liberties.The Lancet39710291June 2021, 22342236
 105 inproceedingsGreen bridges: Reconnecting Europe to avoid economic disaster.Europe in the Time of Covid192020
 106 inproceedingsPARAVER: A tool to visualise and analyze parallel code.Proceedings of Transputer and Occam Developments, WOTUG18.441995
 107 articleMarket sentiments and convergence dynamics in decentralized assignment economies.International Journal of Game Theory491March 2020, 275298
 108 techreportFocus mass testing: How to overcome low test accuracy.Esade Centre for Economic PolicyDecember 2020
 109 articleGreen zoning: An effective policy tool to tackle the Covid19 pandemic.Health Policy1258August 2021, 981986
 110 inproceedingsScalable performance analysis: the Pablo performance analysis environment.Scalable Parallel Libraries Conference1993
 111 thesisToward transparent and parsimonious methods for automatic performance tuning.UGA (Université Grenoble Alpes); USP (Universidade de São Paulo)July 2021
 112 inproceedingsThe eyes have it: A task by data type taxonomy for information visualizations.IEEE Symposium on Visual LanguagesIEEE1996
 113 inproceedingsPower Management and Dynamic Voltage Scaling: Myths and Facts.Proceedings of the 2005 Workshop on Power Aware Realtime ComputingNew Jersey, USASeptember 2005
 114 inproceedingsPotential for Discrimination in Online Targeted Advertising.FAT 2018  Conference on Fairness, Accountability, and Transparency81NewYork, United StatesFebruary 2018, 115
 115 inproceedingsPSINS: An Open Source Event Tracer and Execution Simulator for MPI Applications.EuroPar2009
 116 inproceedingsPrivacy Risks with Facebook’s PIIbased Targeting: Auditing a Data Broker’s Advertising Interface.39th IEEE Symposium on Security and Privacy (S&P)Proceedings of the 39th IEEE Symposium on Security and Privacy (S&P)San Francisco, United States2018
 117 inproceedingsCongestion Avoidance in LowVoltage Networks by using the Advanced Metering Infrastructure.ePerf 2018  IFIP WG PERFORMANCE  36th International Symposium on Computer Performance, Modeling, Measurements and EvalutionToulouse, FranceDecember 2018, 13
 118 inproceedingsDecentralized Optimization of Energy Exchanges in an Electricity Microgrid .ACM eEnergy 2016  7th ACM International Conference on Future Energy SystemsWaterloo, CanadaJune 2016
 119 inproceedingsDecentralized optimization of energy exchanges in an electricity microgrid.2016 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGTEurope)Ljubljana, SloveniaIEEEOctober 2016, 16
 120 inproceedingsScheduling for Reduced CPU Energy.Proceedings of the 1st USENIX Conference on Operating Systems Design and ImplementationOSDI '94USAMonterey, CaliforniaUSENIX Association1994, 2–es
 121 inproceedingsValidation and Uncertainty Assessment of ExtremeScale HPC Simulation through Bayesian Inference.EuroPar2013
 122 articleToward Scalable Performance Visualization with Jumpshot.International Journal of High Performance Computing Applications1331999
 123 inproceedingsBigSim: A Parallel Simulator for Performance Prediction of Extremely Large Parallel Machines.IPDPS2004
 124 articleRobust power management via learning and game design.Operations Research691January 2021, 331345