Since its creation in 2003, TAO activities had constantly but slowly evolved, as old problems were being solved, and new applications arose, bringing new fundamental issues to tackle. But recent abrupt progresses in Machine Learning (and in particular in Deep Learning) have greatly accelerated these changes also within the team. It so happened that this change of slope also coincided with some more practical changes in TAO ecosystem: following Inria 12-years rule, the team definitely ended in December 2016. The new team TAU (for TAckling the Underspecified) has been proposed, and formally created in July 2019. At the same time important staff changes took place, that also justify even sharper changes in the team focus. During the year 2018, the second year of this new era for the (remaining) members of the team, our research topics have now stabilized around a final version of the TAU project.

Following the dramatic changes in TAU staff during the years 2016-2017 (see the 2017 activity report of the team for the details), the research around continuous optimization has definitely faded out in TAU (while the research axis on hyperparameter tuning has focused on Machine Learning algorithms), the Energy application domain has slightly changed direction under Isabelle Guyon's supervision (Section 4.2), after the completion of the work started by Olivier Teytaud, and a few new directions have emerged, around the robustness of ML systems (Section 3.1.2). The other research topics have been continued, as described below.

As discussed by 146, and in the recent collaborative survey paper 120, the topic of ethical AI was non-existent until 2010, was laughed at in 2016, and became a hot topic in 2017 as the AI disruptivity with respect to the fabric of life (travel, education, entertainment, social networks, politics, to name a few) became unavoidable 138, together with its expected impacts on the nature and amount of jobs. As of now, it seems that the risk of a new AI Winter might arise from legal1 and societal2 issues. While privacy is now recognized as a civil right in Europe, it is feared that the GAFAM, BATX and others can already capture a sufficient fraction of human preferences and their dynamics to achieve their commercial and other goals, and build a Brave New Big Brother (BNBB), a system that is openly beneficial to many, covertly nudging, and possibly dictatorial).

The ambition of Tau is to mitigate the BNBB risk along several intricated dimensions, and build i) causal and explainable models; ii) fair data and models; iii) provably robust models.

Participants: Isabelle Guyon, Michèle Sebag, Philippe Caillou, Paola Tubaro

The extraction of causal models, a long goal of AI 143, 113, 144, became a strategic issue as the usage of learned models gradually shifted from prediction to prescription in the last years. This evolution, following Auguste Comte's vision of science (Savoir pour prévoir, afin de pouvoir) indeed reflects the exuberant optimism about AI: Knowledge enables Prediction; Prediction enables Control.
However, although predictive models can be based on correlations, prescriptions can only be based on causal models3.

Among the research applications concerned with causal modeling, predictive modeling or collaborative filtering at Tau are all projects described in section 4.1 (see also Section 3.4), studying the relationships between: i) the educational background of persons and the job openings (FUI project JobAgile and DataIA project Vadore); ii) the quality of life at work and the economic performance indicators of the enterprises (ISN Lidex project Amiqap) 115 ; iii) the nutritional items bought by households (at the level of granularity of the barcode) and their health status, as approximated from their body-mass-index (IRS UPSaclay Nutriperso); iv) the actual offer of restaurants and their scores on online rating systems.
In these projects, a wealth of data is available (though hardly sufficient for applications ii), iii and iv))) and there is little doubt that these data reflect the imbalances and biases of the world as is, ranging from gender to racial to economical prejudices. Preventing the learned models from perpetuating such biases is essential to deliver an AI endowed with common decency.

In some cases, the bias is known; for instance, the cohorts in the Nutriperso study are more well-off than the average French population, and the Kantar database includes explicit weights to address this bias through importance sampling. In other cases, the bias is only guessed; for instance, the companies for which Secafi data are available hardly correspond to a uniform sample as these data have been gathered upon the request of the company trade union.

Causal relationships are being identified using our recently published paper 18. This work will be continued, as TAU is a partner of the PEPR-IA project Causalit-AI (local PI Michèle Sebag), starting next Spring.

Participants: Guillaume Charpiat, Marc Schoenauer, Michèle Sebag

Due to their outstanding performances, deep neural networks and more generally machine learning-based decision making systems, referred to as MLs in the following, have been raising hopes in the recent years to achieve breakthroughs in critical systems, ranging from autonomous vehicles to defense. The main pitfall for such applications lies in the lack of guarantees for MLs robustness.

Specifically, MLs are used when the mainstream software design process does not apply, that is, when no formal specification of the target software behavior is available and/or when the system is embedded in an open unpredictable world. The extensive body of knowledge developed to deliver guarantees about mainstream software

These downsides, currently preventing the dissemination of MLs in safety-critical systems (SCS), call for a considerable amount of research, in order to understand when and to which extent an MLs can be certified to provide the desired level of guarantees.

This activity has been put on hold in the team this year, but will be revived in the PEPR-IA SAIF project (starting next Spring) in which TAU is a partner (local PI Guillaume Charpiat).

Participants: Guillaume Charpiat, Cécile Germain, Isabelle Guyon, Marc Schoenauer, Michèle Sebag

In sciences and engineering, human knowledge is commonly expressed in closed form, through equations or mechanistic models characterizing how a natural or social phenomenon, or a physical device, will behave/evolve depending on its environment and external stimuli, under some assumptions and up to some approximations. The field of numerical engineering, and the simulators based on such mechanistic models, are at the core of most approaches to understand and analyze the world, from solid mechanics to computational fluid dynamics, from chemistry to molecular biology, from astronomy to population dynamics, from epidemiology and information propagation in social networks to economy and finance.

Most generally, numerical engineering supports the simulation, and when appropriate the optimization and control4 of the phenomenons under study, although several sources of discrepancy might adversely affect the results, ranging from the underlying assumptions and simplifying hypotheses in the models, to systematic experiment errors to statistical measurement errors (not to mention numerical issues). This knowledge and know-how are materialized in millions of lines of code, capitalizing the expertise of academic and industrial labs. These softwares have been steadily extended over decades, modeling new and more fine-grained effects through layered extensions, making them increasingly harder to maintain, extend and master. Another difficulty is that complex systems most often resort to hybrid (pluridisciplinary) models, as they involve many components interacting along several time and space scales, hampering their numerical simulation.

At the other extreme, machine learning offers the opportunity to model phenomenons from scratch, using any available data gathered through experiments or simulations. Recent successes of machine learning in computer vision, natural language processing and games, to name a few, have demonstrated the power of such agnostic approaches and their efficiency in terms of prediction 119, inverse problem solving 139, and sequential decision making 167, 168, despite their lack of any "semantic" understanding of the universe. Even before these successes, Anderson's claim was that the data deluge [might make] the scientific method obsolete71, as if a reasonable option might be to throw away the existing equational or software bodies of knowledge, and let Machine Learning rediscover all models from scratch. Such a claim is hampered among others by the fact that not all domains offer a wealth of data, as any academic involved in an industrial collaboration around data has discovered.

Another approach is considered in Tau, investigating how existing mechanistic models and related simulators
can be partnered with ML algorithms: i) to achieve the same goals with the same methods with a gain of accuracy or time; ii) to achieve new goals; iii) to achieve the same goals with new methods.

Toward more robust numerical engineering: In domains where satisfying mechanistic models and simulators are available, ML can contribute to improve their accuracy or usability. A first direction is to refine or extend the models and simulators to better fit the empirical evidence. The goal is to finely account for the different biases and uncertainties attached to the available knowledge and data, distinguishing the different types of known unknowns. Such known unknowns include the model hyper-parameters (coefficients), the systematic errors due to e.g., experiment imperfections, and the statistical errors due to e.g., measurement errors. A second approach is based on learning a surrogate model for the phenomenon under study that incorporate domain knowledge from the mechanistic model (or its simulation). See Section 8.5 for case studies.

A related direction, typically when considering black-box simulators, aims to learn a model of the error, or equivalently, a post-processor of the software. The discrepancy between simulated and empirical results, referred to as reality gap128, can be tackled in terms of domain adaptation 75, 101. Specifically, the source domain here corresponds to the simulated phenomenon, offering a wealth of inexpensive data, and the target domain corresponds to the actual phenomenon, with rare and expensive data; the goal is to devise accurate target models using the source data and models.

Extending numerical engineering: ML, using both experimental and numerical data, can also be used to tackle new goals, that are beyond the current state-of-the-art of standard approaches. Inverse problems are such goals, identifying the parameters or the initial conditions of phenomenons for which the model is not differentiable, or amenable to the adjoint state method.

A slightly different kind of inverse problem is that of recovering the ground truth when only noisy data is available. This problem can be formulated as a search for the simplest model explaining the data. The question then becomes to formulate and efficiently exploit such a simplicity criterion.

Another goal can be to model the distribution of given quantiles for some system: The challenge is to exploit available data to train a generative model, aimed at sampling the target quantiles.

Examples tackled in TAU are detailed in Section 8.5. Note that the "Cracking the Glass Problem", described in Section 8.2.3 is yet another instance of a similar problem.

Data-driven numerical engineering: Finally, ML can also be used to sidestep numerical engineering limitations in terms of scalability, or to build a simulator emulating the resolution of the (unknown) mechanistic model from data, or to revisit the formal background.

When the mechanistic model is known and sufficiently accurate, it can be used to train a deep network on an arbitrary set of (space,time) samples, resulting in a meshless numerical approximation of the model 155, supporting by construction differentiable programming125.

When no mechanistic model is sufficiently efficient, the model must be identified from the data only. Genetic programming has been used to identify systems of ODEs 154, through the identification of invariant quantities from data, as well as for the direct identification of control commands of nonlinear complex systems, including some chaotic systems 94. Another recent approach uses two deep neural networks, one for the state of the system, the other for the equation itself 147. The critical issues for both approaches include the scalability, and the explainability of the resulting models. Such line of research will benefit from TAU unique mixed expertise in Genetic Programming and Deep Learning.

According to Ali Rahimi's test of times award speech at NIPS 17, the current ML algorithms have become a form of alchemy. Competitive testing and empirical breakthroughs gradually become mandatory for a contribution to be acknowledged; an increasing part of the community adopts trials and errors as main scientific methodology, and theory is lagging behind practice. This style of progress is typical of technological and engineering revolutions for some; others ask for consolidated and well-understood theoretical advances, saving the time wasted in trying to build upon hardly reproducible results.

Basically, while practical achievements have often passed the expectations, there exist caveats along three dimensions. Firstly, excellent performances do not imply that the model has captured what was to learn, as shown by the phenomenon of adversarial examples. Following Ian Goodfellow, some well-performing models might be compared to Clever Hans, the horse that was able to solve mathematical exercizes using non verbal cues from its teacher 112; it is the purpose of Pillar I. to alleviate the Clever Hans trap (section 3.1).

Secondly, some major advances, e.g. related to the celebrated adversarial learning 106, 101, establish proofs of concept more than a sound methodology, where the reproducibility is limited due to i) the computational power required for training (often beyond reach of academic labs); ii) the numerical instabilities (witnessed as random seeds happen to be found in the codes); iii) the insufficiently documented experimental settings. What works, why and when is still a matter of speculation, although better understanding the limitations of the current state of the art is acknowledged to be a priority. After Ali Rahimi again, simple experiments, simple theorems are the building blocks that help us understand more complicated systems. Along this line, 135 propose toy examples to demonstrate and understand the defaults of convergence of gradient descent adversarial learning.

Thirdly, and most importantly, the reported achievements rely on carefully tuned learning architectures and hyper-parameters. The sensitivity of the results to the selection and calibration of algorithms has been identified since the end 80s as a key ML bottleneck, and the field of automatic algorithm selection and calibration, referred to as AutoML or Auto-

Tau aims to contribute to the ML evolution
toward a more mature stage along three dimensions. In the short term, the research done in Auto-depending on the nature and amount of the available data (section 3.3.2).
In the longer term, our goal is to leverage the methodologies forged in statistical physics to understand and control the trajectories of complex learning systems
(section 3.3.3).

Participants: Isabelle Guyon, Marc Schoenauer, Michèle Sebag

The so-called Auto-

Several approaches have been used to tackle Auto-any problem instance characterized from its meta-feature values 150, 73, 105. Collaborative filtering, considering that a problem instance "likes better" an algorithm configuration yielding a better performance, learns to recommend good algorithms to problem instances 157, 132. Bayesian optimization proceeds by alternatively building a surrogate model of algorithm performances on the problem instance at hand, and tackling it 97. This last approach currently is the prominent one; as shown in 132, the meta-features developed for AutoML are hardly relevant, hampering both meta-learning and collaborative filtering. The design of better features is another long-term research direction, in which TAU has recently been 93, ans still is very active. more recent approach used in TAU 148 extends the Bayesian Optimization approach with a Multi-Armed Bandit algorithm to generate the full Machine Learning pipeline, competing with the famed AutoSKLearn 97 (see Section 8.2.1).

Participants: Guillaume Charpiat, Marc Schoenauer, Michèle Sebag

In the 60s, Kolmogorov and Solomonoff provided a well-grounded theory for building (probabilistic) models best explaining the available data 151, 107, that is, the shortest programs able to generate these data. Such programs can then be used to generate further data or to answer specific questions (interpreted as missing values in the data). Deep learning, from this viewpoint, efficiently explores a space of computation graphs, described from its hyperparameters (network structure) and parameters (weights). Network training amounts to optimizing these parameters, namely, navigating the space of computational graphs to find a network, as simple as possible, that explain the past observations well.

This vision is at the core of variational auto-encoders 117, directly optimizing a bound on the Kolmogorov complexity of the dataset. More generally variational methods provide quantitative criteria to identify superfluous elements (edges, units) in a neural network, that can potentially be used for structural optimization of the network (Leonard Blier's PhD, started Oct. 2018).

The same principles apply to unsupervised learning, aimed to find the maximum amount of structure hidden in the data, quantified using this information-theoretic criterion.

The known invariances in the data can be exploited to guide the model design (e.g. as translation invariance leads to convolutional structures, or LSTM is shown to enforce the invariance to time affine transformations of the data sequence 160). Scattering transforms exploit similar principles 81.
A general theory of how to detect unknown invariances in the data, however, is currently lacking.

The view of information theory and Kolmogorov complexity suggests that key program operations (composition, recursivity, use of predefined routines) should intervene when searching for a good computation graph. One possible framework for exploring the space of computation graphs with such operations is that of Genetic Programming. It is interesting to see that evolutionary computation appeared in the last two years among the best candidates to explore the space of deep learning structures 149, 126. Other approaches might proceed by combining simple models into more powerful ones, e.g. using “Context Tree Weighting” 162 or switch distributions 96. Another option is to formulate neural architecture design as a reinforcement learning problem 74; the value of the building blocks (predefined routines) might be defined using e.g., Monte-Carlo Tree Search. A key difficulty is the computational cost of retraining neural nets from scratch upon modifying their architecture; an option might be to use neutral initializations to support warm-restart.

Participants: Cyril Furtlehner, Aurélien Decelle, François Landes, Michèle Sebag

Methods and criteria from statistical physics have been widely used in ML. In early days, the capacity of Hopfield networks (associative memories defined by the attractors of an energy function) was investigated by using the replica formalism 70. Restricted Boltzmann machines likewise define a generative model built upon an energy function trained from the data. Along the same lines, Variational Auto-Encoders can be interpreted as systems relating the free energy of the distribution, the information about the data and the entropy (the degree of ignorance about the micro-states of the system) 161. A key promise of the statistical physics perspective and the Bayesian view of deep learning is to harness the tremendous growth of the model size (billions of weights in recent machine translation netwowrks), and make them sustainable through e.g. posterior drop-out 136, weight quantization and probabilistic binary networks 131. Such "informational cooling" of a trained deep network can reduce its size by several orders of magnitude while preserving its performance.

Statistical physics is among the key expertises of Tau, originally only represented by Cyril Furtlehner, later strenghtened by Aurélien Decelle's and François Landes' arrivals in 2014 and 2018. On-going studies are conducted along several directions.

Generative models are most often expressed in terms of a Gibbs distributions

Another direction, explored in TAO/TAU in the recent years, is based on the definition and exploitation of self-consistency properties, enforcing principled divide-and-conquer resolutions. In the particular case of the message-passing Affinity Propagation algorithm for instance 165, self-consistency imposes the invariance of the solution when handled at different scales, thus enabling to characterize the critical value of the penalty and other hyper-parameters in closed form (in the case of simple data distributions) or empirically otherwise 100.

A more recent research direction examines the quantity of information in a (deep) neural net along the random matrix theory framework 84. It is addressed in Giancarlo Fissore's PhD, and is detailed in Section 8.2.3.

Finally, we note the recent surge in using ML to address fundamental physics problems: from turbulence to high-energy physics and soft matter (with amorphous materials at its core) 123 or atrophysics/cosmology as well. TAU's dual expertise in Deep Networks and in
statistical physics places it in an ideal position to significantly contribute to this domain and shape the methods
that will be used by the physics community in the future. In that direction, the PhD thesis of Marion Ullmo and Tony Bonnaire applying statistical method coming either from deep learning or statistical physics to the task of inferring the structure of the cosmic web has show great succes with recents results discussed in Section 8.2.3.
François Landes' recent arrival in the team makes TAU a unique place for such interdisciplinary research, thanks to his collaborators from the Simons Collaboration Cracking the Glass Problem (gathering 13 statistical physics teams at the international level).
This project is detailed in Section 8.2.3.

Independently, François Landes is actively collaborating with statistical physicists (Alberto Rosso, LPTMS, Univ. Paris-Saclay) and physcists at the frontier with geophysics (Eugenio Lippiello, Second Univ. of Naples) 127, 145. A CNRS grant (80Prime) finances a shared PhD (Vincenzo Schimmenti), at the frontier between seismicity and ML (Alberto Rosso, Marc Schoenauer and François Landes).

Participants: Cécile Germain, Isabelle Guyon, Marc Schoenauer, Michèle Sebag

Challenges have been an important drive for Machine Learning research for many years, and TAO members have played important roles in the organization of many such challenges: Michèle Sebag was head of the challenge programme in the Pascal European Network of Excellence (2005-2013); Isabelle Guyon, as mentioned, was the PI of many challenges ranging from causation challenges 108, to AutoML 109. The Higgs challenge 69, most attended ever Kaggle challenge, was jointly organized by TAO (C. Germain), LAL-IN2P3 (D. Rousseau and B. Kegl) and I. Guyon (not yet at TAO), in collaboration with CERN and Imperial College.

Many challenges have been organized in the recent years on the Codalab platform, managed by Isabelle Guyon and maintained at LISN. See details in Section 8.6.

Participants: Philippe Caillou, Isabelle Guyon, Michèle Sebag, Paola Tubaro

Collaboration: Jean-Pierre Nadal (EHESS); Marco Cuturi, Bruno Crépon (ENSAE); Thierry Weil (Mines); Jean-Luc Bazet (RITM)

Computational Social Sciences (CSS) studies social and economic phenomena, ranging from technological innovation to politics, from media to social networks, from human resources to education, from inequalities to health. It combines perspectives from different scientific disciplines, building upon the tradition of computer simulation and modeling of complex social systems 104 on the one hand, and data science on the other hand, fueled by the capacity to collect and analyze massive amounts of digital data.

The emerging field of CSS raises formidable challenges along three dimensions. Firstly, the definition of the research questions, the formulation of hypotheses and the validation of the results require a tight pluridisciplinary interaction and dialogue between researchers from different backgrounds. Secondly, the development of CSS is a touchstone for ethical AI. On the one hand, CSS gains ground in major, data-rich private companies; on the other hand, public researchers around the world are engaging in an effort to use it for the benefit of society as a whole 124. The key technical difficulties related to data and model biases, and to self-fulfilling prophecies have been discussed in section 3.1. Thirdly, CSS does not only regard scientists: it is essential that the civil society participate in the science of society 156.

Tao was involved in CSS for the last five years, and its activities have been strengthened thanks to P. Tubaro's and I. Guyon's expertises respectively in sociology and economics, and in causal modeling. Their departures will impact the team activities in this domain, but many projects are tsill on-going and CSS will remain a domain of choice. Details are given in Section 8.3.

Participants: Isabelle Guyon, Marc Schoenauer, Michèle Sebag

Collaboration: Rémy Clément, Antoine Marot, Patrick Panciatici (RTE), Vincent Renault (Artelys), Thibault Faney (IFPEN)

Energy Management has been an application domain of choice for Tao since the mid 2000s, with main partners SME Artelys (METIS Ilab INRIA; ADEME projects POST and NEXT), RTE (three CIFRE PhDs), and IFPEN (bilateral contract, DATAIA project ML4CFD). The goals concern i) optimal planning over several spatio-temporal scales, from investments on continental Europe/North Africa grid at the decade scale (POST), to daily planning of local or regional power networks (NEXT); ii) monitoring and control of the French grid enforcing the prevention of power breaks (RTE); iii) improvement of house-made numerical methods using data-intense learning in all aspects of IFPEN activities (Section 3.2).

The daily maintainance of power grids requires the building of approximate predictive models on the top of any given network topology. Deep Networks are natural candidates for such modelling, considering the size of the French grid (

Furthermore, predictive models of local grids are based on the estimated consumption of end-customers: Linky meters only provide coarse grain information due to privacy issues, and very few samples of fine-grained consumption are available (from volunteer customers). A first task is to transfer knowledge from small data to the whole domain of application. A second task is to directly predict the peaks of consumption based on the user cluster profiles and their representativity (see Section 8.4.2).

Participants: Guillaume Charpiat, Cécile Germain, Isabelle Guyon, Flora Jay, Marc Schoenauer, Michèle Sebag

As said (section 3.2), in domains where both first principle-based models and equations, and empirical or simulated data are available, their combined usage can support more accurate modelling and prediction, and when appropriate, optimization, control and design, and help improving the time-to-design chain through fast interactions between the simulation, optimization, control and design stages. The expected advances regard: i) the quality of the models or simulators (through data assimilation, e.g. coupling first principles and data, or repairing/extending closed-form models); ii) the exploitation of data derived from different distributions and/or related phenomenons; and, most interestingly, iii) the task of optimal design and the assessment of the resulting designs.

A first challenge regards the design of the model space, and the architecture used to enforce the known domain properties (symmetries, invariance operators, temporal structures). When appropriate, data from different distributions (e.g. simulated vs real-world data) will be reconciled, for instance taking inspiration from real-valued non-volume preserving transformations 90 in order to preserve the natural interpretation.

Another challenge regards the validation of the models and solutions of the optimal design problems. The more flexible the models, the more intensive the validation must be. Along this way, generative models will be used to support the design of "what if" scenarios, to enhance anomaly detection and monitoring via refined likelihood criteria.

In the application domains described by Partial Differential Equations (PDEs), the goal of incorporating machine learning into classical simulators is to speed up the simulations while maintaining as much as possible the accuracy ad physical relevance of the proposed solutions. Many possible tracks are possible for this; one can build surrogate models, either of the whole system, or of its most computationaly costly parts; one can search to provide better initialization heuristics to solvers, which make sure that physical constraints are satisfied. Or one can inject physical knowledge/constraints at different stages of the numerical solver.

Thanks to the pandemia, the impact of our activities regarding carbon footprint have decreased a lot, from our daily commute that have almost completely disappeared as we all switched to tele-working to the transformation of all conferences and workshops into virtual events. We all miss the informal discussions that took place during coffee breaks in the lab as well as during conferences. But when the pandemia vanishes, after the first moments of joy when actually meeting again physically with our colleagues, we will have to think of a new model for the way we work: we were indeed discussing before the pandemia about how to reduce the carbon footpring of the conferences, but now we know that there exist solutions, even though not perfect.

All our work on Energy (see Sections 4.2) is ultimately targeted toward optimizing the distribution of electricity, be it in planning the investments in the power network by more accurate previsions of user consumption, or helping the operators of RTE to maintain the French Grid in optimal conditions.

Herilalaina Rakotoarison, Louisot Milijaona, Andry Rasoanaivo, Michèle Sebag, Marc Schoenauer, Spotlight paper at ICLR (top 5% submissions) for the paper 31 Learning Meta-features for AutoML. International Conference on Learning Representations, 2022.

Isabelle Guyon Keynote, NeurIPS 2022 : The Data-centric Era: How ML is becoming an experimental science.

Herilalaina Rakotoarison, First prize ex æquo at the 2022 PhD Prize in Computer Science, awarded by Labex DigiCosme, the CS doctoral school of Université Paris Saclay and the doctoral school of IPP (i.e., the whole ”Plateau de Saclay” in CS).

Participants: Philippe Caillou, Isabelle Guyon, Michèle Sebag

PhDs: Armand Lacombe, Cyriaque Rousselot, Nicolas Atienza

Post-doc: Shuyu Dong, Shiyang Yan

Collaboration: Olivier Allais (INRAE); Julia Mink (Univ. Bonn); Jean-Pierre Nadal & Annick Vignes (CAMS, EHESS); David Lopez-Paz (Facebook).

This year, the long awaited journal version of SAM (Structural Agnostic Modelling), has been published in JMLR 18, long after Diviyan Kalainathan's PhD 114. The causal modelling activity continues with three main directions in 2022.
The first one is tackled in collaboration with INRAE (Cyriaque Rousselot's PhD), within the Horapest DataIA project. The goal is to assess the causal effects of the diffusion of pesticides in French residential areas, through exploiting the data from the Health Data Hub together with the newly available dataset reporting the concentrations of diverse molecules in 50 stations on a weekly basis (CNEP), and the overall amount of products bought yearly in every postal code (BNVD). The potential effects that will be investigated concern the children' health in the 2019-2022 period, born between 2013 and 2019. The study will contrast the children resident in places with high or low pesticide average concentration on average, and the children with high or low pesticide concentration in utero. Besides getting the data5 the difficulty lies in observational causal modelling from spatio-temporal data with hidden confounders.
A second direction is explored in partnership with Fujitsu (Shuyu Dong's postdoc). The goal is to achieve linear Structural Equation Model (SEM)identification from observational data in the large

Finally, causality is also at the core of TAU participation in the INRIA Challenge OceanIA, that started in 2021 153. Shiyang Yan's post-doc is dedicated to out-of-distribution learning, motivated by the analysis of the TARA images to identify the ecosystems in the diverse sites of the data collection. The high imbalance of the data among the classes, the prevalence of outliers, are handled using generalized contrastive losses and introducing fake outliers extracted from face images, or created as chimeras.

Other motivating applications for causal modeling are described in section 4.1.

Participants: Isabelle Guyon, François Landes, Alessandro Leite, Marc Schoenauer, Michèle Sebag

PhD: Cyriaque Rousselot

Collaboration: MyDataModels; Thales

In Roman Bresson’s PhD 46, (coll. LISN-GALAC, Thales, U. Paderborn; pending patent Bresson-Labreuche-Sebag-Cohen), the goal was to adapt a neuronal architecture to yield an interpretable-by-design model. The extension of this approach is investigated to transform an accurate black-box into a hierarchical choquet integral.

The team is also involved in the proposal for the IPL HyAIAI (Hybrid Approaches for Interpretable AI), coordinated by the LACODAM team (Rennes) dedicated to the design of hybrid approaches that combine state of the art numeric models (e.g., deep neural networks) with explainable symbolic models, in order to be able to integrate high level (domain) constraints in ML models, to give model designers information on ill-performing parts of the model, to provide understandable explanations on its results. On-going collaboration with the Multispeech team in Nancy is concerned with co-supervision of G. Zervakis' PhD (to be defended March 2023), and concerns the use of background knowledge to improve the performances of foundational models in NLP 164 and an analogy based approach for solving target sense verification 34.

An original approach to DNN explainability might arise from the study of structural glasses (8.2.3), with a parallel to Graph Neural Networks (GNNs), that could become an excellent non-trivial example for developing explainability protocols, as we already suggest from results in 66.

Build on collaboration with Raymond Poincaré Hospital, the team is developing tools to increase the interpretability of medical data in applicative context. A first study published in 17 investigates how geometric methods could represent the evolution of patients’ key indicators on a curved manifold to generate meaningful and interpretable representation. These representations could be generalized with minor modifications to temporal data.

Genetic Programming 72 is an Evolutionary Computing technique that evolves models as analytical expressions (Boolean formulae, functions, LISP-like code), that are hopefully easier to understand than black-box NNs with hundreds of thousands of weights. This idea has been picked up by the European FET project TRUST-AI (Transparent, Reliable and Unbiased Smart Tool for AI) that started in October 2020. Alessandro Leite joined the project (and the TAU team) in February 2021 on an ARP position. First work addressed explainable reinforcement learning using GP 33. Current work, recently acepted to EuroGP 2023, concerns the adaptation of the Memetic Semantic Generic Programming 98 to the continuous case. Furthermore, a collaboration with O. Teytaud (Meta), around the follow-up is Mathurin's CIFRE PhD (started Oct. 2022).

Participants: Guillaume Charpiat, Marc Schoenauer, Michèle Sebag

PhDs: Roman Bresson

Collaboration: Johanne Cohen (LISN-GALAC) and Christophe Labreuche (Thalès); Eyke Hullermeier (U. Paderborn, Germany).

Though several on-going activitires in this domain have been put on hold (see Section 3.1.2), new research lines have started to emerge, pertaining to robustness.

The first one, already described in section 8.1.2, concerns the indentifiability of the neural net implementing a hierarchical Choquet integral, in the large sample limit.

Another direction, part of A. Lacombe’s on-going PhD, is concerned with privacy. Our primary motivation was to contribute to the understanding of the pandemy, with no former collaboration with hospitals, and therefore, no access to real data. An approach was developed to achieve excessively private learning through a differential-privacy compliant access to the only marginals of the data 122.

We have also explored relationships between theoretical guarantees provided by differential privacy and membership interence attacks 30, as described in Section 8.5.3.

Participants: Guillaume Charpiat, Isabelle Guyon, Marc Schoenauer, Michèle Sebag

PhDs: Léonard Blier, Adrien Pavao, Herilalaina Rakotoarison, Hoazhe Sun, Manon Verbockhaven, Romain Egele

Collaborations: Vincent Renault (SME Artelys); Yann Ollivier (Facebook); Wei-Wei Tu (4Paradigm, Chine); André Elisseeff (Google Zurich); Prasanna Balaprakash (Argonne National labs), among others (for a full list see https://autodl.chalearn.org/ and https://metalearning.chalearn.org/)

Auto-Tau investigate several research directions.

After proposing MOSAIC 148, that extends and adapts Monte-Carlo Tree Search to explore the structured space of pre-processing + learning algorithm configurations, and performs on par with AutoSklearn, the winner of Auto-

Heri also contributed to a large benchmarking effort together with Olivier Teytaud, former member of the team, now with Facebook AI Research 19.

In a second direction, with the internship and starting PhD thesis of Manon Verbockhaven, we adopt a functional analysis viewpoint in order to adapt on the fly the architecture of neural networks that are being trained. This allows to start training neural networks with very few neurons and layers, and add them where they are needed, instead of training huge architectures and then pruning them, a common practice in deep learning, for optimization reasons. For this, we quantify the lack of expressivity of a neural network being trained, by analyzing the difference between how the backpropagation would like the activations to change and what the tangent space of the parameters offers as possible activation variations. We can then localize the lacks of expressivity, and add neurons accordingly. It turns out that the optimal weights of the added neurons can be computed in closed form.

A last direction of investigation concerns the design of challenges, that contribute to the collective advance of research in the Auto-

Sef-supervised learning seems to be an avenue with great future, allowing to train reprsentations without costly human labeling. A new challenge accepted as part of the WCCI competition program 2022 is currently running. Another challenge on Neural Architecture Search (NAS) has been run together with a workshop at the CVPR 2021 conference. Preliminary results on NAS have been produced by one of our interns (Romain Egele 95). Further developments have led to effective algorithms to conduct simultaneously NAS and hyper-parameter selection 27, 57. More details on challenges are found in Section 8.6).

Participants: Guillaume Charpiat, Isabelle Guyon, Marc Schoenauer, Michèle Sebag

PhDs: Léonard Blier, Zhengying Liu, Adrien Pavao, Haozhe Sun, Romain Egele

Collaboration: Yann Ollivier (Facebook AI Research, Paris)

Although a comprehensive mathematical theory of deep learning is yet to come, theoretical insights from information theory or from dynamical systems can deliver principled improvements to deep learning and/or explain the empirical successes of some architectures compared to others.

During his CIFRE PhD with Facebook AI Research Paris, co-supervised by Yann Ollivier (former TAU member) 45 , Léonard Blier has properly formalized the concepts of successor states and multi-goal functions78, in particular in the case of continuous state spaces. This allowed him to define unbiased algorithms with finite variance to learn such ojects, including the continuous case thanks to approximation functions. In the case of finite environments, new convergence bounds have been obtained for the learning of the value function. These new algorithms capable of learning successor states in turn lead to define and learn new representations for the state space.

The AutoDL challenges, co-organized in TAU (in particular by Isabelle Guyon), also contribute to a better understanding of Deep Learning. It is interesting to note that no Neural Architecture Search algorithm was proposed to solve the different challenges in AutoDL (corresponding to different data types). See section 8.6 for more details.

Our PhD student Haozhe Sun is continuing to work on the problem of modularity in Deep Learning. He wrote a survey under revision and recently submitted a paper on novel algorithms for low-cost AI exploiting modularity. The current trend in Artificial Intelligence (AI) is to heavily rely on systems capable of learning from examples, such as Deep learning (DL) models, a modern embodiment of artificial neural networks. While numerous applications have made it to market in recent years (including self-driving cars, automated assistants, booking services, and chatbots, improvements in search engines, recommendations, and advertising, and heath-care applications, to name a few) DL models are still notoriously hard to deploy in new applications. In particular, the require massive numbers of training examples, hours of GPU training, and highly qualified engineers to hand-tune their architectures. This thesis will contribute to reduce the barrier of entry in using DL models for new applications, a step towards "democratizing AI".

Romain Egele in his PhD in collaboration with Argonne National Labs (USA), is been actively working on Neural Architecture Search (NAS). He developed a package called DeepHyper, allowing users to conduct NAS with genetic algorithms using TensorFlow or PyTorch, the principal Deep Learning frameworks 27. His contributions include applying Recurrent Neural Network Architecture Search for Geophysical Emulation and Scalable Reinforcement-Learning-Based Neural Architecture Search for Cancer Deep Learning Research. In 2022 he published two papers on hyper-parameter optimization. In 57, he proposes to exploit the parallelism in large compute clusters to speed up Bayesian hyper-parameter search in an effective way. In 26 he extends Bayesan optimization algorithms to perform accurate estimations of uncertainties.

Participants: Cyril Furtlehner, Aurélien Decelle, François Landes, Guillaume Charpiat

PhDs: Giancarlo Fissore, Marion Ullmo

Collaboration: Jacopo Rocchi (LPTMS Paris Sud); the Simons team: Rahul Chako (post-doc), Andrea Liu (UPenn), David Reichman (Columbia), Giulio Biroli (ENS), Olivier Dauchot (ESPCI).; Clément Vignax (EPFL);
Yufei Han (Symantec), Nabila Aghanim (Institut d'Astrophysique Spatiale), Tony Bonnaire (ENS Paris).

Generative models constitute an important piece of unsupervised ML techniques which is still under rapid developpment. In this context insights from statistical physics are relevent in particular for energy based models like restricted Boltzmann machines. The information content of a trained restricted Boltzmann machine (RBM) and its learning dynamics can be analyzed precisely with help of ensemble averaging techniques 87, 88. More insight can be obtained by looking at data of low intrinsic dimension, where exact solutions of the RBM can be obtained 89, thanks to a convex relaxation. In particular we have found a 1st order transition mechanisms that may plague the learning in a more advanced part of the learning.In 11 we investigate further this question and show that sampling the equilibrium distribution using the Markov chain Monte Carlo method can be dramatically accelerated when using biased sampling techniques, in particular the Tethered Monte Carlo (TMC) method. This sampling technique can also be used to improve the computation of the log-likelihood gradient during training, leading to dramatic improvements in training RBMs with artificial clustered datasets. On real low-dimensional datasets, this new training method fits RBM models with significantly faster relaxation dynamics than those obtained with standard PCD recipes. Learning dynamics has also been adressed in a different context of feature learning processes in 59 where closed form expressions are obtained for train and test errors via random matrix theory, a characterization of good alignment between the features and the signal and the derivation of a set autonomous equations driving the process at large scale. We have also investigated models of decision-making with approaches grounded in statistical physics, that are able to predict experimental observed data 68. A last point concerns road traffic forecasting, a long standing application of mean-field inference methods based on probabilistic modelling. In 16 we wrap up some of the techniques developed in past works and perform comprehensive experimental tests on various real world Urban traffic dataset, thanks to PTV-SISTeMA, showing the effectiveness of our method.

As mentioned earlier, the use of ML to address fundamental physics problems is quickly growing. In that direction in the context of T. Bonnaire's PhD 80, is undertaken in 12 the first comprehensive and quantitative real-space analysis of the cosmological information content in the environments of the cosmic web (voids, filaments, walls, and nodes) up to non-linear scales, k=0.5 h/Mpc, using a method based on the Gaussian mixture model with a prior forcing the centers to "live" on a tree-graph 13 79, 85. This method has been further developed for handling in particular possible outliers and put into a general formalism 86.

Another place where ML can help address fundamental physics questions is the domain of glasses 40 (how the structure of glasses is related to their dynamics), which is of the major problems in modern theoretical physics (glasses are a key part of the carrer of Giorgio Parisi, a 2021 nobel prize laureate). The idea is to let ML models automatically find the hidden structures (features) that control the flowing or non-flowing state of matter, discriminating liquid from solid states. These models can then help identifying "computational order parameters", that would advance the understanding of physical phenomena 123, 85, on the one hand, and support the development of more complex models, on the other hand. Attacking the problem of amorphous condensed matter by novel Graph Neural Networks (GNN) architectures is a very promising lead, regardless of the precise quantity one may want to predict. Currently GNNs are engineered to deal with molecular systems and/or crystals, but not to deal with amorphous matter. This strategy is currently being investigated by Francesco Pezzicoli (PhD student), who has already demonstrated the generalizing abilities of rotation-equivariant GNNs 66.

Computational Social Sciences (CSS) is making significant progress in the study of social and economic phenomena thank to the combination of social science theories and new insight from data science. While the simultaneous advent of massive data and unprecedented computational power has opened exciting new avenues, it has also raised new questions and challenges.

Several studies are being conducted in TAU, about labor (labor markets, the labor of human annotators for AI data, quality of life and economic performance), about nutrition (health, food, and socio-demographic issues), around Cartolabe, a platform for scientific information system and visual querying.

Participants: Philippe Caillou, Isabelle Guyon, Michèle Sebag, Paola Tubaro

PhDs: Guillaume Bied, Armand Lacombe, Assia Wirth

Engineers: Victor Alfonso Naya

Collaboration: Jean-Pierre Nadal (EHESS); Bruno Crépon (ENSAE); Antonio Casilli, Ulrich Laitenberger (Telecom Paris); Odile Chagny (IRES); Francesca Musiani, Mélanie Dulong de Rosnay (CNRS); José Luis Molina (Universitat Autònoma de Barcelona); Antonio Ortega (Universitat de València); Julian Posada (University of Toronto)

A first area of activity of TAU in Computational Social Sciences is the study of labor, from the functioning of the job market, to the rise of new, atypical forms of work in the networked society of internet platforms, and the quality of life at work.

Job markets
The DataIA project Vadore (partners ENSAE and Pôle Emploi) benefits from the sustained cooperation and from the wealth of data gathered by Pôle Emploi. The data management is regulated along a 3-partite convention (GENES-ENSAE, Univ Paris-Saclay, Pôle Emploi). Extensive efforts have been required to achieve the data pipelines required to enable learning recommendation models and exploiting them in a confidentiality preserving way (G. Bied's PhD). Primary online testing (beta-test campaigns) have assessed the suitability of the recommendations. A second round of testing at the region Rhone-Alpes scale will take place in 2023.

The learned models are inspected w.r.t. several criteria and requirements. A first criterion regards the robustness of the recommender performances under non-stationary distributions, e.g. due to the Covid pandemy 35. Another criterion concerns the congestion of the job market (share of job offers paid attention to by job seekers). Recommender systems tend to increase the congestion due to the so-called popularity bias. Early attempts to prevent the congestion were investigated in 77, 137, using optimal transport; and this direction will be pursued in S. Nathan's PhD.

A third criterion regards the fairness of the recommendation model. A gender-related gap in several utilities (wages, types of contract, distance-to-job) is observed by contrasting the jobs recommended to men and women, everything else being equal. Most interestingly, this gap parallels the gap among the jobs actually occupied by men and women (everything else being equal). Several directions of research are considered based on this fact, depending on the regulations to be enforced (in French or European public services). The first direction consists in integrating a risk-avoidance sub-model in the recommendation model, to decouple the prejudice effects (from the recruiters' decisions) and the social conditioning (from the job seekers' preferences). The second one aims at defining new population-based and individual-based performance criteria for a job recommender system.

A key difficulty for research on ML-based job recommendation is the lack of open and representative datasets, owing to the very sensitive nature of the data and the protection of vulnerable persons. We have co-organized a workshop (Feb. 2023) gathering researchers and industrials on this topic, in collaboration with Actiris and VDAB (public employment services in Belgium), to identify how this lack of open datasets, hindering the benchmarking of existing systems, can be addressed.

The human labor behind AI

We look at business-to-business platform labour 55 and more specifically at the data "micro-workers" who perform essential, yet marginalized and poorly paid tasks such as labeling objects in a photograph, translating or transcribing short texts, or recording utterances. Micro-workers are recruited through specialist intermediaries across supply chains that span the globe and reproduce inherited North-South outsourcing relationships 43. Further observed inequalities are gender-based 21. Despite the opportunity to telework, the COVID-19 pandemic has adversely affected these workers, widening the gap that separates them from the formally employed 20.

Current work extends this research to workers' skills, competencies and workplace learning practices in an environment in which they support machine learning 22, and to the resilience of these emerging labour markets 44

Participants: Philippe Caillou, Michèle Sebag

PhD: Armand Lacombe, Cyriaque Rousselot

Collaboration: Olivier Allais (INRA); Julia Mink (Univ. Bonn, DE).

Continuing our former partnership with INRAE (in the context of the Initiative de Recherche Stratégique Nutriperso; 103), we proposed the Horapest project to uncover the potential causal relationships between pesticide dissemination and children's health (Cyriaque Rousselot's PhD). The demand of access has been approved by the CNIL and the Health Data Hub; the data are expected in Sept. 2023, and contacts have been taken with the CHU Toulouse for cooperation on complementary data.

Participants: Philippe Caillou, Michèle Sebag

Engineers: Anne-Catherine Letournel, Victor Alfonso Naya

Collaboration: Jean-Daniel Fekete (AVIZ, Inria Saclay)

A third area of activity concerns the 2D visualisation and querying of a corpus of documents. Its initial motivation was related to scientific organisms, institutes or Universities, using their scientific production (set of articles, authors, title, abstract) as corpus.
The Cartolabe project (see also Section 7) started as an Inria ADT (coll. Tao and AVIZ, 2015-2017). It received a grant from CNRS (coll. Tau, AVIZ and HCC-LRI, 2018-2019).

The originality of the approach is to rely on the content of the documents (as opposed to, e.g. the graph of co-authoring and citations). This specificity allowed to extend Cartolabe to various corpora, such as Wikipedia, Bibliotheque Nationale de France, or the Software Heritage. Cartolabe was also applied in 2019 to the Grand Debat dataset: to support the interactive exploration of the 3 million propositions; and to check the consistency of the official results of the Grand Debat with the data. Cartolabe has also been applied in 2020 to the COVID-19 kaggle publication dataset (Cartolabe-COVID project) to explore these publications.

Among its intended functionalities are: the visual assessment of a domain and its structuration (who is expert in a scientific domain, how related are the domains); the coverage of an institute expertise relatively to the general expertise; the evolution of domains along time (identification of rising topics). A round of interviews with beta-user scientists has been performed in 2019-2020. Cartolabe usage raises questions at the crossroad of human-centered computing, data visualization and machine learning: i) how to deal with stressed items (the 2D projection of the item similarities poorly reflects their similarities in the high dimensional document space; ii) how to customize the similarity and exploit the users' feedback about relevant neighborhoods. A statement of the current state of the project was published in 2021 82.

Participants: Isabelle Guyon, Marc Schoenauer

PhDs: Balthazar Donon, Wenzhuo Liu

Collaboration: Rémi Clément, Patrick Panciatici (RTE)

Our collaboration with RTE, during Benjamin Donnot's (2016-2019) 91 and Balthazar Donon's 47 CIFRE PhDs, is centered on the maintainance of the national French Power Grid. In order to maintain the so-called "(n-1) safety" (see Section 4.2), fast simulations of the electrical flows on the grid are mandatory, that the home-brewed simulator HADES is too slow to provide. The main difficulty of using Deep Neural Networks surrogate models is that the topology of the grid (a graph) should be taken into account, and because all topologies cannot be included in the training set, this requires out-of-sample generalization capabilities of the learned models.

Balthazar Donon developped in his PhD 47 an approach based on Graph Neural Networks (GNNs). From a Power Grid perspective, GNNs can be viewed as including the topology in the heart of the structure of the neural network, and learning some generic transfer function amongst nodes that will perform well on any topology. His work uses a loss that directly aims to minimize Kirshhoff's law on all lines. Theoretical results as well as a generalization of the approach to other optimization problems had been originaly published at NeurIPS 2021 92.

Eva Boguslawski’s CIFRE PhD, that started in Sept. 2022, addresses the problem of global monitoring of the grid through decentralized decision process (aka multi-agent Reinforcement Learning), in the line of the LR2PN challenge (see Section 8.6) that she contributed to organize during a previous internship 32.

Participants: Isabelle Guyon, Marc Schoenauer, Michèle Sebag

PhDs: Herilalaina Rakotoarison

Collaboration: Vincent Renault (Artelys).

One of the goals of the ADEME Next project, in collaboration with SME Artelys (see also Section 4.2), is the sizing and capacity design of regional power grids. Though smaller than the national grid, regional and urban grids nevertheless raise scaling issues, in particular because many more fine-grained information must be taken into account for their design and predictive growth.

Regarding the design of such grids, and provided accurate predictions of consumption are available (see below), off-the-shelf graph optimization algorithms can be used. However, they require a careful tuning of their hyperparameters, and this was the motiaton of funding Herilalaina Rakotoarison's PhD, that tackles the automatic tuning of such hyer-parameters (see Section 8.2.1); both the Mosaic algorithm 148 and the Metabu algorithm to learn meta-features are being used for Artelys' home optimizer Knitro, and compared to the state-of-the-art in parameter tuning (confidential deliverable).

Participants: Guillaume Charpiat, Marc Schoenauer, Michèle Sebag

PhDs: Matthieu Nastorg

Post-doc: Tamon Nakano

Collaboration: Alessandro Bucci (Safran Tech, former member of the team); Thilbault Faney et Jean-Marc
Gratien (IFPEN).

During the 2.5 years that he spent at TAU, funded by the bilateral project with IFPEN, Alessandro Bucci worked on several use case of IFPEN, with the goal of accelerating some softwares that IFPEN uses daily. This IFPEN/TAU collaboration lead to a successful application to a DATAIA program with the ML4CFD project. Direct follow-up of the previous collaboration , a prominent result was obtained on the simulation of diphasic fluid flow in a distillation column: one of the main timeconsuming step in the simulation is the tracking of the interface between the bubbles of the gaz and the liquid they circulate in within the Volume-of-Fluid numerical method: this critical step was replaced with a Graph Neural Network model directly working on the unstrutured mesh, making the industrial application possible 63. ML4CFD is also funding Matthieu Nastorg’s PhD, who significantly accelerated 38 the numerical resolution of the Poisson equation (ubiquitous in CFD, e.g., to compute the pressure in Navier Stokes simulations), based on B. Donon’s Statistical Solvers 92. Note that this resul is more general than its application to energy problems, but was made possible only because of the collaboration with IFPEN.

Participants: Cyril Furtlehner, Michèle Sebag

Post-doc: Olivier Bui

Collaboration: Jannis Teunissen (CWI)

Space Weather is broadly defined as the study of the relationships between the variable conditions on the Sun and the space environment surrounding Earth. Aside from its scientific interest from the point of view of fundamental space physics phenomena, Space Weather plays an increasingly important role on our technology-dependent society. In particular, it focuses on events that can affect the performance and reliability of space-borne and ground-based technological systems, such as satellite and electric networks that can be damaged by an enhanced flux of energetic particles interacting with electronic circuits.6

Since 2016, in the context of the Inria-CWI partnership, a collaboration between Tau and the Multiscale Dynamics Group of CWI aims to long-term Space Weather forecasting. The goal is to take advantage of the data produced everyday by satellites surveying the sun and the magnetosphere, and more particularly to relate solar images and the quantities (e.g., electron flux, proton flux, solar wind speed) measured on the L1 libration point between the Earth and the Sun (about 1,500,000 km and 1 hour time forward of Earth).
A challenge is to formulate such goals in terms of supervised learning problem, while the "labels" associated to solar images are recorded at L1 (thus with a varying and unknown time lag). In essence, while typical ML models aim to answer the question What, our goal here is to answer both questions What and When. This project has been articulated around Mandar Chandorkar's Phd thesis 83 which has been defended this year in Eindhoven. The continuation of this collaboration is inseured by the hiring of Olivier Bui as a post-doc who's work has consisting in extending preliminary results on solar wind forecasting based on auto-encoded solar magnetograms on a longer period of data corresponding to 2 solar cycles. Negative results have incited us to dig more into physical models of solar wind propagation and try to combine them with ML models in a systematic way.

Participants: Guillaume Charpiat, Flora Jay, Aurélien Decelle, Cyril Furtlehner

PhD: Théophile Sanchez, Jérémy Guez

PostDoc: Jean Cury, Burak Yelmen

Collaboration: Bioinfo Team (LISN), Estonian Biocentre (Institute of Genomics, Tartu, Estonia), UNAM (Mexico), U Brown (USA), U Cornell (USA), TIMC-IMAG (Grenoble), MNHN (Paris), Pasteur Institute (Paris)

Thanks to the constant improvement of DNA sequencing technology, large quantities of genetic data should greatly enhance our knowledge about evolution and in particular the past history of a population. This history can be reconstructed over the past thousands of years, by inference from present-day individuals: by comparing their DNA, identifying shared genetic mutations or motifs, their frequency, and their correlations at different genomic scales. Still, the best way to extract information from large genomic data remains an open problem; currently, it mostly relies on drastic dimensionality reduction, considering a few well-studied population genetics features.

For the past decades, simulation-based likelihood-free inference methods have enabled researchers to address numerous population genetics problems. As the richness and amount of simulated and real genetic data keep increasing, the field has a strong opportunity to tackle tasks that current methods hardly solve. However, high data dimensionality forces most methods to summarize large genomic datasets into a relatively small number of handcrafted features (summary statistics).In Theophile Sanchez' PhD 51, we propose an alternative to summary statistics, based on the automatic extraction of relevant information using deep learning techniques. Specifically, we design artificial neural networks (ANNs) that take as input single nucleotide polymorphic sites (SNPs) found in individuals sampled from a single population and infer the past effective population size history. First, we provide guidelines to construct artificial neural networks that comply with the intrinsic properties of SNP data such as invariance to permutation of haplotypes, long scale interactions between SNPs and variable genomic length. Thanks to a Bayesian hyperparameter optimization procedure, we evaluate the performance of multiple networks and compare them to well established methods like Approximate Bayesian Computation (ABC). Even without the expert knowledge of summary statistics, our approach compares fairly well to an ABC based on handcrafted features. Furthermore we show that combining deep learning and ABC can improve performance while taking advantage of both frameworks. Later, we experimented with other types of permutation invariance, based on similar architectures, and achieved a significative performance gain with respect to the state of the art, including w.r.t. ABC on summary statistics (20% gap), which means that we extract information from raw data that is not present in summary statistics. The question is now how to express this information in a human-friendly way.

In the short-term these architectures can be used for demographic inference 60 or selection inference in bacterial populations (ongoing work with a postdoctoral researcher, J Cury, collab: Pasteur Institute, for ancient DNA: UNAM and U Brown); the longer-term goal is to integrate them in various systems handling genetic data or other biological sequence data. Regarding the bacterial populations, we already implemented a flexible simulator that will allow researchers to investigate complex evolutionary scenarios (e.g. dynamics of antibiotic resistance in 2D space through time) with realistic biological processes (bacterial recombination), which was impossible before (collab. U Cornell, MNHN) 14.

In collaboration with the Institute of Genomics of Tartu, we leveraged two types of generative neural networks (Generative Adversarial Networks and Restricted Boltzmann Machines) to learn the high dimensional distributions of real genomic datasets and create artificial genomes 163. These artificial genomes retain important characteristics of the real genomes (genetic allele frequencies and linkage, hidden population structure, ...) without copying them and have the potential to be valuable assets in future genetic studies by providing anonymous substitutes for private databases (such as the ones hold by companies or public institutes like the Institute of Genomics of Tartu. Ongoing work concerns scaling up to the full genome and developing new privacy scores.

We released dnadna, a flexible open-source python-based software for deep learning inference in population genetics7. It is task-agnostic and aims at facilitating the development, reproducibility, dissemination, and reusability of neural networks designed for genetic polymorphism data. dnadna defines multiple user-friendly workflows 152.

Participants: Isabelle Guyon

PhD: Adrien Pavao

Collaboration: Kristin Bennett and Joe Pedersen (RPI, NY, USA), Wei-Wei Tu (4Paradigm, Chine), Pablo Piantanida (Centrale-Supelec)

While theoretical criteria of privacy preservation, such as “differencial privacy” are important to gain insights into how to protect privacy, they are often impractical, because they put forward pessimistic bounds and impose degrading data and/or model to a point that hampers utility. Additionally, for all practical purposes, data owners seek to obtain guarantees that no privite information is leaked in the form of an empirical statistical test, rather than a more elusive theoretical guarantee. To that end, we have set to work on evaluating the effectiveness of privacy protection agains specific attacks, such as membership inference or attribute inference. We devised an evaluation apparatus called “LTU-attacker” 30, in collaboration with Kristin Bennett, Joe Pedersen, and Wei-Wei Tu and with 2 interns (Rafel Monos-Gomez and Jiangna Huang) have obtained interesting preliminary results demonstrating lack of privacy preservation of most scikit-learn algorithms under membership inference attacks. New directions currently explored in collabotation with Pablo Piantanida include defining a degree of “privacy exposure” of particular individual involving information theoretic arguments.

With Master student Alice Lacan, we have been investigating the modelization of the Covid-19 epidemic propogation using compartimental models, following earlier work by former master student Martin Cepeda. A group of students including Alice entered the "Pandemic response" XPrize and qualified for the final phase. This work was follwed by a paper on estimating uncertainty in time series, in application to prediciting the evolution of the number of Covid cases presented at the BayLearn 2022 conference 121. Alice was invited to give a presentation of this work at the WIDS 2023 conference.

Last but not least regarding Covid-19, F. Landes participated to Inria Saclay collaborative effort to monitor and optimize the emergency bed occupancy in East of France 24.

Participants: Guillaume Charpiat

PhD: Loris Felardos

Collaboration: Jérôme Hénin (IBPC), Bruno Raffin (InriAlpes)

Numerical simulations on massively parallel architectures, routinely used to study the dynamics of biomolecules at the atomic scale, produce large amounts of data representing the time trajectories of molecular configurations, with the goal of exploring and sampling all possible configuration basins of given molecules. The configuration space is high-dimensional (10,000+), hindering the use of standard data analytics approaches. The use of advanced data analytics to identify intrinsic configuration patterns could be transformative for the field.

The high-dimensional data produced by molecular simulations live on low-dimensional manifolds; the extraction of these manifolds will enable to drive detailed large-scale simulations further in the configuration space. We study how to bypass simulations by directly predicting, given a molecule formula, its possible configurations. This is done using Graph Neural Networks 58 in a generative way, producing 3D configurations, and constitutes the main part of Loris Felardos’ PhD 48, funded by the Inria Challenge HPC/Big Data. The goal is to sample all possible configurations, and with the right probability. This year we studied various normalizing flow architectures as well as varied training criteria suitable for distributions (Kullback-Leibler divergence in latent or sample space, in one direction or the other one, as it is not symmetric, but also pairwise distances, optimal transport, etc.). It turns out that mode collapse is frequently observed in most cases, even on simple tasks. Further analysis identified several causes for this, from which we built remedies.

Participants: François Landes, Marc Schoenauer

PhD: Vincenzo Schimmenti

Collaboration: Alberto Rosso (LPTMS)

Earthquakes occur in brittle regions of the Crust typically located at the depth of 5-15 km and characterized by a solid friction, which is at the origin of the stick-slip behaviour. Their magnitude distribution displays thecelebrated Gutenberg-Richter law and a significant increase of the seismic rate is observed after large events (called main shocks). The occurrence of the subsequent earthquakes in the same region, the aftershocks, obeys well established empirical laws that demand to be understood. A change in the seismic rate also happens before a main shock, with an excess of small events compared to the expected rate of aftershocks related to the previous main shock in that region. These additional events are defined as foreshocks of the coming main shock, however they are scarce so that defininig them is a very difficult task. For this reason their statistical fingerprint, so important for human secutiry, remains elusive.

The treatment of rare events by Machine Learning is a challenging yet rapidly evolving domain. At TAU we have a great expertise in data modeling, in particular Bayesian models and Restricted Boltzman Machines (RBMs) have been built to model space weather forecast data (Section 8.5.1). These techniques, inspired from statistical physics, are both based on a probabilistic description of latent variables, allowing the modelling of a large span of data correlations. This kind of models can be extended to study spatially resolved earthquakes, the latent variable here being the local stress within the fault and in the ductile regions. Our goal is to characterize the statistical properties of a sequence of events (foreshocks, main shock and aftershocks) and predict its following history. We will first study the sequences obtained from simulations of the physical model 145. We will answer the following question: given a short sequence of foreshocks, can we predict the future of the sequence? How big will be the main shock? When will it occur? In a second step we will use also the data coming from real sequences, where events are unlabeled. These sequences are public and available (The most accurate catalog is for Southern California, a catalog with 1.81 million earthquakes. It is available at https://scedc.caltech.edu/research-tools/QTMcatalog.html). Concretely, the data consists in the earthquakes’ magnitude, occurrence time and hypocenter locations.

Two parallel directions are being explored, with our PhD Student, Vincenzo Schimmenti:

Partecipants: Michele Alessandro Bucci, Marc Schoenauer

PhD: Emmanuel Menier

Collaboration: Mouadh Yagoubi (IRT-SystemX), Lionel Mathelin (DATAFLOT team, LISN)

Numerical simulations of fluid dynamics in industrial applications require the spatial discretization of complex 3D geometries with consequent demanding computational operations for the PDE integration. The computational cost is mitigated by the formulation of Reduced Order Models (ROMs) aiming at describing the flow dynamics in a low dimensional feature space. The Galerkin projection of the driving equations onto a meaningful orthonormal basis speeds up the numerical simulations but introduces numerical errors linked to the underrepresentation of dissipative mechanisms.

Deep Neural Networks can be trained to compensate missing information in the projection basis. By exploiting the projection operation, the ROM correction consists in a forcing term in the reduced dynamical system which has to i) recover the information living in the subspace orthonormal to the projection one ii) ensure that its dynamic is dissipative. A constrained optimization is then employed to minimize the ROM errors but also to ensure the reconstruction and the dissipative nature of the forcing. We tested this solution on benchmarked cases where it is well known that transient dynamics are poorly represented by ROMs. The results 62 show how the correction term improves the prediction while preserving the guarantees of the ROM. Furthermore, the approach was generalized, and the extension was validated on Michelin use case of rubber calendering process 61.

Participants: Michele Alessandro Bucci (now with SafranTech)

Collaboration: Lionel Mathelin (LISN), Onofrio Semeraro (LISN), Sergio Chibbaro (UPMC), Alexandre
Allauzen (ESPCI)

The inference of a data driven model aiming at reproducing chaotic systems is challenging even for the most performing Neural Network architectures. According to the ergodic theory, the amount of data required to converge to the invariant measure of a chaotic system goes exponentially with its intrinsic dimension. Recent work 54 analyzes the amount of data that is sufficient for a priori guaranteeing a faithful model of the physical system.

Participants: Guillaume Charpiat, Marc Schoenauer, Michèle Sebag

PhDs: Balthazar Donon, Loris Felardos, Wenzhuo Liu, Matthieu Nastorg

Post-doc: Tamon Nakano

Collaboration: Mouahd, Yagoubi (IRT SystemX), Lionel Mathelin (LISN), Alessandro Bucci (Safran Tech, former member of the team); Thilbault Faney et Jean-Marc Gratien (IFPEN).

Many of the works that have been introduced earlier featured the use of Graph Neural Networks to learn how to solve numerical problems invoving data on graphs: Balthazar Donon’s PhD 47 simulated the French Power Grid, Loris Felardos’ PhD 48 learnt the distribution of 3D molecule conformations, Matthieu Nastorg accelerate the numerical resolution of Poisson’s equation on any unstructured mesh with GNNs 38, Tamon Nakano also handled unstructured meshes with GNNs to track the interface between both fluids in multi-phasic flow simulations 63.

But the use of GNNs to approximate the numerical solutions of PDEs on any unstructured mesh, rather than using grid meshes to be able to use the CNNs and the whold zoology of Deep Neural Networks designed for image processing was systematicall studies in Wenzhuo Liu’s PhD: After porting ideas from multi-grid approaches to Finite Elements, and comparing the CNN and GNN approchaes 130, she tackled the poor Out-of-Distribution generalization issue using Meta-Learning 37, improving the OoD learning on CFD simulations of air flow around an airfoil by considering each airfoil shape as a separate task. he is now completing her PhD (defense in March) by applying Transfer Learning to decrease the amount of data to learn accurate simulation on fine meshes using numerous costless simulations on coarse meshes (submitted).

Participants: Cécile Germain, Isabelle Guyon, Adrien Pavao, Anne-Catherine Letournel, Marc Schoenauer, Michèle Sebag

PhD: Eva Boguslawski, Balthazar Donon, Adrien Pavao, Haozhe Sun, Romain Egele

Engineer: Sébastien Tréguer.

Collaborations: D. Rousseau (LAL), André Elisseeff (Google Zurich), Jean-Roch Vilmant (CERN), Antoine Marot and Benjamin Donnot (RTE), Kristin Bennett (RPI), Magali Richard (Université de Grenoble), Wei-Wei Tu (4Paradigm, Chine), Sergio Escalera (U. Barcelona, Espagne).

The Tau group uses challenges (scientific competitions) as a means of stimulating research in machine learning and engage a diverse community of engineers, researchers, and students to learn and contribute advancing the state-of-the-art. The Tau group is community lead of the open-source Codalab platform (see Section 7), hosted by Université Paris-Saclay. The project had grown since 2019 and includes now an engineer dedicated full time to administering the platform and developing challenges (Adrien Pavao), financed in 2021 by a 500k€ project with the Région Ile-de-France. This project will also receive the support of the Chaire Nationale d'Intelligence Artificielle of Isabelle Guyon (2020-2024).

Adrien Pavao has also set to work on the theoretical rationalization of judging competitions. A first work built ties between this problem and the theory of social choice 142. This is applicable, in particular to judging multi-task or multi-objective challenges: each task or objective can be thought of as a “judge” voting towards determining a winner. He devised novel empirical criteria to assess the quality of ranking functions, including the generalization to new tasks and the stability under judge or candidate perturbation and conducted empirical comparisons on 5 competitions and benchmarks. While prior theoretical analyses indicate that no single ranking function satisfies all desired theoretical properties, our empirical study reveals that the classical "average rank" method (often used in practice to judge competitions) fares well. However, some pairwise comparison methods can get better empirical results.

Following the highly successful ChaLearn AutoML Challenges (NIPS 2015 – ICML 2016 109 – PKDD 2018 110)and AutoDL129 was run in 2019 (see http://autodl.chalearn.org), that pointed to the importance of meta-learning, we opened a new line of research on meta-learning from learning curves [28, 64] and cross-domain meta-learning [36]. This led us to explore uses of reinforcement learning as a means to devise policies for meta-learning (on-going). In parallel, a new challenge on automated reinforcement learning (AutoRL) is currently under design.

A new challenge series in Reinforcement Learning for Power Grid control was started in 2021 with the company RTE France on the theme “Learning to run a power network” 134
(L2RPN, http://l2rpn.chalearn.org). The goal is to test the potential of Reinforcement Learning to solve a real world problem of great practical importance: controlling electricity transportation in smart grids while keeping people and equipment safe. The first edition was run in Spring 2019, and aimed at demonstrating the feasibility of applying Reinforcement Learning for controlling electrical flows on a power grid. The 2020 edition 133 introduced a realistically-sized grid environment along with two fundamental real-life properties of power grid systems to reconsider while shifting towards a sustainable world: robustness and adaptability, and the 2022 edition was concerned with changing topology, and was co-organized with RTE and TAILOR challenge task force (TAU team is responsible of the organization of challenges within the European project TAILOR). The analysis paper is under review.

In preparation, with the sponsorship of Paris-Région Ile de France, a competition between statups in the AI challenge for Industry series is being organized, in collaboration with RTE 32. The competition is assorted with a 1 millon Euro prize pool. The objective is to device control policies for the French electricity grid under scenarios of energies of the future, towards attaining carbon neutrality. The participants will be tackling prospective productions and consumption scenarios of the future, emphasing renewable energies. This poses particular difficulties because of solal and wind energies have irregular productions.

Paris Ile-de-France region selected in 2021 Codalab and the Tau team to organize the industry machine learning challenge seris of the Paris Region. Adrien Pavao, who was the project leader, organized with Dassault aviation a project of “numerical twins”, aiming at performing predictive maintenance on airplanes. The Paris Region offered 500K Euros to the winner, a startup, which would then collaborate with Dassault to productize the solution. The challenge took place from February 2021 to May 2021. The results have indicated that, on such problems of time series regression, ensembles of decision trees such as XGBoost dominate over DL methods. This result, which came somewhat as a surprise, but stem from the massive amount of data that had to be processed. Despite the significant compute power made avaliable (10 GPUs for 2 days), search for optimal architectures was difficult. Results of detailed analyses conducted by a consortium of organizers and participants have been published 140. This challenge has demonstrated that Codalab is now “industry grade”, and has paved the way to organizing other AI for Industry challenges. We have currently in preparation a challenge targeting carbon-neutrality by 2025, in collaboration with RTE-France.

It is important to introduce challenges in ML teaching. This has been done (and is on-going) in I. Guyon's Master courses 141 : some assignments to Master students are to design small challenges, which are then given to other students in labs, and both types of students seem to love it. Codalab has also been used to implement reinforcement learning homework in the form of challenges by Victor Berger and Heri Rakotoarison for the class of Michèle Sebag. New directions being explored by students in 2021 include takling fairness and bias in data.

In terms of dissemination, a collaborative book “AI competitions and benchmarks: The science behind the contests ” written by expert challenge organizers is under way and will appear in the Springer series on challenges in machine learning, see http://www.chalearn.org/books.html. Challenge organizaton is now better grounded in theory, with such effort. The thesis of Adrien Pavao will include several advances in devising sound challenge protocols, including two-stage challenges, as described in his recent paper "Filtering participants improves generalization in competitions and benchmarks" 29.

Tau continues its policy about technology transfer, accepting any informal meeting following industrial requests for discussion (and we are happy to be often solicited), and deciding about the follow-up based upon the originality, feasibility and possible impacts of the foreseen research directions, provided they fit our general canvas. This lead to the following 3 on-going CIFRE PhDs, with the corresponding side-contracts with the industrial supervisor, one bilateral contract with IFPEN, one recently started bilateral contract with Fujitsu (within the national "accord-cadre" Inria/Fujitsu), plus at least two new CIFRE PhDs, one with our long-lasting partner RTE, and one with Ekimetrics company, with whom we have never worked before), that will start in 2022.

IFPEN (Institut Français du Pétrole Energies Nouvelles) 2019-2023 (300 kEuros), to hire an Inria Starting Research Position (Alessandro Bucci) to work in all topics mentioned in Section 3.2 relevant to IFPEN activity.

Coordinator: Marc Schoenauer

Participants: Alessandro Bucci, Guillaume Charpiat

Fujitsu, 2021-2022 renewed 2022-2023 (200k€ per year), Causal discovery in high dimensions

Coordinator: Marc Schoenauer

Participants: Shuyu Dong and Michèle Sebag

CIFRE RTE 2021-2024 (72 kEuros), with RTE, related to Eva Boguslawski's CIFRE PhD Decentralized Partially Observable Markov Decision Process for Power Grid Management

Coordinator: Marc Schoenauer and Matthieu Dussartre (RTE)

Participants: Eva Boguslawski, Alessandro Leite

CIFRE Ekimetrics 2022-2025 (45 kEuros), with Ekimetrics, related to Audrey Poinsot's CIFRE PhD Causal incertainty quantification under partial knowledge and low data regimes

Coordinator: Marc Schoenauer

Participants: Guillaume Charpiat, Alessandro Leite, Audrey Poinsot and Michèle Sebag

CIFRE MAIR 2022-2025 (75 kEuros), with Meta (Facebook) AI Research, related to Mathurin Videau's CIFRE PhD Reinforcement Learning: Sparse Noisy Reward

Coordinator: Marc Schoenauer and Olivier Teytaud (Meta)

Participants: Alessandro Leite and Mathurin Videau

CIFRE MAIR 2022-2025 (75 kEuros), with Meta (Facebook) AI Research, related to Badr Youbi's CIFRE PhD Learning invariant representations from temporal data

Coordinator: Isabelle Guyon (now Michèle Sebag) and David Lopez-Paz (Meta)

Participants: Badr Youbi

Adra-e project on cordis.europa.eu

TAILOR project on cordis.europa.eu

VISION project on cordis.europa.eu

Chaire IA HUMANIA 2020-2024 (600kEuros), Democratizing Artificial Intelligence.

Coordinator: Isabelle Guyon (TAU)

Participants: Marc Schoenauer, Michèle Sebag, Anne-Catherine Letournel, François Landes.

HUSH 2020-2023 (348k euros), HUman Supply cHain behind smart technologies.

Coordinator : Antonio A. Casilli (Telecom Paris)

Participants: Paola Tubaro

SPEED 2021-2024 (49k€) Simulating Physical PDEs Efficiently with Deep Learning

Coordinator: Lionel Mathelin (LISN (ex-LIMSI))

Participants: Michele Alessandro Bucci, Guillaume Charpiat, Marc Schoenauer.

RoDAPoG 2021-2025 (302k€) Robust Deep learning for Artificial genomics and Population Genetics

Coordinator:Flora Jay,

Participants: Cyril Furtlehner, Guillaume Charpiat.

ADEME NEXT 2017-2021, extended 2023 (675 kEuros). Simulation, calibration, and optimization of regional or urban power grids (Section 4.2).

ADEME (Agence de l'Environnement et de la Maîtrise de l'Energie)

Coordinator: SME ARTELYS

Participants Isabelle Guyon, Marc Schoenauer, Michèle Sebag, Victor Berger (PhD), Herilalaina Rakotoarison (PhD), Berna Bakir Batu (Post-doc)

IPL $\phantom{\rule{-0.166667em}{0ex}}$ HPC-BigData 2018-2022 (100 kEuros) High Performance Computing and Big Data (Section 8.5.4)

Coordinator: Bruno Raffin (Inria Grenoble)

Participants: Guillaume Charpiat, Loris Felardos (PhD)

Inria Challenge (formerly IPL) HYAIAI, 2019-2023, HYbrid Approaches for Interpretable Artificial Intelligence

Coordinator: Elisa Fromont (Lacodam, Inria Rennes)

Participants: Marc Schoenauer and Michèle Sebag

Les vraies voix de l'Intelligence Artificielle, 2021-2023 (29k euros), funded by Maison des Sciences de l'Homme Paris-Saclay.

Coordinator : Paola Tubaro

Participants: A.A. Casilli (Telecom Paris); I. Vasilescu, L. Lamel, Gilles Adda (CNRS-LISN); J.L. Molina (UAB Barcelona); J.A. Ortega (Univ. València)

Inria Challenge OceanAI 2021-2025, AI, Data, Models for a Blue Economy

Coordinator: Nayat Sanchez Pi (Inria Chile)

Participants: Marc Schoenauer, Michèle Sebag and Shiyang Yan

DATAIA ML4CFD 2020-2022 (105 kEuros) Machine Learning for Computational Fluid Dynamics.

Coordinator: Michele Alessandro Bucci

Participants: Guillaume Charpiat, Marc Schoenauer

Collaboration: IFPEN (Jean-Marc Gratien and Thibault Faney)

DATAIA YARN 2022-2025 (240 kEuros) Automatic Processing of Messy Brain Data with Robust Methods and Transfer Learning.

Coordinator: Sylvain Chevallier, Florent Bouchard (L2S)

Collaboration: Raymond Poincaré Hospital (France), FCAI (Aalto University, Finland), Frédéric Pascal (L2S), Alexandre Gramfort (Meta)

All TAU members are reviewers of the main conferences in their respective fields of expertise.

All members of the team reviewed numerous articles for the most prestigious journals in their respective fields of expertise.