3.1 Context

Telecommunication networks are ubiquitous in our daily life. The network ecosystem and its infrastructure are increasingly complex with more and more involved actors and technologies which have to interoperate. This complexity is illustrated in Figure 1, where

Show image description

Description of the various actors on the ICT (for Information & Communications Technology) ecosystem

the basic network paradigm with network operators simply sending content from/to end users has moved to a topology with many players having different roles. A complete telecommunications system has to integrate traditional networks capabilities but also all other innovative Internet services into a single service platform. Selfishness of actors, heterogeneity of requests and technologies, interoperability of services and automation of decisions are major challenges to be tackled.

ERMINE deals with network operations management and the required associated monitoring and measurements and performance analysis. Operation network management means for us performance-based management, fault analysis, or the design of procedures and strategies for resource provisioning and quality-of-service fulfillment. On this management side, we aim at looking at the whole topic of networks economics, studying the best economic strategies of actors of the digital economy, their interactions and the potential need for regulation. We also want to address resource management through artificial intelligence for automated decision as another main research direction. On the measurement side, monitoring services of all sorts is a major challenge for regulatory bodies to ensure a fair and legal behavior of providers, and is also of primary importance for providers themselves to optimize decisions; this often requires new modeling, analysis and artificial intelligence tools. This is detailed below.

3.2 Managing network operations

The heterogeneity and complexity of the ICT (for Information & Communications Technology) world bring a number of challenges to network operations management. ERMINE addresses key issues for network management: i) the challenges of ICT economics; ii) network control and interoperability iii) automated decisions through AI.

ICT is omnipresent in our modern society, and the economy has gone beyond the industrial economy to the Internet and ICT economy. Thanks to hyper-connectivity, there are now lots of opportunities for innovation. As of May 2020, among the top-10 most valuable companies worldwide, seven are ICT companies, trusting the places from 2 to 8 (Microsoft, Apple Inc., Amazon Inc., Alphabet Inc., Facebook, Alibaba Group, Tencent). Internet Service Providers (ISPs), Content Delivery Networks (CDNs) and cloud providers, social network actors, all services and content providers are among actors needing a business model as profitable as possible. Designing and analyzing such business models and their acceptance by end users are issues to be addressed, leading to challenges in terms of outcome for all players (who act in most cases selfishly) and in terms of benefits for society. On that last point, regulators have to determine rules that actors need to follow in order to avoid harmful behaviors and maximize social welfare. The issues we address include: the design of economic rules for new services, the analysis of the impact of players decisions and interactions, and the potential design of rules or incentives from regulatory bodies leading to the most adequate (social) outcome. One of the frameworks to be used is that of game theory, and in particular of mechanism design.

There has been a metamorphosis in the last few years on the management of network operations, driven by the virtualization of networks and services. This evolution allows to meet the needs in terms of the dynamic scaling of infrastructures and the agility of the decision-making, namely, a necessary prerequisite for reducing operating costs as well as improving return on investment. These developments are radically changing the way services are managed, as they become more complex (i.e., services in the form of graphs, micro-services, etc.), their expectations can be very diverse and strategies for their placement could consider a single domain or spanning across several domains, whether cooperative or not. Thus, the operation of the network becomes extremely difficult and requires not only optimization of resources but also economic considerations, especially when management involves several domains. We believe that automated management and control is the key direction for an efficient solution. We deal with the automation of the network by contributing to ongoing standardization efforts, notably by the “Zero touch network & Service Management” (ZSM) group of the ETSI (the European Telecommunications Standards Institute), and through the elaboration of solutions based on the most recent advances in machine learning techniques, and in particular in deep reinforcement learning. Some of the challenges we tackle include dynamic placement of complex and constrained – QoS/QoE – services (i.e., network slicing), automatic service scaling, congestion control in the context of new generation networks, including IoT networks (e.g., massive Machine-Type Communications (mMTC)). Different aspects are taken into account when developing the various solutions, including reliability, resilience, guaranteed performance (i.e., deterministic networks), but also the energy efficiency on which the viability of the latter depends. Ensuring the performance of the solutions when using machine learning techniques is, however, a major problem that is poorly addressed in the literature. We address it by proposing techniques that offer at least similar performances to heuristics. Indeed, even if heuristics can be very efficient, they often have, however, some limitations. One of these is the blocking at the level of local minima. AI techniques can address this issue encountered in combinatorial problems (of the NP-hard type, such as the travelling salesman one). They sometimes yield near-optimal results. AI techniques have the additional advantage of being able to learn by interacting with the environment, and can therefore find solutions that heuristics could not.

3.3 Measuring

A proper management of network services and operations cannot be effective without measurement/monitoring and without the analysis of relevant metrics. Indeed, decision making cannot be realized without knowledge and data on the past, present and even future status of the activity. This is typically the case for agile capacity planning and resource sharing for which usage needs to be computed and even predicted. This is a challenge for which tools coming from AI are very promising. Another relevant example of challenge is measurements for regulation purposes, in relation with our activity on network economics: defining regulation rules means designing measurement procedures to verify that the rules are followed. It is for example required to monitor a neutral behavior of ISPs as expected from the Net neutrality principles. While there exist a few tools towards that goal, they are actually all devoted to very specific hindrances to neutrality (blocking, degradation) for specific types of flows or traffic; a major challenge is to detect any type of non-authorized behavior. Many other actors of the ICT economy, if not all, are barely monitored while an unfair behavior could be seen as harming end users and society. To name some noticeable examples, we can cite CDNs which cache some content at the edge of the network and could unfairly propose a better quality to selected customers, or search engines who can prioritize the web. In all those cases, different and specific techniques have to be designed to analyze and detect unexpected behaviors.

Similarly, managing resources and operations calls for evaluating (measuring) the impact of decisions to determine the most appropriate ones. Modeling and performance evaluation techniques are appropriate and useful solutions at a low cost, without the need to build and run the real system. While we have in the group a long experience on performance evaluation, new challenges keep popping up due to new types of services translated into new problems to be modeled and analyzed through performance evaluation tools, and adaptations or extensions of existing techniques for better decision making. They usually require to develop new appropriate network metrics to assess network service operations effectiveness. Illustrative examples are blockchains. The popularity of blockchain lies in the introduction of the concept of a public distributed ledger shared by all participants of the system without relying on a centralized authority. This distributed ledger records all the transactions between parties efficiently and in a verifiable and permanent way. In order to enjoy higher throughput and self-adaptivity to transactions demand in ICT, it requires the development of a new architecture and thus specific modeling and analysis techniques.

3.4 Research axes

Our activity is organized in five subtopics (or axes). The first two axes are on network management, and the next two are orthogonal on the associated and required measurements. Axis 5 on AI is transverse. The axes and their intersections are described in Figure 2.

Show image description

The five research axes: two on network management, two orthogonal on measurements and transverse on AI, with intersections

As a brief description:

• Network Economics. The digital economy has gained and keeps gaining in scale, scope, and significance. The ecosystem is quickly evolving and one of our main goals is to answer all questions related to the Internet & digital economy that pop up in line with what we started to do, and to be reactive to the news in that domain. We want to address issues concerning Internet resource allocation and pricing models, the economics of services, issues with vertical integration, the economics of structuring platforms, as well as economics and regulation (including network neutrality).
• Managing next generation networks & massive IoT communication. One of the fundamental challenges in the traffic management of new and emerging networking activities is to describe, analyze and control heterogeneous resources and complex services, under energy constraints. While AI is a main research direction to address this problem, there exist many other possibilities. The main questions of interest include the automation of infrastructure management (a typical example being network slicing) or the massive access in IoT networks.
• Measuring, monitoring & regulation. With the rapid evolution of networks, many needs are appearing for the design of measurement techniques associated with new services, but also for regulators to monitor network's activity. Measurements are of two possible categories: passive, monitoring existing traffic, or active, i.e., based on injecting traffic to investigate the network behavior. The goal of this research direction is to develop practical network operations measurement techniques, from the two different points of view: operators (for a better management) and regulators (for monitoring).
• Measuring performance metrics based on models. Measuring performances has to be done not only to observe and monitor directly a service like in previous subsection, but also at different phases such as at conception or to propose enhancements. Indeed, network services management and decision making cannot be applied efficiently without a valid and accurate evaluation of performance metrics through the construction and analysis of models. We are carrying out such evaluations.
• Artificial Intelligence (AI) in networking. AI is a transverse research axis, since it is used for both management and measuring. We are not only users of the technology but also interested in the methodology itself, that is, we also develop new techniques, based on our networking problems but usable outside. This concerns mainly supervised learning, in particular time series forecasting using this type of learning, and reinforcement learning. From the point of view of the tools, our skills and work are essentially related to Random Neural Networks, Reservoir Computing and deep neural architectures in Reinforcement Learning.

5.1 Impact of research results

Our research on network economics and particularly neutrality issues aims to view how, and if, regulation can help to improve social welfare, among other goals. Typical examples are how tools SNIDE and DemoWehe described later explaining how potential deviations from rules by respectively search engines and Internet service providers can be detected. Our modeling works on neutrality also allow to investigate the appropriate rules toward a benefice to society. This has led to collaborations with specialists of human sciences in information networks (R. Badouard from Pantheon-Assas and I. Lyubareva from IMT Atlantique) to tackle the socially sensitive issue of bias in information networks, in particular from search engines.

We also work on so-called green streaming to analyze the interactions and strategies of various vendors in the market and provide valuable insights that can contribute to the sustainable growth of the Content Delivery Network (CDN) industry. By understanding the implications of energy pricing and its impact on the market dynamics, this research endeavors to offer potential solutions and recommendations to address the challenges and foster a more sustainable and efficient CDN ecosystem. Similarly, we have started to work on the societal, environmental, and economic impacts of digital infrastructures through a Cifre PhD with Orange Labs.

6.1 Awards

• Paper 25 won the best paper award at the 11th International Conference on Network Games, Control and Optimization (NETGCOOP'2024).
• Paper 16 won the best paper award at the 50th Latinoamerican Conference in Computer Science (CLEI 2024).
• Paper 6 won the Editor’s Choice Awards 2024 in Elsevier, Internet of Things Journal.

7.1 New software

8.1 Network Economics

Participants: Patrick Maillé, Bruno Tuffin, Maël Le Treust, Yassine Hadjadj-Aoul.

Economics and regulation: network neutrality and network slicing. A non-negligible part of our activity on network economics was focusing in 2024 on network neutrality and its consequences.

Network slicing is a key component of 5G-and-beyond networks but induces many questions related to an associated business model and its need to be regulated due to its difficult co-existence with the network neutrality debate. We propose in 5 a slicing model in the case of heterogeneous users/applications where a service provider may purchase a slice in a wireless network and offer a “premium” service where the improved quality stems from higher prices leading to less demand and less congestion than the basic service offered by the network owner, a scheme known as Paris Metro Pricing. We obtain thanks to game theory the economically-optimal slice size and prices charged by all actors. We also compare with the case of a unique “pipe” (no premium service) corresponding to a fully-neutral scenario and with the case of vertical integration to evaluate the impact of slicing on all actors and identify the “best” economic scenario and the eventual need for regulation.

Slicing is therefore seen as a key characteristic of 5G-and-beyond networks, however it seems in contradiction with neutrality principles promoted worldwide. We detail in 20 the two contradictory but considered compulsory notions, and discuss how they can be accommodated.

Search neutrality. The digital transformation has revolutionized information consumption, with search engines playing a pivotal role in shaping user access to diverse media. Employing algorithms, these engines influence content visibility and aggregate news sources, significantly molding public opinion. As gatekeepers of information, search engines impact media outlet visibility, affecting online traffic, revenue, and journalistic diversity. In breaking news and societal issues, search engines expedite information dissemination, influencing initial narratives. Understanding their role is crucial for transparency and user access to diverse information. Focusing on movements against police violence, our paper 37 conducts a comparative analysis across 12 search engines for terms "Black Lives Matter" and "Justice pour Adama". Our innovative methodology identifies biases in information diversity, providing insights into the dynamics shaping visibility of societal issues.

Platforms regulation. The regulation of platforms has been a hot topic, particularly in Europe with the Digital Markets Act (DMA) passed in 2022. The purpose of 25 is to design a mathematical model representing a game between long-term revenue-oriented platforms in competition playing with their ranking strategy of items. The objective of platforms is to apply a trade-off between short-term revenue from each visit by displaying most profitable items and long-term number of visits due to the satisfaction of users from the relevance of the displayed items. We analyze the output of the game and the impact of proposed regulation rules on platforms and users.

Information and decision making. In 30, we investigate the mismatched distortion-rate problem formulated by Lapidoth in 1997, where the characterization of the distortion-rate tradeoff is a difficult open problem. This problem also relates to distributed decision-making problems which is under study in 33 for the counter-example of the Witsenhausen (1968). In 2, we investigate the zero-error coding for computing problems with encoder side information formulated by Shannon (1956). An encoder communicates with a decoder that possesses side information Y and aims to retrieve some function with zero probability of error. In this article, we study the side information design problem, which consists of finding the best trade-offs between the quality of the encoder’s side information optal rate. We construct two greedy algorithms that give an achievable set of points in the side information design problem, based on partition refining and coarsening.

8.2 Managing next generation networks & massive IoT communication

Participants: Soraya Ait Chellouche, Yassine Hadjadj-Aoul, Patrick Maillé, Fatima Zahra Mardi.

Optimization of Massive-IoT emission strategies. The expanding Internet of Things landscape, combined with a significant reduction in the cost of connected devices, has enabled the widespread deployment of sensors. These sensors, often situated in close proximity to each other, frequently produce overlapping data. The paper 40 aims to identify such overlaps in sensor data to effectively cluster similar sensors in order to reduce data transmissions. For that, we develop a similarity metric that employs interpolation techniques to manage noisy, irregular, and unsynchronized data. Simulations highlight the superiority of our method compared to the state-of-the-art Dynamic Time Warping distance and a hierarchical clustering with complete linkage inspired by related works.

Efficient resource allocation in Wireless Sensor Networks (WSNs), particularly in LoRaWAN environments, has become increasingly challenging due to limited bandwidth and diverse service requirements. The paper 6 presents three novel resource allocation solutions for LoRaWAN network slicing to address this challenge. These solutions are based on the Multi-Armed Bandit (MAB) algorithm, which is known for balancing the exploration of available actions with the exploitation of optimal decisions. Our objective is to dynamically and efficiently allocate resources to network slices by treating the resource allocation as a MAB problem. This approach aims to maximize Packet Delivery Rate (PDR) performance while ensuring each service’s Service Level Agreement (SLA). The first solution, UCB-MAB, uses the Upper Confidence Bound (UCB) strategy to balance exploration and exploitation to improve network performance. The second solution, Q-UCB-MAB, continuously updates Q-values using the Q-learning update equation and incorporates the UCB strategy for further optimization. Finally, the third solution, ARIMA-UCB-MAB, leverages the predicted reward value from the Autoregressive Integrated Moving Average (ARIMA) model within the UCB framework to enhance network performance. Our results demonstrate that all three solutions offer efficient resource allocation in terms of PDR and SLA satisfaction. Specifically, the ARIMA-UCB-MAB solution outperforms the two other solutions.

Service management in the cloud-edge continuum. With the increasing sophistication and heterogeneity of network infrastructures, the need for Virtual Network Embedding (VNE) is becoming more critical than ever. VNE consists of mapping virtual networks on top of the physical infrastructure to optimize network resource use and improve overall network performance. Considered as one of the most important bricks of network slicing, it has been proven to be an NP-hard problem. Several heuristics and meta-heuristics were proposed to solve it. As heuristics do not provide satisfactory solutions, meta-heuristics allow a good exploration of the solutions’ space, though they require testing several solutions, which is generally unfeasible in a real world environment. Other methods relying on deep reinforcement learning (DRL) and combined with heuristics yield better performance without revealing issues such as sticking at local minima or poor space exploration limits. Nevertheless, these algorithms present varied performances according to the employed approach and the problem to be treated, resulting in robustness problems. To overcome these limits, we propose a robust placement approach based on the Algorithm Selection paradigm. In 1, we propose to dynamically select the best algorithm from a set of learning strategies regarding reward and sample efficiency at each time step. The proposed strategy acts as a meta-algorithm that brings more robustness to the network since it dynamically selects the best solution for a specific scenario. We propose two selection algorithms. First, we consider an offline selection in which the placement strategies are updated outside the selection period. In the second algorithm, the different agents are updated simultaneously with the selection process, which we call an online selection. Both solutions proved their efficiency and managed to dynamically select the best algorithm regarding acceptance ratio of the deployed services. Besides, the proposed solutions succeed in commuting to the best placement strategy depending on the strategies’ strengths while outperforming all standalone algorithms.

Despite the opportunities offered by the recent advances in network virtualization and distributed cloud infrastructure, these developments introduce complexities in the context of multi-domain networks. We propose, in 15 a collaborative orchestration approach based on the message-passing paradigm. Orchestration actions are performed using a Horizontal Pod Autoscaling (HPA) algorithm on each domain independently according to the required and allocated resources. We have developed a simulation testbed based on OMNeT++, allowing the creation and manipulation of multi-domain topologies and customized network slices. We show that exploiting inter-domain communication through message exchanges enables more accurate scheduling decisions than non-communicative orchestration, reducing data rate losses and enabling more efficient resource utilization. The paper 14 presents a distributed version that leverages deep reinforcement learning (DRL) to enhance network function orchestration (NFO) with intelligent scaling decisions, facilitating seamless cross-domain information exchange. First, we develop a DRL agent designed to handle fluctuating traffic loads and generate scaling actions tailored for the considered network function (NF). This trained DRL agent is then integrated into a multi-domain message exchange scaling framework with traffic prediction capabilities. Moreover, we develop a simulation testbed to manipulate multi-domain topologies, customize network slices, and enable precise per-slice and per-domain scaling decisions. Our DRL-based solution outperforms the Horizontal Pod Autoscaling heuristic used by Kubernetes, improving resource utilization and reducing data rate losses.

8.3 Measuring, monitoring & regulation

Participants: Burak Kara, Patrick Maillé, Gerardo Rubino, Bruno Tuffin.

Our activity on monitoring has focused in 2024 on specific contexts but notice that we still work on network neutrality monitoring, starting to develop a demonstrator of the efficiency of existing tools which will help to suggest improvements.

Network neutrality monitoring. With network neutrality regulation imposed worldwide at different extents, there is a need for monitoring tools to verify if Internet service providers (among other actors) comply with the existing rules. Among the very few maintained tools, Wehe has been developed by Northeastern University and has been publicized by the French regulator, ARCEP. The tool runs traces and produces a positive or negative response about service differentiation, but users have to trust that result without any visual perception of the degradation. Our goal in 24 is to describe a demonstrator of Wehe for video streaming, from which we can simultaneously see a baseline video and its manuallydifferentiated counterpart, as well as run Wehe to evaluate its response with respect to the perceived quality degradation. The demonstrator allows to study the efficiency of Wehe in order to propose potential improvements, if needed.

CDNs. The Content Delivery Networks (CDNs) in a multi-CDN delivery have variable requirements and capabilities. Multi-CDN communication over content steering server aims to address this problem in real-time by allowing a CDN to manage its traffic load based on internal metrics via communication with other CDNs over the content steering server while aligning with the standard. The communication process involves operation requests, acceptances, rejections, or negotiations for client offloading/onloading. 21, 22 implement this approach and utilize power efficiency as the internal CDN metric. We present the implementation details and the possible lifecycles of a communication process. Future work includes the comprehensive evaluation and introduction of learning-based approaches by predicting future traffic load and initiating communication process before the internal metrics catastrophe.

Network slices monitoring. Network virtualization enables 5G slicing, a technique for sharing physical resources among isolated slices managed by different actors. However, monitoring slices' performance is becoming challenging due to the significant network overhead associated with direct measurements. To overcome this problem, we propose, in 17, using network tomography to estimate slices' delays in the network. In particular, we investigate the problem of finding the minimal combination of end-to-end simple monitoring paths needed to minimize the estimation error of the slices' delays in a network. To improve the search for the optimal solution, we investigate both a fixed mutation approach and our proposed adaptive mutation approach. Our evaluations show the effectiveness of both approaches; however, the adaptive mutation method outperforms the fixed method by exploring new solutions and avoiding local minima, leading to faster convergence and better results. In 18, we introduce the use of the Relational Graph Convolutional Network (RGCN) model to meet the challenge of generalizing monitors' placement across a network. We conduct a comparative analysis between the RGCN and a traditional Neural Network (NN) across two network topologies, considering every possible node configuration within these topologies. Our findings indicate that the RGCN model, once trained on a specific node pair, exhibits superior generalization ability and accuracy. It consistently transfers its learning to changes in monitors' placement and accurately estimates links delays beyond its initial training monitors. This paper not only demonstrates the effectiveness of the RGCN model in generalizing the monitors' placement problem but also paves the way for its broader application in dynamic network monitoring contexts.

Passive network monitoring and troubleshooting from within the browser. Despite recent advances in terms of network performance, end users still face slow web browsing situations, which can have a range of causes, such as a congested Wi-Fi, a bad wireless signal, or a loaded network or end host. It is thus crucial to monitor the network and troubleshoot the specific causes of slow web browsing, as this benefits end users, operators, and internet service providers alike. Various tools attempting to actively troubleshoot the network through the injection of probes exist. However, these tools are, on the one hand, expensive to run and, on the other hand, not general enough to be able to identify the specific cause of web browsing slowness. The paper 23 addresses the problem by proposing a new lightweight passive measurement solution capable of transforming the web performance measurements collected from within the browser into indicators of network performance anomalies. We validate our solution by emulating a controlled network environment with manually injected anomalies; then, we leverage the measurement data available within the browser to build a predictive model that uses a random forest classifier to correctly categorize the causes of web browsing performance degradation, with an accuracy of over 95%. This implies that one can build on our solution to propose a tool, in the form of a browser extension, that can be used in the wild to monitor the network and shed light on its anomalies by solely relying on a regular user’s web activity.

Resilience in 5G Radio Access networks. Resilience is defined as the ability of a network to resist, adapt to and quickly bounce back from disruptions, to continue keeping an acceptable level of service from users’ perspective. To characterize and analyze the resilience of a 5G Radio Access Network, we must be able to measure current and prospective performance levels using relevant metrics. This year Soumeya Kaada defended her PhD 36 on this topic. In the project, made in collaboration with Nokia, we proposed Markovian models and tools to evaluate the resilience, and we studied methods to improve it on real communication networks.

8.4 Measuring performance metrics based on models

Participants: Bruno Tuffin, Gerardo Rubino, Bruno Sericola.

Our work on performance analysis in 2024 can be decomposed in simulation, with application to resilience, transient analysis, with application to protocols, and distributed systems.

Monte Carlo simulation. We have continued our work on rare event simulation. The paper 26 examines the relative errors (REs) of quantile estimators of various stochastic models under different asymptotic regimes. Depending on the particular limit considered and the Monte Carlo method applied, the RE may be vanishing, bounded, or unbounded. We provide examples of these possibilities.

Network reliability. Network reliability computation is an NP-hard problem which has attracted much attention in literature. In 16, we discuss the Permutation Monte Carlo method, an estimation algorithm which has shown much promise but that is prone to numerical problems in the case of networks with a large number of links. We discuss this situation and we present a simple way to rewrite the algorithm’s computations which is more numerically stable. We present computational results showing the efficiency of our technique for highly reliable networks, in particular when it is applied to topologies with hundreds of links.

Finite sample Ranking and Selection. The paper 28 introduces a new perspective on the problem of finite sample Ranking and Selection. An asymptotically equivalent approximation to the probability of correct selection in terms of power series is derived beyond the classic large deviations principle that has been widely adopted for the design and measurement of allocation policies. The novel approximation method provides more information on the finite sample performance of allocation policies, based on which a new finite computing budget allocation policy is proposed. The asymptotically equivalent approximation may also serve as an estimate of policy performance after allocating the samples. We develop a simple finite computing budget allocation policy based on our approximation and carry out experiments in various settings to show its superiority.

Randomized Quasi-Monte Carlo. Another domain in simulation concerns Randomized Quasi-Monte Carlo (RQMC) methods. Randomized quasi-Monte Carlo methods have been introduced with the main purpose of yielding a computable measure of error for quasi-Monte Carlo approximations through the implicit application of a central limit theorem over independent randomizations. But to increase precision for a given computational budget, the number of independent randomizations is usually set to a small value so that a large number of points are used from each randomized low-discrepancy sequence to benefit from the fast convergence rate of quasi-Monte Carlo. While a central limit theorem has been previously established for a specific but impractical type of randomization, it is also known in general that fixing the number of randomizations and increasing the length of the sequence used for quasi-Monte Carlo can lead to a non-Gaussian limiting distribution. 7 presents sufficient conditions on the relative growth rates of the number of randomizations and the quasi-Monte Carlo sequence length to ensure a central limit theorem and also an asymptotically valid confidence interval. We obtain several results based on the Lindeberg condition and expressed in terms of the regularity of the integrand and the convergence speed of the quasi-Monte Carlo method. We also analyze the resulting estimator's convergence rate. 12 recalls the results and further illustrates that t-bootstrap, thought in the community to be a relevant way to obtain a confidence interval with a good coverage is not as efficient as the regular Student-t one.

Markov chains. In 9 and 11 we continue the development of two transformations of an arbitrary square matrix $M$, one designed to facilitate the calculation of any power of $M$, the other one for the evaluation of the exponential of $M$. In many situations, calculating on the transformed matrices is easier than working with the original ones. The main application of this tool is the evaluation of the transient distribution of an arbitrary Markov chain, in discrete or in continuous time. The idea generalizes the concept of duality introduced by Siegmund, and appears to be very efficient in the analysis of several fundamental models.

Measuring the performance of distributed systems. Distributed systems protocols and network protocols sometimes make use of random walks (i.e., token circulation) on such systems, to explore them in order to collect information from them and to make systems reliable. Important performancemeasures of these protocols such as the search time and the time to visit all the sites obviously depend on some quality measures on random walks such as the hitting and the cover times. The cover time of a discrete-time homogeneous Markov chain or of a random walk is the time needed to visit all the states of the process. We analyze in 8 both the distribution and the moments of the cover time of a general Markov chain and we are interested in exact results instead of asymptotic values of the mean cover time which are generally considered in the literature. These results are applied to particular graphs namely the generalized cycle graph, the complete graph and the generalized path graph. We analyze in detail in 39 the moments and the distribution of the hitting and cover times of a random walk in the specific case of the complete graph. We obtain recurrence relations for the moments of all orders and we use these relations to analyze the asymptotic behavior of the hitting and cover times distributions when the number of states tends to infinity. We perform a similar analysis in 38 on the lollipop graph which is the graph exhibiting the maximum expected hitting time among all the graphs having the same number of nodes.

8.5 Artificial Intelligence (AI) in networking

Participants: Yassine Hadjadj-Aoul, Gerardo Rubino, César Viho.

On Time Series Forecasting in climate change.

In 10, we describe two projects whose last works were done in 2024, and will finish in 2025. One of them concernes Echo State Networks, and the other one, being the object of this paragraph, concernes forecasting macro-characteristics of a future period of time, typically next summer. The project involved people from Uruguay, Argentina, Chile and France. The goal was to forecast the maximal temperature reached in, say, next summer. In the paper, we report partial result concerning the project. The main tool used was Auto Machine Learning.

Drift detection in time series. Modeling non-stationary data in time series analysis is a challenging problem in the field of continual learning, and data distribution shifts may result in negative consequences on the performance of a machine learning model. Classic learning tools are often vulnerable to perturbations of the input covariates, and are sensitive to outliers and noise, and some tools are based on rigid algebraic assumptions. Distribution shifts are frequently occurring due to changes in raw materials for production, seasonality, a different user base, or even adversarial attacks. Therefore, there is a need for more effective distribution shift detection techniques. In 27, we propose a continual learning framework for monitoring and detecting distribution changes. We explore the problem in a latent space generated by a bio-inspired self-organizing clustering and statistical aspects of the latent space. In particular, we investigate the projections made by two topology-preserving maps: the Self-Organizing Map and the Scale Invariant Map. Our method can be applied in both a supervised and an unsupervised context. We construct the assessment of changes in the data distribution as a comparison of Gaussian signals, making the proposed method fast and robust. We compare it to other unsupervised techniques, specifically Principal Component Analysis (PCA) and Kernel-PCA. Our comparison involves conducting experiments using sequences of images (based on MNIST and injected shifts with adversarial samples), chemical sensor measurements, and the environmental variable related to ozone levels. The empirical study reveals the potential of the proposed approach.

Network Traffic Classification for Multi-Activity Situations Detection. Classifying network traffic is one of the most crucial aspects of monitoring. In order to facilitate network management, network traffic classification has historically concentrated on identifying individual activities, such as streaming or talking. However, new methods are required to precisely identify and react to these patterns when users exhibit more multi-activity behaviors. We have proposed a strategy to identify single and multiple activities within a network. Experimental results from both studies show that these multi-activity recognition methods achieve performance levels that match or even exceed those of traditional, single-activity-focused approaches, highlighting the potential for these techniques to improve quality of service and network management strategies in modern, dynamic network environments. Similarly, this year we propose to go a step further by identifying the specific activities and applications being used. This more granular approach provides deeper insights into user behavior patterns, enabling more fine-tuned network optimizations 34.

Adaptative AI for an Efficienct Schedulers in Wireless Networks. With the increased demands for 5G networks and the limited radio resources, providing high spectral efficiency, low delay, low energy consumption, and other Key Performance Indicators (KPIs) is a challenging task. Extensive research has been conducted to propose efficient solutions for specific objectives and contexts. Although these solutions (often heuristicbased) are highly effective in specific contexts, their performance diminishes when applied in different conditions. This implies difficulties in adapting to environmental variations and/or changes in objectives. In order to overcome this problem, we propose, in 29, an approach employing reinforcement learning to dynamically derive the formula for a scheduler that can be adapted to any context and objective. The proposed solution is validated with a Proof of Concept (PoC), which highlights the AI ability to identify the adequate scheduler to optimize spectral efficiency in different traffic loads contexts.

ML-based Network Slicing. Network slicing, a key component of post-5G networks, has brought virtual network embedding (VNE) to the forefront of networking research. However, existing VNE approaches are limited by their consideration of static network topologies, failing to account for the dynamic nature of real-world networks. This limitation becomes particularly problematic in the face of link failures or topology modifications, rendering these approaches ineffective. To address this, the paper 4 proposes a new service placement strategy that remains effective even when the network topology changes. Furthermore, we introduce several service migration strategies and thoroughly investigate their effectiveness. Our results demonstrate the adaptability of our proposed strategy, which leverages graph neural networks, in handling link failures without necessitating relearning. This adaptability underlines the potential of our approach to significantly enhance the robustness and flexibility of service migration in VNE, thereby contributing to the evolution of network slicing in post-5G networks.

Using Self Organizing Maps to model cognitive deseases. Since the 1990s, Self-Organizing Maps (SOMs) have been instrumental in reducing dimensionality and visualizing high-dimensional data. In 13 and 19, we modified SOMs to explore the neural representation of human concepts, their neural ‘word net’ mapping, and the deterioration of these mappings in certain neurological disorders. Our model draws inspiration from semantic dementia, a severe condition that degrades semantic knowledge in the brain. Although our exploration utilizes a low-dimensional model —- a rough simplification with respect of our brains —-, it successfully replicates observed clinical patterns. These promising results inspire further research to enhance our understanding of language pathophysiology in neurological disorders. The article 19 continues the work presented in 13, exploring complemetary aspects of the model.

Enhancing Autonomous Driving Navigation Using AI. Autonomous vehicles have gained extensive attention in recent years, both in academia and industry. For these self-driving vehicles, decision-making in urban environments poses significant challenges due to the unpredictable behavior of traffic participants and intricate road layouts. While existing decision-making approaches based on Deep Reinforcement Learning (DRL) show potential for tackling urban driving situations, they suffer from slow convergence, especially in complex scenarios with high mobility. In 3, we present a new approach based on the Soft Actor-Critic (SAC) algorithm to control the autonomous vehicle to enter roundabouts smoothly and safely and ensure it reaches its destination without delay. For this, we introduce a destination vector concatenated with extracted features using Convolutional Neural Networks (CNN). To evaluate the performance of our model, we conduct extensive experiments in the CARLA simulator and compared it with the Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) models. Qualitative results reveal that our model converges rapidly and achieves a high success rate in scenarios with high traffic compared to the DQN and PPO models.

9.1 Bilateral contracts with industry

9.2 Bilateral Grants with Industry

10.1 International initiatives

10.2 National initiatives

10.3 Regional initiatives

11.1 Promoting scientific activities

11.2 Teaching - Supervision - Juries

11.3 Popularization

12.1 Publications of the year