7.1.1 Queueing Systems

• Functional Description: This tool aims at providing a simple web interface to promote the use of our proposed solutions to numerically solve classical queueing systems.
• URL: http://queueing-systems.ens-lyon.fr/
• Contact: Thomas Begin
• Participants: Alexandre Brandwajn, Thomas Begin

7.1.2 WSNet

• Keyword: Network simulator
• Functional Description: WSNet is a modular event-driven simulator targeted to Wireless Sensor Networks. Its main goals are to offer scalabiliy, extensibility and modularity for the integration of new protocols/hardware models and a precise radio medium simulation. We still hope to find the proper resource to make WSNet evolve into a wireless capillary network simulator suitable for conducting simulations at the urban scale.
• URL: https://gforge.inria.fr/projects/wsnet-3/
• Contact: Guillaume Chelius
• Participants: Rodrigue Domga Komguem, Fabrice Valois
• Partner: CEA-LETI

7.1.3 Wi-Fi experimentation platform

• Keywords: Wi-Fi, Hardware and Software Platform
• Functional Description: Networking experimental platform in order to test and evaluate the Wi-Fi communication technology in various configurations. The platform consists in a server and Wi-Fi nodes. The server, remotely manageable, can deploy the images on the Wi-Fi nodes. These latter and the server are connected via a wired network.
• Contact: Isabelle Guérin-Lassous

7.1.4 FAuST

• Keywords: Learning, Sparsity, Fast transform, Multilayer sparse factorisation
• Scientific Description: FAuST allows to approximate a given dense matrix by a product of sparse matrices, with considerable potential gains in terms of storage and speedup for matrix-vector multiplications.
• Functional Description:

  FAUST is a C++ toolbox designed to decompose a given dense matrix into a product of sparse matrices in order to reduce its computational complexity (both for storage and manipulation).

  Faust includes Matlab and Python wrappers and scripts to reproduce the experimental results of the following papers: - Le Magoarou L. and Gribonval R,. "Flexible multi-layer sparse approximations of matrices and applications", Journal of Selected Topics in Signal Processing, 2016. - Le Magoarou L., Gribonval R., Tremblay N. "Approximate fast graph Fourier transforms via multi-layer sparse", IEEE Transactions on Signal and Information Processing over Networks, 2018

• Release Contributions:

  Faust 1.x contains Matlab routines to reproduce experiments of the PANAMA team on learned fast transforms.

  Faust 2.x contains a C++ implementation with preliminary Matlab / Python wrappers.

  Faust 3.x includes Python and Matlab wrappers around a C++ core with GPU acceleration.

• News of the Year:

  In 2020, major efforts were put into finalizing Python wrappers, producing tutorials using Jupyter notebooks and Matlab livescripts, as well as substantial refactoring of the code to optimize its efficiency and exploit GPUs.

  In april 2018, a Software Development Initiative (ADT REVELATION) started in for the maturation of FAuST. A first step was to complete and robustify Matlab wrappers, to code Python wrappers with the same functionality, and to setup a continuous integration process. A second step was to simplify the parameterization of the main algorithms. The roadmap for next year includes showcasing examples and optimizing computational efficiency. – In 2017, new Matlab code for fast approximate Fourier Graph Transforms have been included. based on the approach described in the papers:

  -Luc Le Magoarou, Rémi Gribonval, "Are There Approximate Fast Fourier Transforms On Graphs?", ICASSP 2016 .

  -Luc Le Magoarou, Rémi Gribonval, Nicolas Tremblay, "Approximate fast graph Fourier transforms via multi-layer sparse approximations", IEEE Transactions on Signal and Information Processing over Networks,2017.

• URL: http://faust.inria.fr/
• Publications: hal-01416110, hal-01627434, hal-01167948, hal-01254108, tel-01412558, hal-01156478, hal-01104696, hal-01158057, hal-03132013
• Contact: Rémi Gribonval
• Participants: Luc Le Magoarou, Nicolas Tremblay, Rémi Gribonval, Nicolas Bellot, Adrien Leman, Hakim Hadj-Djilani

8.1.1 Diffused Wasserstein Distance for Optimal transport between attributed graphs

Participants: Paulo Gonçalves, Rémi Gribonval, Amélie Barbe, Titouan Vayer.

We proposed the Diffusion Wasserstein (DW) distance, as a generalization of the standard Wasserstein distance to undirected and connected graphs where nodes are described by feature vectors. DW is based on the Laplacian exponential kernel and benefits from the heat diffusion to catch both structural and feature information from the graphs. We further derived lower/upper bounds on DW and showed that it can be directly plugged into the Fused Gromov Wasserstein (FGW) distance that has been recently proposed, leading - for free - to a DifFused Gromov Wasserstein distance (DFGW) that allows a significant performance boost when solving graph domain adaptation tasks 31, 30.

8.1.2 Sparse tensor dimensionality reduction with application to the clustering of functional connectivity in the brain

Participants: Paulo Gonçalves, Gaetan Frusque.

Within the framework of G. Frusque PhD work 46, we addressed the problem of clustering FC into relevant ensembles of simultaneously activated components, yielding a multiplex network that reveals characteristic patterns of the epileptic seizures of a given patient. While k-means is certainly the most popular method for data clustering, it is known to perform poorly on large dimensional data sets, and to be highly sensitive to noise. To overcome the so- called curse of dimensionality, we proposed a new tensor decomposition to reduce the size of the data set formed by FC time-series recorded for several seizures, prior to apply k-means. We proposed an adapted procedure to infer a multiplex network from several FC time series, and we emphasised one particular variant that imposes sparsity constraint 20, 34, 33. Then, we conducted a real case study, applying the proposed sparse tensor decomposition to iEEG data to infer a multiplex network corresponding to the different stages of an epileptic seizure 19.

In another collaboration, we conducted a theoretical analysis of a variant of the Wasserstein distance for optimal transport, where minibatches are used to reduce complexity 44.

8.1.3 Neural networks : approximation theory and compression algorithms

Participants: Rémi Gribonval, Pierre Stock.

From an approximation theoretic perspective, we studied the expressivity of deep neural networks and highlighted the role of depth to enable efficient approximation of functions with very limited smoothness 23.

From a more computational perspective, within the framework of the Ph.D. of Pierre Stock, we proposed a technique to drastically compress neural networks using product quantization 45, and an approach to learn networks that can be more efficiently quantized 50

8.1.4 Compressive learning : guarantees on privacy preservation and statistical significance

Participants: Rémi Gribonval.

The compressive learning framework proposes to deal with the large scale of datasets by compressing them into a single vector of generalized random moments, called a sketch, from which the learning task is then performed. We provided a theoretical analysis establishing statistical guarantees on the generalization error of this procedure, first in a general abstract setting illustrated on PCA 22, then for the specific case of compressive $k$-means and compressive Gaussian Mixture Modeling 52.

In the context of the Ph.D. thesis of Antoine Chatalic (in the PANAMA team in Rennes) we showed 16 that a simple perturbation of this mechanism with additive noise is sufficient to satisfy differential privacy, a well established formalism for defining and quantifying the privacy of a random mechanism. We combined this with a feature subsampling mechanism, which reduces the computational cost without damaging privacy. The framework was applied to the tasks of Gaussian modeling, k-means clustering and principal component analysis (PCA), for which sharp privacy bounds were derived. Empirically, the quality (for subsequent learning) of the compressed representation produced by this mechanism is strongly related with the induced noise level, for which we gave analytical expressions.

A tutorial paper on the principle and the main guarantees of compressive learning was prepared 53.

8.1.5 Sparse matrix factorizations: algorithms and identifiability guarantees

Participants: Rémi Gribonval, Elisa Ricietti, Marion Foare, Léon Zheng, Quoc-Tung Le.

Matrix factorization with sparsity constraints plays an important role in many machine learning and signal processing problems such as dictionary learning, data visualization, dimension reduction. From a theoretical perspective, we investigated the uniqueness properties of sparse matrix factorizations. From an algorithmic perspective, we analyzed and fixed a weakness of proximal algorithms in sparse matrix factorization. We also described a new tractable proximal operator called Generalized Hungarian Method, associated to so-called $k$-regular matrices, which are useful for the factorization of a class of matrices associated to fast linear transforms. We further illustrated the effectiveness of our proposals by numerical experiments on the Hadamard Transform and magnetoencephalography matrix factorization 37, 56. The new proximal operator was implemented in the FA$\mu$ST software library.

8.1.6 Sparsity-based audio declipping

Participants: Rémi Gribonval.

We provided a selected overview of sparsity-based audio declipping methods, as well as some new algorithms and a large-scale evaluation. Recent advances in audio declipping have substantially improved the state of the art. Yet, practitioners need guidelines to choose a method, and while existing benchmarks have been instrumental in advancing the field, larger-scale experiments are needed to guide such choices. First, we showed that the clipping levels in existing small-scale benchmarks are moderate and call for benchmarks with more perceptually significant clipping levels. We then proposed a general algorithmic framework for declipping that covers existing and new combinations of variants of state-of-the-art techniques exploiting time-frequency sparsity: synthesis vs. analysis sparsity, with plain or structured sparsity. Finally, we systematically compared these combinations and a selection of state-of-the-art methods. Using a large-scale numerical benchmark and a smaller scale formal listening test, we provided guidelines for various clipping levels, both for speech and various musical genres. The code was made publicly available for the purpose of reproducible research and benchmarking 21.

8.2.1 Interpretable socioeconomic status inference from aerial imagery through urban patterns

Participants: Márton Karsai, Jacobo Levy Abitbol.

Urbanization is a great challenge for modern societies, promising better access to economic opportunities, but widening socioeconomic inequalities. Accurately tracking this process as it unfolds has been challenging for traditional data collection methods, but remote sensing information offers an alternative way to gather a more complete view of these societal changes. By feeding neural networks with satellite images, the socioeconomic information associated with that area can be recovered. However, these models lack the ability to explain how visual features contained in a sample trigger a given prediction. In our work published in 11, we closed this gap by predicting socioeconomic status across France from aerial images and interpreting class activation mappings in terms of urban topology. We showed that trained models disregard the spatial correlations existing between urban class and socioeconomic status to derive their predictions. These results pave the way to build more interpretable models, which may help to better track and understand urbanization and its consequences.

8.2.2 Dynamics of cascades on burstiness-controlled temporal networks

Participants: Márton Karsai, Samuel Unicomb.

In the work published in 28, we developed a master equation formalism to study cascades on temporal networks with burstiness modelled by renewal processes. Supported by numerical and data-driven simulations, we described the interplay between heterogeneous temporal interactions and models of threshold-driven and epidemic spreading. We found that increasing interevent time variance can both accelerate and decelerate spreading for threshold models, but can only decelerate epidemic spreading. When accounting for the skewness of different interevent time distributions, spreading times collapse onto a universal curve. Our framework uncovers a deep yet subtle connection between generic diffusion mechanisms and underlying temporal network structures that impacts a broad class of networked phenomena, from spin interactions to epidemic contagion and language dynamics

8.2.3 Temporal social network reconstruction using wireless proximity sensors: model selection and consequences

Participants: Márton Karsai, Sicheng Dai.

In our study published in 17, we analysed a proximity dataset collected during a longitudinal social experiment aiming to understand the co-evolution of children’s language development and social network. Physical proximity and verbal communication of hundreds of pre-school children and their teachers were recorded over three years using autonomous wearable low power wireless devices. The dataset is accompanied with three annotated ground truth datasets, which record the time, distance, relative orientation, and interaction state of interacting children for validation purposes. We used this dataset to explore several pipelines of dynamical event reconstruction including earlier applied naïve approaches, methods based on Hidden Markov Model, or on Long Short-Term Memory models, some of them combined with supervised pre-classification of interaction packets. We found that while naïve models propose the worst reconstruction, Long Short-Term Memory models provide the most precise way to reconstruct real interactions up to  90% accuracy. Finally, we simulated information spreading on the reconstructed networks obtained by the different methods. Results indicate that small improvement of network reconstruction accuracy may lead to significantly different spreading dynamics, while sometimes large differences in accuracy have no obvious effects on the dynamics. This not only demonstrates the importance of precise network reconstruction but also the careful choice of the reconstruction method in relation with the data collected. Missing this initial step in any study may seriously mislead conclusions made about the emerging properties of the observed network or any dynamical process simulated on it.

8.2.4 Socio-geography of mobile telephony

Participants: Eric Guichard, Thomas Begin, Anthony Busson.

Within the “Better-Net’’ IPL leaded by Isabelle Chrisment (Loria), we conducted an original research on the socio-geography of mobile telephony. By combining the division of inhabited France into two million 200 m2 squares proposed by INSEE and public data relating to mobile telephony media, antennas and transmitters, an online atlas was created. It is available at the URL http://betternet-rumbi.lhs.loria.fr/website/antennes.html. It proposes an interactive map of France, that can be zoomed in up to the scale of the dwelling, with the precise location of the transmitters, completed (when it exists) with information relating to the place and the nature of their medium, their operator and their generation. This map also integrates the 900m2 squares with more than 10 inhabitants, enriched with the following socio-demographic details:

• the percentages of young people aged 15 to 25,
• the percentages of poor households
• and the total population of these small squares.

Readers can choose a department, a postcode, and one of the three variables mentioned above. They can also hide sparsely populated squares. And any visualization is memorizable via its URL.

Such mapping attempts are rare. Those that do exist in France (Arcep, etc.) are not correlated with sociological data. This work was conducted in collaboration with Antoine Chemardin, an Inria engineer and will give rise to an article in the review Mappemonde (expected publication: 2021).

8.3.1 Quantum communications

The work on performance analysis of a multipartite quantum switch has been continued (cf. description in Section 7.3.1 in Dante 2019 activity report). In 26 the stability of a multipartite switch with infinite memory and identical links is derived via Lyapounov techniques for Markov chains. The capacity and the number of stored Qbits are found in explicit form. In 29 a bipartite switch is analyzed via a discrete-time Markov chain (DTMC). The aim is to compare the results with that obtained in 26 where a continuous-time Markov chain (CTMC) is used, the latter model being considerably easier to manipulate than the DTMC model. We conclude that the CTMC model is overall a suitable model. In 39 we study a quantum switch serving a set of users in a star topology. The function of the switch is to create bipartite or tripartite entangled state among users at the highest possible rates at a fixed ratio. We model a set of randomized policies for sharing the switch bipartite and tripartite state generation and find that the best policy outperforms a TDM (time-division multiplexing) policy. However, the performance improvement decreases as the number of users grows. The model can be augmented to capture Qbit decoherence and we show that the latter has little effect on the switch capacity.

8.3.2 Network caching

Caching systems have long been crucial for improving the performance of a wide variety of network and web based online applications. In such systems, end-to-end application performance heavily depends on the fraction of objects transfered from the cache, also known as the cache hit proba-bility. Many cache eviction policies have been proposed and implemented to improve the hit probability. In 38, we propose a new method to compute an upper bound on hit probability for all non-anticipative caching policies, i.e. for policies that have no knowledge of future requests. At object request arrival, we use hazard rate HR function based ordering to classify the request as a hit or not. Under some statistical assumptions, we prove that our proposed HR based ordering model computes the maximum achievable hit probability and serves as an upper bound forall non-anticipative caching policies. We also provide simulation results to validate its correctness and to compare it to Belady’s upper bound. We find it to almost always be tighter than Belady’s bound.

8.3.3 Resource Allocation

The work on 1-D ressource allocation has been continued (see description in Section 7.3.2 of 2019 Dante activity report). A first set of extensions in 27 consists of a preliminary cost model where the communication cost is a function of the request distance (this model is generalized in 55), an extension in which each user requests two resources, and an approximation algorithm for the optimal allocation problem in a 2-D setting where users and servers are located on a 2-D Euclidean plane. In 55 we compute the expected communication cost associated with the move-to-right policy in a 1-D service network. Users are distributed according to a Poisson process. This 1-D network setting applies to various service networks, in particular in the context of vehicular wireless ad-hoc networks on a one-lane roadway, where users are in vehicles and resources are attached to fixed infrastructure such as lamp posts. We use results in 27 to map the request distance of an assigned user in a 1-D service network to the sojourn time of a customer in an exceptional service accessible batch queueing system. We compute the Laplace-Stieltjes transform of the sojourn time distribution for this queueing system and in the process also generate new results for batch service queues.

8.3.4 VoD broadcasting over vehicular networks

Participants: Thomas Begin, Anthony Busson, Isabelle Guérin Lassous.

In this work 14, we introduce a relatively simple and versatile analytical model to the performance computation of an IEEE 802.1 1-basedVideo-on-Demand (VoD) service for connected vehicles traveling along a highway or an expressway. We use a probabilistic approach to account for the intrinsic dynamics (mobility) of vehicles, the density of subscriber vehicles, the distance between Road Side Units (RSUs), as well as the video codec bit rate. Our solution delivers estimates for the attained throughput of each subscriber vehicle as well as for their interruption times in the playback of their video. We studied several scenarios to assess the accuracy of our approximate solution comparing its results with those of a discrete-event simulator with realistic vehicular traffic trajectories, video traffic, and networking protocols. Overall, our solution is found to be accurate delivering estimates in good agreement with the simulation results. Through several applications, we illustrate how our analytical solution can help in the design of a VoD service for vehicles for capacity planning purposes, or at improving its performance with the use of a blocking policy.

8.3.5 Towards a Fast and Efficient Strategy to Assign Channels in WLANs with Channel Bonding

Participants: Thomas Begin, Anthony Busson, Marija Stojanova.

With the aim of increasing wireless data rates, IEEE 802.11n introduced the possibility for WLAN nodes to bond two channels into a single channel. However, channel bonding also limits spacial reutilization and complexifies channel assignment. In this work 42, we present a fast and efficient solution for channel width selection and channel assignment in 802.11 WLANs using channel bonding. The proposed algorithm uses a novel, graph-centric metric to propose a single channel width for all the APs of the WLAN aiming at avoiding starvation in any of the network's APs. Decoupling the choice of channel width and channel assignment results in a scalable approach that bypasses the usual complexity issues of classic channel assignment schemes. We test the solution's precision in choosing a suited channel width and assignment by comparing its results with those delivered by the ns-3 network simulator. We obtain that, in the large majority of the cases, the choice made by our solution matches the simulation results.

8.3.6 Distributed Congestion Control mechanism for NANs

Participants: Thomas Begin, Anthony Busson, Juan Pablo Astudillo.

The need for significant improvements in the management and efficient use of electrical energy has led to the evolution from the traditional electrical infrastructures towards modern Smart Grid networks. Taking into account the critical importance of this type of networks, multiple research groups focus their work on issues related to the generation, transport and consumption of electrical energy. One of the key research points is the data communication network associated with the electricity transport infrastructure, and specifically the network that interconnects the devices in consumers' homes, the so-called Neighborhood Area Networks (NANs). In this paper, a new distributed congestion control mechanism is proposed, implemented and evaluated for NANs. Besides, different priorities have been considered for the traffic flows transmitted by different applications. The main goal is to provide with the needed Quality of Service (QoS) to all traffic flows, especially in high traffic load situations. The proposal is evaluated in the context of a wireless ad hoc network made up by a set of smart meter devices, using the Ad hoc On-Demand Distance Vector (AODV) routing protocol and the IEEE 802.11ac physical layer standard. The application of the proposed congestion control mechanism, together with the necessary modifications made to the AODV protocol, lead to performance improvements in terms of packet delivery ratio, network throughput and transit time, fairness between different traffic sources and QoS provision 13.

8.3.7 Leveraging Antenna Orientation to Optimize Network Performance of Fleets of UAVs

Participants: Rémy Grünblatt, Isabelle Guérin-Lassous.

In this work 35, we investigate the problem of optimizing the network performance of a fleet of unmanned aerial vehicles (UAVs) in static positions. More precisely, we allow each UAV to change its orientation in order to improve the quality of communication with its neighbours. This form of controlled mobility takes advantage of the effective radiation pattern of each UAV. We build a decentralized scheme based on the hill climbing optimization approach without a priori knowledge of the antennas radiation patterns. Then, we propose a simulation framework, based on ns–3, allowing to evaluate the gain in network performance. We provide results in several deployment scenarios involving different rate adaptation algorithms and network sizes.

8.3.8 A Passive Method to Infer the Weighted Conflict Graph of a IEEE 802.11 Network

Participants: Lafdal Abdelwedoud, Anthony Busson, Isabelle Guérin-Lassous, Marion Foare.

We consider an IEEE 802.11 network composed of several Access Points (APs) managed by one controller. The controller relies on pieces of information describing the network state as channels, load, associated stations, conflicts, etc. to configure and optimize the network. In this work 10, we propose a method that infers the way the different channels are shared between APs according to the Clear Channel Assessment (CCA) mechanism. It is represented through a conflict graph where an edge exists if two APs are able to detect each other. As this detection is sometimes partial, the links are weighted. Our method relies on measures already available on most of Wi-Fi products and does not generate any traffic except the transmission of these measures to the controller. A Markov network and an optimization problem are then proposed to infer the weights of the conflict graph. Our solution is shown accurate on a large set of simulations performed with the network simulator ns-3.

Informal international partners

• Univ. of California, Santa Cruz (USA), Collaboration with Pr. Brandwajn

  Participants: Thomas Begin.

  Goals: Designing efficient solutions to solve multi-server queueing systems.

• Polytechnic University of Catalonia (Spain), Collaboration with the network Research Team from UPC (Ass. Pr. M. Aguilar and Ass. Pr. Luis J. de la Cruz Llopis)

  Participants: Thomas Begin, Anthony Busson, Isabelle Guérin Lassous.

  Goals: Designing tools to improve the sharing of resources in vehicular networks.

• Univ. of Ottawa (Canada), Collaboration with the Diva lab headed by Pr. Boukerche

  Participants: Thomas Begin, Anthony Busson, Isabelle Guérin Lassous.

  Goals: Evaluating the feasbility of distributing video on demand (VoD) on vehicular networks.
• Univ. Louvain (Belgium), Collaboration with the team of Laurent Jacques

  Participants: Rémi Gribonval.

  Topic: compressive learning and privacy guarantees
• Ohio State Univ. (USA), Collaboration with the group of Phil Schniter

  Participants: Rémi Gribonval.

  Topic: algorihms for compressive learning

10.2.1 Visits of international scientists

• Visit from Pr. Brandwajn (Univ. of California, Santa Cruz) in September 2020, funded by the ENS Lyon.

10.3.1 ANR IA Chaire : AllegroAssai

Participants: Rémi Gribonval, Paulo Gonçalves, Elisa Ricietti, Marion Foare, Léon Zheng, Quoc-Tung Le, Titouan Vayer, Ayoub Belhadji, Clement Lalanne.

Duration of the project: 2020 - 2024.

AllegroAssai focuses on the design of machine learning techniques endowed both with statistical guarantees (to ensure their performance, fairness, privacy, etc.) and provable resource-efficiency (e.g. in terms of bytes and flops, which impact energy consumption and hardware costs), robustness in adversarial conditions for secure performance, and ability to leverage domain-specific models and expert knowledge. The vision of AllegroAssai is that the versatile notion of sparsity, together with sketching techniques using random features, are key in harnessing these fundamental tradeoffs. The first pillar of the project is to investigate sparsely connected deep networks, to understand the tradeoffs between the approximation capacity of a network architecture (ResNet, U-net, etc.) and its “trainability” with provably-good algorithms. A major endeavor is to design efficient regularizers promoting sparsely connected networks with provable robustness in adversarial settings. The second pillar revolves around the design and analysis of provably-good end-to-end sketching pipelines for versatile and resource-efficient large-scale learning, with controlled complexity driven by the structure of the data and that of the task rather than the dataset size.

10.3.2 ANR DataRedux

Participants: Paulo Gonçalves, Rémi Gribonval, Marion Foare, Israel Campero Jurado.

Duration of the project: February 2020 - January 2024.

DataRedux puts forward an innovative framework to reduce networked data complexity while preserving its richness, by working at intermediate scales (“mesoscales”). Our objective is to reach a fundamental breakthrough in the theoretical understanding and representation of rich and complex networked datasets for use in predictive data-driven models. Our main novelty is to define network reduction techniques in relation with the dynamical processes occurring on the networks. To this aim, we will develop methods to go from data to information and knowledge at different scales in a human-accessible way by extracting structures from high-resolution, diverse and heterogeneous data. Our methodology will involve the identification of the most relevant subparts of time-resolved datasets while remapping the remaining parts of the system, the simultaneous structural-temporal representations of time-varying networks, the development of parsimonious data representations extracting meaningful structures at mesoscales (“mesostructures”), and the building of models of interactions that include mesostructures of various types. Our aim is to identify data aggregation methods at intermediate scales and new types of data representations in relation with dynamical processes, that carry the richness of information of the original data, while keeping their most relevant patterns for their manageable integration in data-driven numerical models for decision making and actionable insights.

10.3.3 ANR Darling

Participants: Paulo Gonçalves, Rémi Gribonval, Marion Foare.

Duration of the project: February 2020 - January 2024.

This project meets the compelling demand of developing a unified framework for distributed knowledge extraction and learning from graph data streaming using in-network adaptive processing, and adjoining powerful recent mathematical tools to analyze and improve performances. The project draws on three major parallel directions of research: network diffusion, signal processing on graphs, and random matrix theory which DARLING aims at unifying into a holistic dynamic network processing framework. Signal processing on graphs has recently provided a comprehensive set of basic instruments allowing for signal on graph filtering or sampling, but it is limited to static signal models. Network diffusion on the opposite inherently assumes models of time varying graphs and signals, and has pursued the path of proposing and understanding the performance of distributed dynamic inference on graphs. Both areas are however limited by their assuming either deterministic graph or signal models, thereby entailing often inflexible and difficult-to-grasp theoretical results. Random matrix theory for random graph inference has taken a parallel road in explicitly studying the performance, thereby drawing limitations and providing directions of improvement, of graph-based algorithms (e.g., spectral clustering methods). The ambition of DARLING lies in the development of network diffusion-type algorithms anchored in the graph signal processing lore, rather than heuristics, which shall systematically be analyzed and improved through random matrix analysis on elementary graph models. We believe that this original communion of as yet remote areas has the potential to path the pave to the emergence of the critically needed future field of dynamical network signal processing.

10.3.4 ANR DylNet

Participants: Márton Karsai.

Duration of the project: September 2016 - September 2020.

The DylNet project aims to observe and to characterise the relationships between childhood sociability and oral-language learning at kindergarten. With a view to this, it takes an multidisciplinary approach combining work on language acquisition, sociolinguistics, and network science. It will be implemented by following all the children ($\approx 220$) and teaching staff in one kindergarten over a 3-year period. The use of wireless proximity sensors will enable collection of social contacts throughout the study. The data on sociability will be linked to the results of language tests and recordings of verbal interactions used to follow the children's progress on both a psycholinguistic level (lexicon, syntax, pragmatics) and a sociolinguistic level (features showing belonging to a social group). The aim is to better understand the mechanisms of adaptation and integration at work when young children first come into contact with the school context.

10.3.5 CNRS GreenHMR

Participants: Isabelle Guérin Lassous, Thomas Begin, Anthony Busson.

Duration of the project: March 2020 - Decembre 2020.

This project is an emergent collaborative project. The goal of the project is to propose end-to-end solutions to reduce energy consumption in heterogeneous wireless networks while offering a good level of performance to users. This project is based on different scientific steps: energy real measurements, modeling and solution designing.

10.3.6 Inria IPL BetterNet

Participants: Éric Guichard, Thomas Begin, Anthony Busson.

BetterNet aims at building and delivering a scientific and technical collaborative observatory to measure and improve the Internet service access as perceived by users. In this Inria Project Lab, we will propose new original user-centered measurement methods, which will associate social sciences to better understand Internet usage and the quality of services and networks with a particular focus on geography and cartography.

10.4.1 Idex Lyon - ACADEMICS

Participants: Paulo Gonçalves, Rémi Gribonval, Marion Foare, Amélie Barbe, Titouan Vayer, Gaetan Frusque.

Duration of the project: September 2018 - December 2021.

The project brings together a consortium of 4 teams from Laboratories of Université de Lyon (UdL) and will form a working group with complementary expertise in machine learning (deep learning, statistical learning, data mining), in data science (complex data analysis, adaptive and/or data-driven methods, network science) and in the studies of climate modeling and of computational social science. It comprises:

• Laboratoire Informatique du Parallélisme (LIP): P. Gonçalves (PI), M. Karsai (PI for Comp. Social Sc.)
• Laboratoire de Physique (LP): P. Borgnat (Coordinator), F. Bouchet (PI for Climate)
• Laboratoire Hubert Curien (LabHC), Université Jean Monnet: M. Sebban (PI)
• Laboratoire d'InfoRmatique en Images et Systèmes d'information (LIRIS): C. Robardet (PI)

The impacts of the project will stem from the efficiency of our proposed methods to learn from complex and dynamic data, and if so, future applications will naturally follow in many areas: social science and study of social interactions, climate and environmental science but also in technological networks, neuroscience with the study of brain networks and more generally in any domain where effective dynamical models of complex situations are to be learned from data. All these situations go beyond the current classical applicative frameworks of ML (time measurements, 2D images, or texts) and compel us to work out a major scientific breakthrough.

11.1.1 Scientific events: organisation

• Thomas Begin was co-organiser of the “Journées Cloud 2020” (with the financial help of the action “Virtualisation et Cloud” of the GdR RSD (Réseaux et Systèmes Distribués) of CNRS). More than 90 participants

  https://journeescloud20.sciencesconf.org/

11.1.2 Scientific events: selection

11.1.3 Journal

11.1.4 Invited talks

• Rémi Gribonval was an invited speaker at the IPAM Workshop “PDE and Inverse Problem Methods in Machine Learning”, Los Angeles, April 2020 (participation canceled due to Covid-19), a plenary speaker at the SIGMA’2020 workshop (Signal, Image, Géométrie, Modélisation, Approximation), CIRM, Marseille, April 2020 (canceled due to Covid-19), at the workshop “Optimization for Machine Learning”, CIRM, Luminy, March 2020 (participation canceled due to Covid-19)

11.1.5 Scientific expertise

• Isabelle Guérin Lassous is an expert for the MESRI / DAEI (Ministère de l'Enseignement Supérieur, de la Recherche et de l'Innovation / Délégation aux affaires européennes et internationales)
• Rémi Gribonval is a member of the Scientific Advisory Board (vice-president) of the Acoustics Research Institute of the Austrian Academy of Sciences, and a member of the Commission Prospective of Institut de Mathématiques de Marseille
• Rémi Gribonval is a member of the Cellule ERC of INS2I

11.1.6 Research administration

• P. Gonçalves is scientific correspondent for the international partnerships of Inria GRA
• P. Gonçalves is member of the executive board of the Labex MILyon

11.2.1 Teaching

• Licence :
  • Thomas Begin: Computer Networks, 18.5h, L3, Université Lyon 1
  • Isabelle Guérin Lassous: Computer Networks, 15h, L3, Université Lyon 1
  • Márton Karsai: Introduction to Complex Networks, 6h, L3, ENS Lyon
  • C. Crespelle: Introduction au reseau et au Web (L2, UCBL); Algorithmique et programmation avancee (L2, UCBL); Algorithmique, programmation et complexite (L3, UCBL) ; Algorithmique Avancé (L3, ENS de Lyon)
• Master :
  • Thomas Begin: Distributed Algorithms, 18h, M1, Université Lyon 1
  • Thomas Begin: Computer Networks, 44h, M1, Université Lyon 1
  • Thomas Begin: System Administration & Security, 10.5h, M2, Université Lyon 1
  • Thomas Begin: Advanced Networks, 58h, M2, Université Lyon 1
  • Thomas Begin: Cloud Computing, 9h, M2, Université Lyon 1
  • Isabelle Guérin-Lassous: Distributed Algorithms, 30h, M1, Université Lyon 1
  • Isabelle Guérin-Lassous: Networking, 14h, M1, Université Lyon 1
  • Isabelle Guérin-Lassous: Wireless networks, 9h, M2, Université Lyon 1
  • Isabelle Guérin-Lassous: Quality of Service, 5h, M2, Université Lyon 1
  • Isabelle Guérin-Lassous: ToIP and streaming, 12h, M1, Université Lyon 1
  • Éric Guichard: Économie du web et du document, 36h, M2, Enssib & Univ. Lyon 1
  • Éric Guichard: Programmation éditoriale, 18h, M2, Enssib & Univ. Lyon 1
  • Márton Karsai: Complex Networks, 36h, M2, ENS Lyon
  • Márton Karsai: Data Bases Data Mining, 20h, M1, ENS Lyon
  • Rémi Gribonval: Inverse problems and high dimension; Mathematical foundations of deep neural networks; Concentration of measure in probability and high-dimensional statistical learning; M2, ENS Lyon
  • C. Crespelle: Optimisation et Recherche Opérationnelle (M1, UCBL)
• Engineer cycle (Bac+3 to Bac+5):
  • Paulo Gonçalves: Traitement du Signal (déterministe, aléatoire, numérique), Estimation statistique. 80 heures Eq. TD. CPE Lyon, France
  • Marion Foare: Traitement du Signal (déterministe, aléatoire), Traitement d'images, Compression, Projets. 115 heures Eq. TD. CPE Lyon, France
  • Elisa Riccietti: M2 course Optimization and Approximation (28h) and 19h of tutor responsibility at ENS Lyon
• Responsabilities
  • Thomas Begin : Head of the Master 2 speciality SRIV (Systèmes, Réseaux et Infrastructures Virtuelles) at UCBL - http://master-info.univ-lyon1.fr/SRIV
  • Isabelle Guérin Lassous has been appointed as the president of the CAPES NSI (Numérique et Science Informatique) committee since November 2019. This committee is in charge of the recruitment of the computer science teachers in high schools
  • Anthony Busson is head of the computer science department at IUT Lyon 1 Doua

11.2.2 Supervision

• PhD defense: Jacobo Levy-Abitbol Détection computationnelle des inégalités socioéconomiques, jan. 9 2020. M. Karsai (Dir.)
• PhD defense: Samuel Unicomb Threshold driven contagion on complex networks, jan. 14 2020. M. Karsai (Dir.)
• PhD defense: Gaetan Frusque Inférence et décomposition modale de réseaux dynamiques en neurosciences, December 7, 2020. P. Gonçalves (Dir.)
• PhD defense : Rémy Grünblatt, Controlled mobility for UAV networks, jan. 8, 2021. I. Guérin Lassous (Dir.) and O. Simonin
• PhD in progress: Mohamed Adbelwedoud Lafdal, Inference of conflict graph in IEEE 802.11 networks. A. Busson (Dir.) and I. Guérin Lassous. Started sept. 2017
• PhD in progress: Sicheng Dai, Dynamic Multilayer Network Modelling, M. Karsai (Dir.). Started oct. 1st, 2017
• PhD in progress: Amélie Barbe, From local to global learning, P. Gonçalves (Dir., with M. Sebban and P. Borgnat). Started oct. 1st, 2018
• PhD in progress: Nour el Houda Bouzouita, Wi-Fi network Optimization through crowd sensing applications. A. Busson (Dir., with Hervé Rivano). Nov. 2018 - Nov. 2021
• PhD in progress: Quoc Tung Le, Towards robust adversarial learning with sparse deep networks . R. Gribonval (Dir., co-supervised with E. Ricietti). Started sept. 2020
• PhD in progress: Clément Lalanne, The tradeoffs of learning with privacy. R. Gribonval (Dir., with A. Garivier). Started sept. 2020
• PhD in progress: Pierre Stock, Deep neural networks for large-scale machine learning . R. Gribonval (Dir., with H. Jegou from Facebook AI Research). Expected defence march 2021
• PhD in progress: Anthony Bardou, Machine Learning for the spatial reuse of Wi-Fi networks. T. Begin (Dir.). Started oct. 2020
• PhD in progress: Samir Si-Mohammed, Optimization of IoT Networks. T. Begin (Dir.) and I. Guérin-Lassous. Started Sept. 2020
• PhD in progress: Israel Campero Jurado, Temporal network reduction for modelling social contagion processes, P. Gonçalves (Dir.) and M. Karsai and R. Gribonval. Started nov. 1st, 2020
• PhD in progress: Alexandre Bonnefond, Dynamic flocking based on link quality in swarm of UAVs, I. Guérin Lassous and O. Simonin (Dir.). Started Oct. 2019

11.2.3 Juries

• Paulo Gonçalves was reviewer of the Ph.D thesis examination board of Barbara Pilastre, Univ. de Toulouse, nov. 2020
• Paulo Gonçalves was reviewer of the Ph.D thesis examination board of Léo Gautheron, Univ. de Saint-Etienne, dec. 2020
• Isabelle Guérin Lassous was a reviewer of the PhD thesis examination board of Imane Oussakel, Univ. de Toulouse, July 2020
• Isabelle Guérin Lassous was a reviewer of the PhD thesis examination board of Lina Aliouat, Univ. Bourgogne Franche-Comté, Nov. 2020
• Isabelle Guérin Lassous was the president of the PhD thesis examination board of Baptiste Jonglez, Univ. Grenoble Alpes, Oct. 2020
• Isabelle Guérin Lassous was a member of the PhD thesis examination board of Robin Pelle, Univ. Paris-Saclay, Dec. 2020
• Rémi Gribonval was the president of the HDR jury of Vincent Gripon, ENS de Lyon, dec. 2020, and an external examiner for the HDR defense of Yannis Avrithis, Univ. Rennes 1, july 2020
• Rémi Gribonval was an reviewer for the Ph.D. thesis of Alia Abbara, ENS Paris, nov. 2020 and the Ph.D. thesis of Hermina Petric Maretic, EPFL, oct. 2020
• Rémi Gribonval was the president of the Ph.D. defense of Ayoub Belhadji, Ecole Centrale Lille, nov. 2020
• Rémi Gribonval was an external examiner for the Ph.D. thesis of Luc Giffon, Univ. Aix-Marseille, dec. 2020
• C. Crespelle was reviewer of the Ph.D thesis examination board of Nicolas Martin, Université Grenoble Alpes, Automatique, feb. 2020
• C. Crespelle was reviewer of the Ph.D thesis examination board of Léo Rannou, Sorbonne Université, Informatique, dec. 2020

11.3.1 Articles and contents

• Isabelle Guérin Lassous is the author of the article "Enfin un CAPES en informatique !", published in the newsletter of the CNRS INS2I. https://ins2i.cnrs.fr/fr/cnrsinfo/enfin-un-capes-en-informatique
• The portrait of Isabelle Guérin Lassous belongs to the exposition of "La science taille XX elles". https://www.cnrs.fr/en/node/5282

International journals

• 10 article LafdalL. Abdelwedoud, AnthonyA. Busson, IsabelleI. Guérin-Lassous, MarionM. Foare, Mohammed L.M. Diakite and MohamedadeM. Farouk Nanne. Inference of a clear channel assessment based conflict graph Internet Technology Letters September 2020
  HAL DOI back to text
• 11 articleJacob LevyJ. Abitbol and MàrtonM. Karsai. Interpretable socioeconomic status inference from aerial imagery through urban patternsNature Machine Intelligence211November 2020, 684-692
  HALDOIback to text
• 12 articlePatriceP. Abry, NellyN. Pustelnik, Stéphane G.S. Roux, PabloP. Jensen, PatrickP. Flandrin, RémiR. Gribonval, Charles-GérardC.-G. Lucas, ÉricÉ. Guichard, PierreP. Borgnat and NicolasN. B. Garnier. Spatial and temporal regularization to estimate COVID-19 reproduction number R(t): Promoting piecewise smoothness via convex optimizationPLoS ONE158August 2020, e0237901
  HALDOIback to text
• 13 articleJuan PabloJ. Astudillo León, ThomasT. Begin, AnthonyA. Busson and Luis JavierL. de la Cruz Llopis. A Fair and Distributed Congestion Control Mechanism for Smart Grid Neighborhood Area NetworksAd Hoc Networks1042020, 1-13
  HALDOIback to text
• 14 articleThomasT. Begin, AnthonyA. Busson, IsabelleI. Guérin-Lassous and AzzedineA. Boukerche. Delivering Video-on-Demand services with IEEE 802.11p to major non-urban roads: A stochastic performance analysisComputer Networks182December 2020, 1-12
  HALDOIback to text
• 15 articleJean-ClaudeJ.-C. Bermond, DorianD. Mazauric, VishalV. Misra and PhilippeP. Nain. Distributed Link Scheduling in Wireless NetworksDiscrete Mathematics, Algorithms and Applications1252020, 1-38
  HALDOI
• 16 article AntoineA. Chatalic, VincentV. Schellekens, FlorimondF. Houssiau, Yves-AlexandreY.-A. De Montjoye, LaurentL. Jacques and RémiR. Gribonval. Compressive Learning with Privacy Guarantees Information and Inference 2021
  HAL back to text
• 17 article SichengS. Dai, HélèneH. Bouchet, AurélieA. Nardy, EricE. Fleury, Jean-PierreJ.-P. Chevrot and MàrtonM. Karsai. Temporal social network reconstruction using wireless proximity sensors: model selection and consequences EPJ Data Science 9 1 December 2020
  HAL DOI back to text
• 18 articleSimonS. Foucart, RémiR. Gribonval, LaurentL. Jacques and HolgerH. Rauhut. Jointly Low-Rank and Bisparse Recovery: Questions and Partial AnswersAnalysis and Applications18012020, 25--48
  HALDOI
• 19 articleGaëtanG. Frusque, PierreP. Borgnat, PauloP. Gonçalves and JulienJ. Jung. Semi-automatic extraction of functional dynamic networks describing patient's epileptic seizuresFrontiers in Neurology2020, 1-24
  HALDOIback to text
• 20 articleGaetanG. Frusque, JulienJ. Jung, PierreP. Borgnat and PauloP. Gonçalves. Multiplex network inference with sparse tensor decomposition for functional connectivityIEEE transactions on Signal and Information Processing over Networks62020, 316 - 328
  HALDOIback to text
• 21 article ClémentC. Gaultier, SrđanS. Kitić, RémiR. Gribonval and NancyN. Bertin. Sparsity-based audio declipping methods: selected overview, new algorithms, and large-scale evaluation IEEE/ACM Transactions on Audio, Speech and Language Processing 2021
  HAL DOI back to text
• 22 article RémiR. Gribonval, GillesG. Blanchard, NicolasN. Keriven and YannY. Traonmilin. Compressive Statistical Learning with Random Feature Moments Mathematical Statistics and Learning 2021
  HAL back to text
• 23 article RémiR. Gribonval, GittaG. Kutyniok, MortenM. Nielsen and FelixF. Voigtlaender. Approximation spaces of deep neural networks Constructive Approximation 2020
  HAL back to text
• 24 articleRémiR. Gribonval and MilaM. Nikolova. A characterization of proximity operatorsJournal of Mathematical Imaging and Vision62July 2020, 773–789
  HALDOI
• 25 article SébastienS. Lerique, Jacob LevyJ. Abitbol and MártonM. Karsai. Joint embedding of structure and features via graph convolutional networks Applied Network Science 5 1 January 2020
  HAL DOI
• 26 articlePhilippeP. Nain, GayaneG. Vardoyan, SaikatS. Guha and DonD. Towsley. On the Analysis of a Multipartite Entanglement Distribution SwitchProceedings of the ACM on Measurement and Analysis of Computing Systems June 2020, 1-38
  HALDOIback to textback to text
• 27 article Nitish KN. Panigrahy, PrithwishP. Basu, PhilippeP. Nain, DonD. Towsley, AnanthramA. Swami, Kevin SK. Chan and Kin KK. Leung. Resource Allocation in One-dimensional Distributed Service Networks with Applications Performance Evaluation 142 September 2020
  HAL DOI back to text back to text
• 28 article SamuelS. Unicomb, GerardoG. Iñiguez, JamesJ. Gleeson and MàrtonM. Karsai. Dynamics of cascades on burstiness-controlled temporal networks Nature Communications 12 1 December 2021
  HAL DOI back to text
• 29 article GayaneG. Vardoyan, SaikatS. Guha, PhilippeP. Nain and DonD. Towsley. On the Exact Analysis of an Idealized Quantum Switch Performance Evaluation 144 December 2020
  HAL back to text

International peer-reviewed conferences

• 30 inproceedings AmélieA. Barbe, MarcM. Sebban, PauloP. Gonçalves, PierreP. Borgnat and RémiR. Gribonval. Graph Diffusion Wasserstein Distances European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases Ghent, Belgium September 2020
  HAL back to text
• 31 inproceedings AmélieA. Barbe, MarcM. Sebban, PauloP. Gonçalves, PierreP. Borgnat and RémiR. Gribonval. Transport Optimal entre Graphes exploitant la Diffusion de la Chaleur Conférence sur l'Apprentissage Automatique Vannes, France 2020
  HAL back to text
• 32 inproceedings DiegoD. Di Carlo, ClémentC. Elvira, AntoineA. Deleforge, NancyN. Bertin and RémiR. Gribonval. BLASTER: An Off-Grid Method for Blind and Regularized Acoustic Echoes Retrieval -- with supplementary material ICASSP 2020 - IEEE International Conference on Acoustic Speech and Signal Processing Barcelona, Spain Inria February 2020
  HAL
• 33 inproceedingsGaëtanG. Frusque, PierreP. Borgnat, PauloP. Gonçalves and JulienJ. Jung. Extraction of functional dynamic networks describing patient's epileptic seizuresCCS 2020 - Conference on Complex SystemsPalma de Majorque, SpainDecember 2020, 1
  HALback to text
• 34 inproceedingsGaetanG. Frusque, JulienJ. Jung, PierreP. Borgnat and PauloP. Gonçalves. Regularized Partial Phase Synchrony Index Applied to Dynamical Functional Connectivity EstimationICASSP 2020 - IEEE International Conference on Acoustic Speech and Signal ProcessingBarcelona, SpainMay 2020, 1-5
  HALback to text
• 35 inproceedings RémyR. Grünblatt, IsabelleI. Guérin-Lassous and OlivierO. Simonin. Leveraging Antenna Orientation to Optimize Network Performance of Fleets of UAVs MSWiM'20 - The 23rd International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems Alicante, Spain http://mswimconf.com/2020/ November 2020
  HAL DOI back to text
• 36 inproceedingsSidharthS. Gupta, RémiR. Gribonval, LaurentL. Daudet and IvanI. Dokmanić. Fast Optical System Identification by Numerical InterferometryICASSP 2020 - IEEE International Conference on Acoustic Speech and Signal ProcessingBarcelone, SpainMay 2020, 1-5
  HAL
• 37 inproceedings Quoc-TungQ.-T. Le and RémiR. Gribonval. Structured Support Exploration For Multilayer Sparse Matrix Factorization ICASSP 2021 - IEEE International Conference on Acoustics, Speech and Signal Processing Toronto, Ontario, Canada https://2021.ieeeicassp.org/ June 2021
  HAL back to text
• 38 inproceedingsNitish KN. Panigrahy, PhilippeP. Nain, GiovanniG. Neglia and DonD. Towsley. A New Upper Bound on Cache Hit Probability for Non-anticipative Caching PoliciesPerformance 2020 - 38th International Symposium on Computer Performance, Modeling, Measurements and EvaluationMilan / Virtual, ItalyNovember 2020, 1-6
  HALback to text
• 39 inproceedingsGayaneG. Vardoyan, SaikatS. Guha, PhilippeP. Nain and DonD. Towsley. On the Capacity Region of Bipartite and Tripartite Entanglement SwitchingPerformance 2020 - 38th IFIP International Symposium on Computer Performance, Modeling, Measurements and EvaluationMilan, ItalyNovember 2020, 1-6
  HALback to text

Conferences without proceedings

• 40 inproceedingsNour el houdaN. Bouzouita, AnthonyA. Busson and HervéH. Rivano. Analytical study of frame aggregation level to infer IEEE 802.11 network loadIWCMC 2020 - 16th International Wireless Communications and Mobile ComputingLimassol, CyprusJune 2020, 952-957
  HALDOI
• 41 inproceedings Nour El HoudaN. Bouzouita, AnthonyA. Busson and HerveH. Rivano. Etude du niveau d'agrégation des trames dans les réseaux IEEE 802.11 pour l'évaluation du niveau de charge CORES 2020 – 5ème Rencontres Francophones sur la Conception de Protocoles, l’Évaluation de Performance et l’Expérimentation des Réseaux de Communication Lyon, France September 2020
  HAL
• 42 inproceedings AmelA. Chadda, MarijaM. Stojanova, ThomasT. Begin, AnthonyA. Busson and Isabelle GuérinI. Lassous. Towards a fast and efficient strategy to assign channels in WLANs with channel bonding ACM PE-WASUN Alicante, Spain November 2020
  HAL DOI back to text
• 43 inproceedingsClémentC. Elvira, Jérémy EJ. Cohen, CédricC. Herzet and RémiR. Gribonval. Continuous dictionaries meet low-rank tensor approximationsiTwist 2020 - International Traveling Workshop on Interactions between low-complexity data models and Sensing TechniquesNantes, FranceJune 2020, 1-3
  HAL
• 44 inproceedingsKilianK. Fatras, YounesY. Zine, RémiR. Flamary, RémiR. Gribonval and NicolasN. Courty. Learning with minibatch Wasserstein : asymptotic and gradient propertiesAISTATS 2020 - 23nd International Conference on Artificial Intelligence and Statisticsvolume 108PMLRPalermo, Italyhttps://www.aistats.org/June 2020, 1-20
  HALback to text
• 45 inproceedingsPierreP. Stock, ArmandA. Joulin, RémiR. Gribonval, BenjaminB. Graham and HervéH. Jégou. And the Bit Goes Down: Revisiting the Quantization of Neural NetworksICLR 2020 - Eighth International Conference on Learning RepresentationsAddis-Abeba, EthiopiaFebruary 2020, 1-11
  HALback to text

Doctoral dissertations and habilitation theses

• 46 thesis GaëtanG. Frusque. Inference and Modal Decomposition of Dynamic Networks in Neuroscience Université de Lyon December 2020
  HAL back to text
• 47 thesis RémyR. Grünblatt. From WiFi Performance Evaluation to Controlled Mobility in Drone Networks Université Claude Bernard Lyon 1 January 2021
  HAL
• 48 thesis JacoboJ. Levy Abitbol. Computational detection of socioeconomic inequalities Université de Lyon January 2020
  HAL
• 49 thesis Samuel LeeS. Unicomb. Contagion à effet de seuil dans les réseaux complexes Université de Lyon January 2020
  HAL

Reports & preprints

• 50 misc AngelaA. Fan, PierreP. Stock, BenjaminB. Graham, EdouardE. Grave, RémiR. Gribonval, HerveH. Jegou and ArmandA. Joulin. Training with Quantization Noise for Extreme Model Compression February 2021
  HAL back to text
• 51 report Jean-MarieJ.-M. Gorce, MalcolmM. Egan and RémiR. Gribonval. An efficient algorithm to estimate Covid-19 infectiousness risk from BLE-RSSI measurements Inria Grenoble Rhône-Alpes May 2020
  HAL back to text
• 52 misc RémiR. Gribonval, GillesG. Blanchard, NicolasN. Keriven and YannY. Traonmilin. Statistical Learning Guarantees for Compressive Clustering and Compressive Mixture Modeling April 2020
  HAL back to text
• 53 misc RémiR. Gribonval, AntoineA. Chatalic, NicolasN. Keriven, VincentV. Schellekens, LaurentL. Jacques and PhilipP. Schniter. Sketching Datasets for Large-Scale Learning (long version) January 2021
  HAL back to text
• 54 misc ÉricÉ. Guichard. Graphiques Covid-19 avec leurs commentaires quotidiennement mis à jour: Approche graphique, épistémologique et méthodologique April 2020
  HAL DOI
• 55 misc PhilippeP. Nain, Nitish KN. Panigrahy, PrithwishP. Basu and DonD. Towsley. One-dimensional Service Networks and Batch Service Queues December 2020
  HAL back to text back to text

Other scientific publications

• 56 thesisQuoc-TungQ.-T. Le. Multilayer Sparse Matrix FactorizationENS Paris SaclayLyonSeptember 2020, 42
  HALback to text