# Seminar

Since 2017 I have been organizing a regular seminar on information theory and related topics within the CITI laboratory.

TBA

## Past Seminars

Speaker: Dr Paulo Goncalves (INRIA, LIP)

Date14th November 2018.

Title: $L^{\gamma}$ Semi-Supervised Learning for Classification

Abstract: Graph-based semi-supervised learning are methods for classification that combine expert knowledge from (generally few) labelled samples with the structural organization (natural or inferred) of large datasets. PageRank is one famous algorithm that undergoes a standard random walk interpretation to explain how the a priori information diffuses over the entire graph. After extending PageRank to long-range diffusion processes, such as Levy flights, we will show how powers of the combinatorial Laplacian can improve classification performances in critical situations where standard SSL methods are known to behave poorly. Then, we will show that $L^{\gamma}$-SSL yields a pertinent   proxy for the NP-hard problem of unsupervised normalized cut.

Speaker: Dr Aline Roumy (INRIA Rennes)

Date: 12th September 2018.

TitleSource coding under massive random access: theory and applications.

Abstract: In this presentation we will introduce a novel source coding problem allowing massive random access to large databases. Indeed, we consider a database that is so large that, to be stored on a single server, the data have to be compressed efficiently, meaning that the redundancy/correlation between the data have to be exploited. The dataset is then stored on a server and made available to users that may want to access only a subset of the data. Such a request for a subset of the data is indeedrandom, since the choice of the subset is user-dependent. Finally, massive requests are made, meaning that, upon request, the server can only perform low complexity operations (such as bit extraction but no decompression/compression). After describing the problem, information theoretical bounds of the source coding problem will be derived. Then two applications will be presented: Free-viewpoint Television (FTV) and massive requests to a database collecting data from a large-scale sensor network (such as Smart Cities).

Speaker: Dr Anne Savard (IMT Lille Douai)

Date: July 9th, at 10h30, TD-D (Claude Chappe Building)

Title: IF Neuron: theoretical study and application to digital communication

Abstract: In the context of digital communication, one main mechanism proposed in the literature to overcome the large consumption of MAC layers when establishing communications is called wake-up radio: The main processor is only waking up when receiving a specific signal, as for instance the node ID in the network. Unfortunately, since most of the wake-up receivers rely on standard micro-controller, they suffer a large decrease of energy efficiency. Nevertheless, if the wake-up receivers was designed with neuromorphic circuits, one could achieve high energy efficiency for IoT and ad hoc networks.

The main question that is tackled in this presentation is whether a neuro-inspired detection scheme using an Integrate-and-Fire neuron is reliable enough when one needs to detect a weak signal surrounded by noise.

Speaker: Giulia Cervia (ETIS, ENSEA, Université Paris Seine)

Date: 4th July, 14h, TD-D Building Chappe, INSA Lyon.

Title: Strong coordination of signals and actions over noisy channels with two-sided state information

Abstract: In decentralized networks, communication devices must be able to cooperate, to take decisions in a distributed fashion and to reconfigure dynamically by reacting to changes in the environment. To achieve such behavior, efficient techniques to coordinate the actions of different nodes must be developed.
In this talk, we consider a two-node network with a noisy channel and two-sided state information, in which the input and output signals have to be coordinated with the source and its reconstruction.
We propose a joint source-channel coding scheme and derive inner and outer bounds for the strong coordination region. Moreover, we are able to give a complete characterization of the coordination region in some particular cases. Finally, we show that polar codes achieve the best known inner bound for the strong coordination region.

Speaker: Dr Vyacheslav Kungurtsev (Czech Technical University in Prague)

Date: 7th May 2018, 14h00 in TDC.

Title: Optimization Algorithms for Solving Problems Arising from Large Scale Machine Learning

Abstract: In the contemporary “big data” age, the use of Machine Learning models for analyzing large volumes of data has been instrumental in a lot of current technological development. These models necessitate solving very large scale optimization problems, presenting challenges in terms of developing appropriate solvers. In addition, especially for problems arising from Deep Neural Network architectures, the resulting problems are often nonconvex, and sometimes nonsmooth, giving additional difficulty. In this talk I present the standard structural elements of this class of problems, and how these structures can be handled with appropriate parallel architectures. I discuss the state of the art in terms of optimization algorithms for this setting and summarize the prognosis for ongoing and future research.

Speaker: Dr. Nourddine Azzaoui (Laboratoire de Mathématiques, Université Clermont Auvergne)

Date: 16th March 2018, 14h00 in TD-C.

Title: Heavy tailed distributions characterisations and examples of applications in channel modeling.

Abstract: Currently, we are witnessing the proliferation of wireless sensor networks and the superposition of several communicating objects which have an heterogeneous nature. The advent of Internet of Things networks as well as the increasing demand for improved quality and services will increase the complexity of communications and puts a strain on current techniques and models. Indeed, they must firstly adapt to the temporal and spatial evolution and secondly, they must take into account the rare and unpredictable events that can have disastrous consequences for decision-making. This talk provides an overview of the various spectral techniques used in litterature describe a communication channel having an impulsive behavior. This is mainly motivated by the historical success of interactions between probabilities, statistics and the world of communications, information theory and signal processing. The presentation will be divided into two parts: the first is devoted to the synthesis of various developments on alpha-stable variables and processes in a purely mathematical mind. The second part will be devoted to applications in the context of communications. The two sides will combine two fundamentally linked aspects: rst, a theoretical approach, necessary for a good formalization of problems and identifying the best solutions. Secondly, the use of these models in real work of channel modelling.