Jan 13

PhD Seminar – Clément Lalanne’s mini lectures on Differential Privacy

Clément Lalanne, PhD at Dante and UMPA, will give a series of mini lectures on Differential Privacy.

Differential Privacy is a tractable property that aims at protecting the privacy of the individuals composing a dataset.

  • December 9th 2020 at 3:30pm : Introductory lecture on Differential Privacy. On this first lecture we will go through the basic definitions and properties of this concept while explaining why it is appealing for real world applications. We will also cover some modest examples. For the next lectures, Clément plans to cover some general techniques to turn an existing algorithm into a differentially private one and to present some advanced techniques in order to track the privacy loss through composition of algorithms.
  • December 16th 2020 at 3:45pm : We will focus on the techniques that add noise to the output of an algorithm in order to enhance privacy.
  • January 13th 2021 at 9:00am : We will study the graphical representation of differential privacy in an hypothesis testing setup and use it to deduce some properties including the advanced sharp composition theorem for Differential Privacy.


After the lectures, the slides and a detailed pdf will be available at https://clemlal.github.io/privacy.

Thank you !

Jan 11

Mathurin Massias. Fast resolution resolution of structured inverse problems: extrapolation and iterative regularization

This Thursday 14 Jan at 4.pm we will welcome Mathurin Massias post-doctoral researcher at University of Genova who will present his works on structured inverse problem:

Title: Fast resolution resolution of structured inverse problems: extrapolation and iterative regularization

Abstract: Overparametrization is common in linear inverse problems, which poses the question of stability and uniqueness of the solution. A remedy is to select a specific solution by minimizing a bias functional over all interpolating solutions. This functional is frequently neither smooth nor convex (e.g. L1, L2/L1, nuclear norm, TV). In the first part of the talk, we study fast solvers for the so called Tikhonov approach, where the bilevel optimization problem is relaxed into “datafit + regularization” (e.g., the Lasso). We show that, for separable problems arising in ML, coordinate descent algorithms can be accelerated by Anderson extrapolation, which surpasses full gradient methods and inertial acceleration.
The Tikhonov approach can be costly, as it requires to calibrate the regularization strength over a grid of values, and thus to solve many optimization problems. In the second part of the talk, we present results on iterative regularization: a single optimization problem is solved, and the number of iterations acts as the regularizer. We derive novel results on the early stopped iterates, in the case where the bias is convex but not strongly convex.

The presentation is based on:

– https://arxiv.org/abs/2011.10065 Anderson acceleration of coordinate descent
– https://arxiv.org/abs/2006.09859 Iterative regularization for convex regularizers
– https://arxiv.org/abs/1907.05830 Dual extrapolation for sparse generalized linear models

(joint works with Alexandre Gramfort, Joseph Salmon, Samuel Vaiter, Quentin Bertrand, Cesare Molinari, Lorenzo Roscasco and Silvia Villa)

Feb 11

PhD and Postdoc seminar: “Recent advances on solving the Mumford-Shah model for discrete images
” by Marion Foare

Friday 11th February, 2018 at 11am.  – LIP Meeting room M7 (3rd floor)

Our next seminar will be given by Marion Foare, who will present on “Recent advances on solving the Mumford-Shah model for discrete images
”

Abstract-“Essential image processing and analysis tasks, such as image segmentation, simplification and denoising, can be conducted in a unified way by minimizing the Mumford-Shah functional. Although seductive, this minimization is in practice difficult because it requires to jointly define a sharp set of contours and a smooth version of the initial image. For this reason, various relaxations of the original formulations have been proposed, together with optimisation methods. In this talk, we propose several discrete approximations of the Mumford-Shah and their numerical resolution for image processing tasks. We compare the results with state-of-the-art convex relaxations of the Mumford–Shah functional, and show that the proposed methods lead to competitive denoising, restoration and segmentation results.“

Oct 29

PhD Seminar & DL Journal Club: Anomalous diffusion for graph-based data classification

Friday, October 26th, 2018 at 11am.  – LIP Meeting room M7

Esteban Bautista  will present on “Anomalous diffusion for graph-based data classification”.

The talk will be a one time-merging of the PhD seminar and the DL Journal club of the DANTE team.

 

Oct 03

Márton Karsai receives the Junior Scientific Award of the Complex System Society

Since 2014, the Complex System Society gives the Junior Scientific Award to young members of the society (within ten years of completing their PhD)  to recognise their extraordinary scientific achievements within the field of complex systems/

For the year 2018, the Junior Scientific Award was given to Márton Karsai “for his many outstanding contributions to the science of complex systems, especially for his work on temporal networks and computational social sciences” at the CSS2018 annual conference, which was held between the 23 to 28 September 2018 in Thessaloniki, Greece.

 

Sep 29

Stic-AmSud MOTIf – Mo​bile phone sensing of human dynamics in ​techno-social environment

The MOTIf project, funded by the Stic-AmSud program, is a two-years long international project (2018-2019) between French and Latin American partners from Argentina and Brazil. The general goal of the MOTIf project is to ​understand, model, and predict individual behavior embedded in social and technological environments. We propose to work in two directions in order to tackle this challenge:

  • aim to understand ​spatiotemporal patterns of service usage of individuals to learn when, where, and what people are doing​.
  •  aim to understand the ​fine-grained sociodemographic structure of society and see how the ​demographic characteristics of individuals in a social network correlate with the dynamics of their egocentric and global network evolution.

Information and Communication Technology (ICT) is becoming increasingly social, as demonstrated by the multitude of emerging technologies and technology platforms that facilitate social interactions, taking place as communication via telephone, text message, email, online social networks etc. At the same time, our social activities are increasingly embedded in the ICT environments that enable and enhance our ability to transact, share experiences, and maintain social relationships.

One of the best ways to explore these developments is through the mining and analysis of data, which are collected through mobile phones and allow us to investigate how individuals act when embedded in a technology-enabled environment. Unlimited access to a wide range of mobile applications and services may change our way to gain information, to communicate, or even to behave in different contextual places like home, work, or anywhere else. Thus understanding individual activity patterns and the source of decisions behind them is moreover important for the design of future services and to estimate the demand on the infrastructure.

The MOTIf project builds on the analysis and modeling of geo-localized temporally detailed but fully anonymised mobile phone call networks. These datasets allow us to address the two scientific objectives about ​spatiotemporal patterns of service usage of anonymised individuals to learn when, where, and what people are doing; ​and about the ​fine-grained sociodemographic structure of society and its effect on the the individual social behaviour. In other words our goal in general is to understand how individuals behave in a dynamic techno-social environment.

Partners:

  • INRIA (DANTE and INFINE teams) – France
  • Grandata – Argentina
  • Universidad de Buenos Aires – Argentina
  • Universidade Federal de Minas Gerais – Brazil
  • National Laboratory for Scientific Computing – Brazil
  • Pontifícia Universidade Católica de Minas Gerais – Brazil

 

International coordinator

  • Márton KARSAI (ENS Lyon/DANTE)

 

National coordinators:

  • Jussara M. ALMEIDA (Universidade Federal de Minas Gerais)
  • Alejo SALLES (Universidad de Buenos Aires)

Results and publications

  • Y. Leo, M. Karsai, C. Sarraute, E. Fleury, Correlations and dynamics of consumption patterns in social-economic networks. Soc. Netw. Analys. Min. 8, 9 (2018)
  • Y. Liao, W. Du, M. Karsai, C. Sarraute, M. Minnoni, Eric Fleury, Novel Methods of Subscription Type Prediction in Mobile Phone Services. Lect. Notes in Soc. Netw. Analys. Min. (2018)
  • G. Chen, A. C. Viana, C. Sarraute, M. Fiore, S. Hoteit. Enriching Sparse Mobility Information in Call Detail Records.  To appear at Computer Communications Elsevier journal.2018.
  • Lima, A. Aguiar, A. C.Vianaand P. Carvalho. Impacts of Human Mobility in Data Offloading. ACM CHANTS, jointly with ACM MobiCom. New Delhi, India. October 2018.
  • G. Chen, A. C. Viana, M. Fiore. Takeaways in Large-scale Human Mobility Data Mining(invited paper). IEEE International Symposium on Local and Metropolitan Area Networks, Jun 2018, Washington, United States.
  • G. Chen, S. Hoteit, A. C. Viana,M. Fiore, C. Sarraute. Forecasting Individual Demand in Cellular Networks. Rencontres  Francophones sur la Conception de Protocoles, l’Évaluation de Performance et l’Expérimentation des Réseaux de Communication, May 2018, Roscoff, France. 2018.
  • G. Chen, A. C. Viana, M. Fiore. Human Trajectory Recovery via Mobile Network Data. Rencontres Francophones sur la Conception de Protocoles, l’Évaluation de Performance et l’Expérimentation des Réseaux de Communication, May 2018, Roscoff, France. 2018.
  • R. Costa, L. Sampaio, A. Ziviani, A. C. Viana. *Humanos no ciclo de Comunicação*. Short course of 36th edition of the Brazilian Symposium on Computer Networks and Distributed Systems (SBRC), Brazilian Computer Society (SBC). To appear as a chapter of book named “Mini-curso SBRC 2018”. May 2018. (ISBN to be defined).
  • H. T. Marques-Neto, F. H. Z. Xavier, W. Z. Xavier, C. H. S. Malab, A. Ziviani, L. M. Silveira, J. M. Almeida, Understanding  Human Mobility and Workload Dynamics Due to Different Large-Scale Events using Mobile Phone Data, Journal of Network and Systems Management, Vol 26, Issue 4, pp. 1079-1100, October 2018.
  • W. Z. Xavier,H. T. Marques-Neto, Modal — A Platform for Mobility Analyses Using Open Datasets, Workshop on Big Social Data and Urban Computing, in conjunction with VLDB’2018.
  • R. L. Costa, L. Sampaio, A. Ziviani, A. C. Viana,  Short Course in conjunction with XXXVI Brazilian Symposium on Computer Networks and Distributed Systems, 2018.

 

Invited talks and conference presentations

  • Aline C. Viana: IEEE International Symposium on Local and Metropolitan Area Networks (Jun. 2018, Washington, US). ACM CHANTS (Oct. 2018) jointly with ACM MobiCom. New Delhi, India. Rencontres Francophones sur la Conception de Protocoles, l’Évaluation de Performance et l’Expérimentation des Réseaux de Communication (May 2018), Roscoff, France. 2018. Brazilian Symposium on Computer Networks and Distributed Systems (May 2018), Campus do Jordao, Brazil.
  • Karsai– Socioeconomic Correlations and Stratification in Social Communication Networks (invited talk), Mobile Tartu 2018, International Conference on 27-29th of June 2018 in Tartu, Estonia.
  • Karsai– Socioeconomic correlations and stratification in social-communication networks (invited talk),Conference on Complex Systems CCS2018 Satellite Meeting “Complex Systems for the Most Vulnerable”, Thessaloniki, 27 September 2018
  • Z. Xavier,H. T. Marques-Neto, Modal — A Platform for Mobility Analyses Using Open Datasets, Workshop on Big Social Data and Urban Computing, in conjunction with VLDB’2018.

Sep 20

PhD and Postdoc seminar: “Conflict graph-based models for Wifi networks” by Marija Stojanova

Friday 28th September, 2018 at 2pm. Location TBA

The first Phd / Postdoc seminar of the 2018 / 2019 calendar will follow the work of DANTE’s Marija Stojanova, who will present her recent research on wireless networks. In particular, we will discuss conflict-graph based models of Wifi.

Abstract – “We describe a performance modeling method that can provide guidance for configuring WLANs and be used as a decision-support tool by a network architect or as an algorithm embedded within a WLAN controller. The proposed approach estimates the attained throughput of each AP using a Divide-and-Conquer strategy which breaks down the original problem into multiple sub-problems. Every sub-problem is solved as a Markov chain and the solutions are then combined to provide the solution to the original problem. ”

 

Sep 20

DL Journal Club: “Human-level control through deep reinforcement learning” digest by Jacobo Levy-Abitbol

Friday, September 21st, 2018 at 11am – LIP meeting room M7 3rd floor

Jacobo Levy-Abitbol will be giving a presentation of the paper “Human-level control through deep reinforcement learning” published by Mnih et al. in Nature in 2015. From the abstract:

Here we use recent advances in training deep neural networks to develop a novel artificial agent, termed a deep Q-network, that can learn successful policies directly from high-dimensional sensory inputs using end-to-end reinforcement learning. We tested this agent on the challenging domain of classic Atari 2600 games.

 

Full details on the Journal Club’s repository.

Human-level control through deep reinforcement learning, Mnih et al. (2015)

 

Jun 28

PhD and Postdoc seminar: “Graph-based semi-supervised learning for classification” by Sarah De Nigris

Thursday, June 28th, 2018 at 3pm – IXXI Conference room M7

Sarah De Nigris will present her latest work on graph-based semi-supervised learning for classification.

The talk is the second in this year’s series of monthly presentations aimed at diffusing graduate research in the DANTE team.

Jun 12

DL Journal Club: “Geometric deep learning on graphs and manifolds” digest by P. Borgnat after M. Bronstein’s talk

Friday, July 13th, 2018 at 1:30pm – IXXI Conference room M7

Pierre Borgnat will be giving a (simplified) presentation of the tutorial “Geometric deep learning on graphs and manifolds” done by Michael Bronstein at the Graph Signal Processing workshop held at EPFL, june 6–8, 2018 . We should prepare for “a mix of Convolutional Neural Networks and Graph Signal Processing”.

There is a tutorial/review article for reference, which can be good to read before the meeting: M. M. Bronstein, J. Bruna, Y. LeCun, A. Szlam, P. Vandergheynst, Geometric deep learning: going beyond Euclidean data, IEEE Signal Processing Magazine 2017.

Full details on the Journal Club’s repository.

Example Laplacian eigenfunctions on Euclidian and non-Euclidian domains used for Spectral CNN, Bronstein et al. (2017)