January 9, 2020
Keynote: Radu Horaud
Keynote: Radu Horaud
–
January 9, 2020
-
News
- Journée au vert POLARIS 2022/05/23
- DATAMOVE/POLARIS picnic 2021/06/22
- DATAMOVE/POLARIS BBQ 2019 2019/06/14
- POLARIS Bootcamp (May 2019) 2019/05/24
- slides of Andras Gyorgy 2016/01/15
Next seminars
- 2:00 pm – 3:00 pm, April 30, 2024 – Seminar Rémi Castera
Events
Events in January–February 2020
MMonday TTuesday WWednesday TThursday FFriday SSaturday SSunday 30December 30, 201931December 31, 2019January
1January 1, 20202January 2, 20203January 3, 20204January 4, 20205January 5, 20206January 6, 20207January 7, 20208January 8, 2020Keynote: Radu Horaud
–
January 9, 202010January 10, 202011January 11, 202012January 12, 202013January 13, 202014January 14, 202015January 15, 2020Capacity of a LoRaWAN cell, by Martin Heusse (Drakkar)
–
January 16, 2020We propose a model to estimate the packet delivery rate in a LoRaWAN cell, when all nodes have the same traffic generation process and may use repetitions. The model predicts the transmission success rate for any cell range and node density, with similar traffic from all nodes. We find that the transmission success depends on striking a balance between the adverse effects of attenuation and collisions; in small cells, it is highly dependent on the suitable allocation of spreading factors, whereas using packet repetitions is more effective in large cells.
Bâtiment IMAG (442)17January 17, 202018January 18, 202019January 19, 202020January 20, 202021January 21, 202022January 22, 2020Can random matrices change the future of machine learning? by Romain Couillet (Gipsa)
–
January 23, 2020Romain COUILLET (professor at CentraleSupélec, University ParisSaclay; IDEX GSTATS Chair & MIAI LargeDATA Chair, University Grenoble-Alpes)
Title: Can random matrices change the future of machine learning?
Abstract: Many standard machine learning algorithms and intuitions are known to misbehave, if not dramatically collapse, when operated on large dimensional data. In this talk, we will show that large dimensional statistics, and particularly random matrix theory, not only can elucidate this behavior but provides a new set of tools to understand and (sometimes drastically) improve machine learning algorithms. Besides, we will show that our various theoretical findings are provably applicable to very realistic and not-so-large dimensional data.24January 24, 202025January 25, 202026January 26, 202027January 27, 202028January 28, 202029January 29, 2020HDR Nicolas Gast
–
January 30, 2020Abstract:
The design of efficient algorithms is closely linked to the evaluation of their performances. My work focuses on the use of stochastic models for the performance evaluation of large distributed systems. I am interested in developing tools that can characterize the emergent behavior of such systems and improve their performance. This leads me to solve stochastic control and optimization problems, notably through operations research methods. These problems suffer from combinatorial explosion: the complexity of a problem grows exponentially with the number of objects that compose it. It is therefore necessary to design models but also algorithmic processes whose complexity does not increase too rapidly with the size of the system.In this presentation, I will summarize a few of my contributions on the use and the refinements of mean field approxiamtion to study the performance of distributed systems and algorithms. I will introduce the key concepts behind mean field approximation, by giving some examples of where it can be applied. I will review some of the classical models and try to answer a very natural question: how large should a system be for mean field to apply?
Bâtiment IMAG (amphitheater)Saint-Martin-d'Hères, 38400France31January 31, 2020February
1February 1, 20202February 2, 20203February 3, 20204February 4, 20205February 5, 2020keynote LIG
–
February 6, 20207February 7, 20208February 8, 20209February 9, 202010February 10, 2020Distributed Variational Representation Learning, by Abdellatif Zaidi (Huawei)
–
February 11, 2020We connect the information flow in a neural network to sufficient statistics; and show how techniques that are rooted in information theory, such as the source-coding based information bottleneck method can lead to improved architectures, as well as a better understanding of the theoretical foundation of neural networks, viewed as a cascade compression network. We discuss distributed architectures and illustrate our results and view through some numerical examples.
44212February 12, 2020Seminar Francois Durand: "Analysis of Approval Voting in Poisson Games"
–
February 13, 2020Titre :Analysis of Approval Voting in Poisson GamesRésumé :Poisson games were introduced by Roger Myerson to model large elections. In this framework, the number of players of each type follows a Poisson distribution. Typically, the expected number of players is considered large, and players base their strategic response on rare events, such as a "pivot" where two candidates are tied for victory, and where one vote can make the difference. The main part of the talk will be dedicated to presenting Poisson games and the main theoretical results about them. Then I will say a few words about our work on a voting rule called Approval voting.
References:
R. Myerson (2000). Large Poisson games. Journal of Economic Theory, 94, 7–45.
R. Myerson (2002). Comparison of scoring rules in Poisson voting games. Journal of Economic Theory, 103, 219–251.
F. Durand, A. Macé and M. Núñez (2019). Analysis of Approval Voting in Poisson Games. ACM Conference on Economics and Computation.Python implementation: https://poisson-approval.readthedocs.io/.
14February 14, 202015February 15, 202016February 16, 2020Immanuel Bomze (univ. Vienna)
–
February 17, 2020Robust clustering in social networks
(joint with M. Kahr and M. Leitner)
During the last decades the importance of considering data uncertainty in optimization problems has become increasingly apparent, since small
fluctuations of input data may lead to comparably bad decisions in many
practical problems when uncertainty is ignored. If the probability
distribution of the uncertain data is not known (or cannot be sufficiently estimated), a common technique is to estimate bounds on the uncertain data (i.e., define uncertainty sets) and to identify optimal solutions that are robust against data fluctuations within these bounds. This approach leads to the robust optimization paradigm that allows to consider uncertain objectives and constraints [1].Optimization problems where only the objective is uncertain arise, for instance, prominently in the analysis of social networks.
This stems from the fact that the strength of social ties (i.e., the amount of influence individuals exert on each
other) or the willingness of individuals to adopt and share information can, for example, only be roughly estimated based on observations.
A fundamental problem arising in social network analysis regards the identification of communities (e.g., work groups, interest groups), which can be modeled as a Dominant Set Clustering Problem [5,6,7] which in turn leads to a Standard Quadratic Optimization Problem (StQP); see [2]. Here the link strengths enter the objective while the constraints are familiar probability constraints, so that they can be considered certain.Hence we investigate data uncertainty in the objective function of StQPs, considering different uncertainty sets, and derive implications for the complexity of robust variants of the corresponding deterministic counterparts. We can show that considering data uncertainty in a StQP results in another StQP of the same complexity if ellipsoidal, spherical or boxed uncertainty sets are assumed [4]. Moreover we discuss implications when considering polyhedral uncertainty sets, and derive rigorous bounds for this case, based upon copositive optimization [3].
References
[1] Ben-Tal A, El Ghaoui L, Nemirovski AS (2009) Robust optimization. Princeton Series in Applied Mathematics (Princeton NJ: Princeton University Press).
[2] Bomze IM (1998) On standard quadratic optimization problems. Journal of Global Optimization 13(4):369–387.
[3] Bomze IM (2012) Copositive optimization – Recent developments and applications. European Journal of Operational Research 216(3):509–520.
[4] Bomze IM, Kahr M, Leitner M. (2020) Trust your data or not - StQP remains StQP: Community Detection via Robust Standard Quadratic Optimization. To appear in Mathematics of OR.
[5] Pavan M, Pelillo M (2007) Dominant sets and pairwise clustering. IEEE Transactions on Pattern Analysis and Machine Intelligence 29(1):167–172.
[6] Rota Bulò S, Pelillo M (2017) Dominant-set clustering: A review. European Journal of Operational Research 262(1):1–13.
[7] Rota Bul\`{o} S, Pelillo M, Bomze IM (2011) Graph-based quadratic optimization: A fast evolutionary approach. Computer Vision and Image Understanding 115(7):984–995.
Bâtiment IMAG (442)18February 18, 202019February 19, 2020Seminar Abhijnan Chakraborty
–
February 20, 202021February 21, 202022February 22, 202023February 23, 202024February 24, 202025February 25, 202026February 26, 2020Strategic information transmission with receiver's type-dependent decision sets, by Stephan Sémirat (GAEL, Grenoble)
–
February 27, 2020Strategic information transmission with receiver's type-dependent decision sets.
Abstract:
We consider a sender-receiver game, in which the sender has finitely many types and the receiver makes decisions in a bounded real interval. We assume that utility functions are concave, single-peaked and supermodular. After the cheap talk phase, the receiver makes a decision, which must fulfill a constraint (e.g., a participation constraint) that depends on the sender's type. Hence a necessary equilibrium condition is that the receiver maximizes his expected utility subject to the constraints of all positive probability types. This necessary condition may not hold at the receiver's prior belief, so that a non-revealing equilibrium may fail to exist. We propose a constructive algorithm that always achieves a partitional perfect Bayesian equilibrium
Bâtiment IMAG (442)28February 28, 202029February 29, 2020March
1March 1, 20202March 2, 20203March 3, 20204March 4, 2020keynote LIG
–
March 5, 20206March 6, 20207March 7, 20208March 8, 2020Meta
