SEMINARS

The week's events

Monday Tuesday Wednesday Thursday Friday Saturday Sunday
March 9, 2020
March 10, 2020
March 11, 2020
March 12, 2020(1 event)

Seminar Anastasios Giovanidis

–
March 12, 2020

Title : Ranking Online Social Users by their Influence

Abstract:

In this talk I will introduce an original mathematical model to analyse the diffusion of posts within a generic online social platform. As a main result, using the developed model, we derive in closed form the probabilities that posts originating from a given user are found on the Wall and Newsfeed of any other inside the platform. By combining these probabilities we get a measure of per user influence on the entire network. This constitutes a new centrality measure which is more expressive than existing ones, in the sense that it combines the user position on the graph with the user posting activity. Comparisons with simulations show the accuracy of this model and its robustness with respect to the modelling assumptions. Furthermore, its application on large data traces from real platforms asserts its validity for real world applications, and the possibilities it opens for explaining real diffusion phenomena and predicting actual user influence.

Bio:

Anastasios Giovanidis received the Diploma degree in electrical and computer engineering from the National Technical University of Athens, Greece, in 2005, and the Dr. Ing. degree in wireless communications and information theory from the Technical University of Berlin, Germany, in 2010. He has been a postdoctoral fellow, first with the Zuse Institute Berlin, Germany (with Prof. Martin Grötschel), and later with INRIA, Paris, France (with Prof. François Baccelli). Since 2013 he is a permanent researcher of the French National Center for Scientific Research (CNRS, CR1). From 2013 until 2016 he was affiliated with the Télécom ParisTech CNRS-LTCI laboratory. Since 2016 he is affiliated with the computer science laboratory LIP6 of the Sorbonne University. He has served as the General co-chair for WIOPT 2017, CCDWN 2018, and GameNets 2019. His current research interests include performance analysis and optimisation of telecom and social networks, supported by data analysis and learning.
March 13, 2020(1 event)

F. Falniowski: "Robust routes to chaos in congestion games: The effects of scale on learning dynamics"

–
March 13, 2020

We study the effects of increasing the population size/scale of costs in congestion games and generalize recent results for the well known Multiplicative Weights Update dynamic to a large class of Follow-the-Regularized Leader dynamics (FoReL). We prove that even in simple linear congestion games with two parallel links as the population/scale increases, learning becomes unstable and (unless the game is fully symmetric) eventually Li-Yorke chaotic. Despite their chaotic instability, the dynamics provably converge in a time-average sense to an exact equilibrium for any choice of learning rate and any scale of costs.
March 14, 2020
March 15, 2020
  • April 3, 2024 @ Bâtiment IMAG (442) -- [Seminar] Victor Boone

    Who: Victor Boone

    When: Wednesday, April 3, 14:00-15:00

    Where: 447

    What: Learning MDPs with Extended Bellman Operators

    More: Efficiently learning Markov Decision Processes (MDPs) is difficult. When facing an unknown environment, where is the adequate limit between repeating actions that have shown their efficiency in the past (exploitation of your knowledge) and testing alternatives that may actually be better than what you currently believe (exploration of the environment)? To bypass this dilemma, a well-known solution is the "optimism-in-face-of-uncertainty" principle: Think of the score of an action as being the largest that is statistically plausible.

    The exploration-exploitation dilemma then becomes the problem of tuning optimism. In this talk, I will explain how optimism in MDPs can be all rephrased using a single operator, embedding all the uncertainty in your environment within a single MDP. This is a story about "extended Bellman operators" and "extended MDPs", and about how one can achieve minimax optimal regret using this machinery.

  • April 11, 2024 @ Bâtiment IMAG (442) -- [Seminar] Charles Arnal

    Who: Charles Arnal

    When: Thursday, April 11, 14:00-15:00

    Where: 442

    What: Mode Estimation with Partial Feedback

    More: The combination of lightly supervised pre-training and online fine-tuning has played a key role in recent AI developments. These new learning pipelines call for new theoretical frameworks. In this paper, we formalize core aspects of weakly supervised and active learning with a simple problem: the estimation of the mode of a distribution using partial feedback. We show how entropy coding allows for optimal information acquisition from partial feedback, develop coarse sufficient statistics for mode identification, and adapt bandit algorithms to our new setting. Finally, we combine those contributions into a statistically and computationally efficient solution to our problem.

  • April 30, 2024 @ Bâtiment IMAG (442) -- Seminar Rémi Castera

    Correlation of Rankings in Matching Markets

Comments are closed.