Corelab Seminar
2014-2015
Emmanouil Vlatakis-Gkaragkounis (NTUA)
Structured Cover for Multi-Parametrized Distribution
Sparse Cover for Sum of Independent Indicators
Abstract.
Much research has been done in the last decade on the problem of estimation
approximated Nash equillibria in different settings. In a series of
scientific publications, Daskalakis and Papadimitriou studied this problem
at the level of anonymous games. During their studies, they are led to the
construction of a sparse cover for a specific family of distributions that
generlize the concept of the well-known Binomial Distributions. This
generalized family distribution (named Poisson Binomial Distributions), the
cover and its structure, independently from the seminal cause of their
creation, are appeared to play a dominant role in Computer Science and have
application on both fields of Algorithmic Game Theory and Computational
Statistical Learning.
Initially, our goal is to present, explain and deepen on the construction
and the properties of the structure of the cover based on this publication.
At a later stage we hope to study some of the innovating applications of the
construction focused on Distribution Learning