Random Matrices in Machine Learning
Feb 23, 2016
from 10:45 to 12:00
|Where||ENS de Lyon, salle de conférence du CBP|
|Add event to calendar||
Abstract: In this talk, we will discuss the recent theoretical findings in advanced random matrix models, allowing for a better understanding of classical machine learning algorithms. Moving from the classical "sample covariance matrix" model, where random matrices have been rooted for long, to more elaborate robust scatter estimates, kernel random matrices, graph Laplacians, and neural network models, we shall discuss their applications in particular to kernel spectral clustering, community detection on graphs, echo-state neural network performance and optimization, and so on.