Document Actions
Seminar by R. Bailly, tue june 5th, on spectral methods
Title A spectral learning framework for graphical models
Speaker Raphaël Bailly, Universtité Marseille
(http://www.cmi.univ-mrs.fr/~rbailly/)
Time Tue, june 5th, 2012, 11:00h
Place Campus Nord UPC, Omega-S208
Abstract:
I will present an extended version of the spectral algorithm which can be
applied to graphs. This algorithm can be used as a learning algorithm for
graphical models - directed and undirected. It can also be used in a
density estimation task on a distribution of labelled graphs. This algorithm
is proven to converge, and is not prone to local extrema.
Speaker Raphaël Bailly, Universtité Marseille
(http://www.cmi.univ-mrs.fr/~rbailly/)
Time Tue, june 5th, 2012, 11:00h
Place Campus Nord UPC, Omega-S208
Abstract:
I will present an extended version of the spectral algorithm which can be
applied to graphs. This algorithm can be used as a learning algorithm for
graphical models - directed and undirected. It can also be used in a
density estimation task on a distribution of labelled graphs. This algorithm
is proven to converge, and is not prone to local extrema.
