Yoshua is in an overtime phone meeting with DARPA people. We will postpone the Tea-talk to 3pm. Cheers, Aaron On Thu, Feb 17, 2011 at 9:59 AM, Aaron Courville <aaron.courville@gmail.com> wrote:

We're kicking off our new (and belated) tea talk season with Ilaria Castelli who will talk about her recent work building and using massive quantities of Youtube data to learn models of faces.

Date: Today, Thursday Feb. 17th 2011 Time: 14h30 Where: LISA lab (AA3256)

Abstract:  I'll explain how the youtube dataset has been collected and how it is organized, in case someone else would like to use it in the future. I'll then speak about the experiments I have tried with denoising autoencoders and the olivetti dataset on a pose regression task, and the issues I encountered when training together with our youtube data.

Hope to see you there. Cheers, Aaron

-- Aaron C. Courville Département d’Informatique et de recherche opérationnelle Université de Montréal email:Aaron.Courville@gmail.com

-- Aaron C. Courville Département d’Informatique et de recherche opérationnelle Université de Montréal email:Aaron.Courville@gmail.com