Greetings, As suggested by Yoshua, we're going to have a run up to NIPS tea talk this Thursday. The idea is we give very brief pitches for the various projects underway with an eye to a NIPS submission. Students not on a project can shop around and get involved. List of speakers / subjects: (YOU MIGHT BE ON THE LIST!) - Ian Goodfellow: new estimation criteria for energy based models and a new reconstruction semi-restricted Boltzmann machine - Guillaume Desjardins: tracking the partition function and bilinear RBMs - Xavier Glorot & Yann Dauphin: reconstruction sampling with sparse connections - Xavier Glorot & Antoine Bordes: word sense disambiguation with semantic triplet models - James Bergstra: I'd like to essentially scale up the mu-ssRBM to work on big images using pyramidal image representations. - Remi Bardenet & James Bergstra: adaptive model selection - François Savard: deep auto-encoders with (recurrent) top-down connections for coordinating internal representations - Salah Rifai: semantic hashing with contrastive auto-encoders - David Warde-Farley & Razvan Pascanu: sparse coding for animations compression - Eric Thibodeau-Laufer: hybrid non-parametric & parametric (embeddings-based) content-based collaborative filtering - Razvan Pascanu: long-term dependencies in recurrent networks Time / Date: 15h00 / Thursday, May 5th, 2011 Location: LISA Lab (AA3256) 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