[Lisa_seminaires] Talk by Surya Ganguli, this Friday 17th Jan, 14h00-15h00 AA3195

Razvan Pascanu r.pascanu at gmail.com
Mar 14 Jan 00:51:52 EST 2014


Hi all,

 This Friday we have a talk by Surya Ganguli at 14h00 in room 3195 Andre
Aisenstadt.

Exact Solutions to the Nonlinear Dynamics of Learning in Deep Linear
Networks

Despite the widespread practical success of deep learning methods, our
theoretical understanding of the dynamics of learning in deep neural
networks remains quite sparse. We attempt to bridge the gap between the
theory and practice of deep learning by systematically analyzing learning
dynamics for the restricted case of deep linear neural networks. Despite
the linearity of their input-output map, such networks have nonlinear
gradient descent dynamics on weights that change with the addition of each
new hidden layer. We show that deep linear networks exhibit nonlinear
learning phenomena similar to those seen in simulations of nonlinear
networks, including long plateaus followed by rapid transitions to lower
error solutions, and faster convergence from greedy unsupervised
pretraining initial conditions than from random initial conditions. We
provide an analytical description of these phenomena by finding new exact
solutions to the nonlinear dynamics of deep learning. Our theoretical
analysis also reveals the surprising finding that as the depth of a network
approaches infinity, learning speed remains finite: for a special class of
initial conditions on the weights, very deep networks incur only a finite
delay in learning speed relative to shallow networks. We further show that,
under certain conditions on the training data, unsupervised pretraining can
find this special class of initial conditions.  We also discuss application
to infant semantic development.

I hope to see many of you there !

Best,
Razvan
-------------- section suivante --------------
Une pièce jointe HTML a été nettoyée...
URL: https://webmail.iro.umontreal.ca/mailman/private/lisa_seminaires/attachments/20140114/3e0456e4/attachment.html 


Plus d'informations sur la liste de diffusion Lisa_seminaires