I would like to give a talk but I don't have a topic per se. I encountered a paper that attempts to do deep learning with kernel methods (cseweb.ucsd.edu/~yoc002/paper/nips09_arccos.pdf) if no one can offer a better option.
On Thu, Jun 12, 2014 at 11:31 AM, Kyung Hyun Cho <cho.k.hyun@gmail.com> wrote:
_______________________________________________Dear all,Ian Goodfellow will guide us to the wonderful world of adversarial sampling away from the hell of MCMC and iterative inference at the tea talk on Monday (http://arxiv-web3.library.cornell.edu/abs/1406.2661, the paper attached)
Hope to see many of you there!- Cho
P.S. I am looking for volunteers who want to share their ideas or discuss interesting papers at a tea talk! See https://docs.google.com/spreadsheets/d/1_bbHxcm4r-rs63chcKfHuBsM3roHuud4CJJjCEqQfII/edit#gid=0
========- Speaker: Ian Goodfellow- Date and Time: 16 June 2014 @13.00- Place: AA3195- Abstract:We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to 1/2 everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.
Lisa_labo mailing list
Lisa_labo@iro.umontreal.ca
https://webmail.iro.umontreal.ca/mailman/listinfo/lisa_labo
--