An Overview of Deep Learning
Rob Fergus, New York University

My presentation will give a brief introduction to a new area of machine learning called "Deep Learning" that is showing great promise for tackling a wide range of problems in the artificial intelligence arena. I will start by introducing several perceptual problems within this domain that are easy for humans, but very challenging for machines. These differ from many other problems in science in that they are difficult to address with the traditional analytic tools of math and physics so must instead be tackled with statistical approaches, such as machine learning. I will then outline how "deep" models differ from existing "shallow" machine learning approaches, and give theoretical and empirical arguments why these are more likely to lead to solutions to these AI problems.

Background Review Article:
Hinton, G. E. and Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, Vol. 313. no. 5786, pp. 504 - 507, 28 July 2006.