Speaker: Dr. Geoffrey Hinton
Department of Computer Science
University of Toronto
Title: Deep Learning for Object Recognition and Motion Generation
Abstract: I will describe how multiple layers of non-linear features can be learned one layer at a time from unlabeled images. The multiple layers of features form a good generative model of images and the highest level features are very good for object recognition. I will show how the same learning methods can be applied to high-dimensional sequential data to produce good generative models of people walking in many different styles.
Live from Toronto, Ontario
The Coast-To-Coast seminar is a Canada wide seminar series on topics in
mathematics and computing science, presented via AccessGrid technology.
The main goal of the series is to provide an opportunity to scientific
communities from various Canadian universities to collaborate and share
their interests. The secondary goal is to set the standard for these
types of events and to explore further the possibilities of the use of
the available technology.
Organized and distributed by AARMS, ACEnet, the IRMACS Centre, SciNet,
SHARCNET, and WestGrid
For more information go to http://www.scinet.utoronto.ca/resources/courses.htm
November 3 - Dr. Simon Haykin, McMaster University,
November 17 - Dr. Richard Vaughan, Simon Fraser University,
December 1 - Dr. Thomas Trappenberg, Dalhousie University.