Department of Computer Science, University of Toronto
Modeling Natural Images Using Higher-Order Boltzmann Machines
Tuesday 22nd of June 2010 at 12:00pm
Learning a generative model of natural images is a useful way of extracting features that capture interesting regularities. Previous work on learning such models has focused on methods in which the latent features are used to determine the mean and variance of each pixel independently, or on methods in which the hidden units determine the covariance matrix of a zero-mean Gaussian distribution. In this talk, I will describe a probabilistic model that combines these two approaches into a single framework. Each image is represented by using one set of binary latent features that model the image-specific covariance and a separate set that model the mean. I will give a few different interpretations of this approach and show how it relates to the widely used simple-cell complex-cell model of early vision. I will demonstrate this model by generating images and by using the features to recognize objects in the CIFAR 10 dataset.
508-20 Evans Hall
This is joint work with Geoff Hinton.
Join Email List
You can subscribe to our weekly seminar email list by sending an email to
firstname.lastname@example.org that contains the words
subscribe redwood in the body of the message.
(Note: The subject line can be arbitrary and will be ignored)