Seminar Details

Seminar Details


Monday

Mar 9

3:30 pm

Overview of the Intent and Goals of the Programme on Neural Networks and Machine Learning at the Isaac Newton Institute

Thomas Richardson

Seminar

University of Washington - Department of Statistics

In this talk I will give a summary of the recent Newton Institute Programme on Neural Networks and Machine Learning, which I attended.

I will discuss the main topics covered in the workshop, including a brief sketch (time permitting) of the following, which struck me as being of particular interest:

1. Slice sampling - a new technique in MCMC, primarily due to Neal.
2. Independent component analysis - a technique developed by Sejnowski and Amari (among others) for carrying out blind separation of non-gaussian sources.
3. Error Correcting codes: MacKay and Neal have used Pearl's Bayesian network belief propagation algorithm to come up with very efficient error correcting codes that operate close to the Shannon Limit.