Density Estimation: Views From Machine Learning And Statistics

Date: 13th December 2016

Speaker: Professor Iain Murray

Abstract

Machine learning methods are getting rapidly better at fantasizing novel complex objects, such as images or audio, based on large training sets of examples. Some of the methods behind these impressive demonstrations are also useful for scientific data analysis in two ways. First, the probabilistic models behind the machine learning methods can be a useful component of larger scientific models. Second, some of the same machine learning methods can be used to perform statistical inference, where traditional algorithms (such as Monte Carlo methods) are hard to apply.