Talk Title: Novel Mixture Models to Learn Complex Patterns in High-Dimensional Data
　

Abstract:

While classical mixture model and its variants remain excellent tools to develop generative models for data, we can learn more
informative models under certain real life data generation scenarios by making a few subtle yet fundamental changes to the classical
model. In order to generate a data point, the classical mixture model selects one of the generative component by performing a multinomial
mixture trial over the mixing proportions, and then manifests the various data attributes based on the selected component. Thus, for any given
data point, only a single component is a possible generator. However, there are many real life situations where it makes far more sense to model
a data point as being generated using multiple components. We have recently proposed two such novel mixture modeling frameworks that allow
multiple components to influence data generation, and associated learning algorithms using both Expectation Maximization (EM) and Markov Chain
Monte Carlo (MCMC) based approaches. In this talk, we will describe our models, their applications and key optimization and learning challenges.