Efficient Methods for Unsupervised Learning of Probabilistic Models

TL;DR

This thesis introduces new techniques for training, evaluating, and sampling from complex probabilistic models that are intractable and high-dimensional, advancing unsupervised learning methods.

Contribution

It develops novel methods for handling intractable probabilistic models, improving training and sampling efficiency in unsupervised learning.

Findings

Enhanced training algorithms for complex models

Improved sampling techniques for high-dimensional data

Better evaluation metrics for probabilistic models

Abstract

In this thesis I develop a variety of techniques to train, evaluate, and sample from intractable and high dimensional probabilistic models. Abstract exceeds arXiv space limitations -- see PDF.