Variational Methods for Graph Models with Hidden Variables

Date

2018-05

ORCID

Journal Title

Journal ISSN

Volume Title

Publisher

item.page.doi

Abstract

In this thesis, we focus on the Markov Random Field graph. We gradually transition to Hidden Markov Model and work exclusively with binary graphs. We then utilize produced graphical model to train machine learning model that would classify images of written digits. The thesis is comprised of two parts: review of the articles and implementation (coding) of the key models and methods introduced in the articles. The articles are chosen from the seminal work as well as from the recent advances in graphical models, graphical models with latent variables and their application to various image recognition problems. The last part of the thesis applies developed inference framework to produce machine learning model for written digits recognition task.

Description

Keywords

Markov random fields, Graph theory, Hidden Markov models, Machine learning, Graphic methods

item.page.sponsorship

Rights

©2018 The Author. Digital access to this material is made possible by the Eugene McDermott Library. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Citation