Skip to main content
Duke University Libraries
DukeSpace Scholarship by Duke Authors
  • Login
  • Ask
  • Menu
  • Login
  • Ask a Librarian
  • Search & Find
  • Using the Library
  • Research Support
  • Course Support
  • Libraries
  • About
View Item 
  •   DukeSpace
  • Theses and Dissertations
  • Duke Dissertations
  • View Item
  •   DukeSpace
  • Theses and Dissertations
  • Duke Dissertations
  • View Item
JavaScript is disabled for your browser. Some features of this site may not work without it.

Towards Better Representations with Deep/Bayesian Learning

Thumbnail
View / Download
13.9 Mb
Date
2018
Author
Li, Chunyuan
Advisor
Carin, Lawrence
Repository Usage Stats
836
views
503
downloads
Abstract

Deep learning and Bayesian Learning are two popular research topics in machine learning. They provide the flexible representations in the complementary manner. Therefore, it is desirable to take the best from both fields. This thesis focuses on the intersection of the two topics— enriching one with each other. Two new research topics are inspired: Bayesian deep learning and Deep Bayesian learning.

In Bayesian deep learning, scalable Bayesian methods are proposed to learn the weight uncertainty of deep neural networks (DNNs). On this topic, I propose the preconditioned stochastic gradient MCMC methods, then show its connection to Dropout, and its applications to modern network architectures in computer vision and natural language processing.

In Deep Bayesian learning: DNNs are employed as powerful representations of conditionals in traditional Bayesian models. I will focus on understanding the recent adversarial learning methods for joint distribution matching, through which several recent bivariate adversarial models are unified. It further raises the non-identifiability issues in bidirectional adversarial learning, and propose ALICE algorithms: a conditional entropy framework to remedy the issues. The derived algorithms show significant improvement in the tasks of image generation and translation, by solving the non-identifiability issues.

Description
Dissertation
Type
Dissertation
Department
Electrical and Computer Engineering
Subject
Artificial intelligence
Statistics
Computer science
adversarial learning
Bayesian learning
deep learning
generative models
neural networks
Permalink
https://hdl.handle.net/10161/18207
Citation
Li, Chunyuan (2018). Towards Better Representations with Deep/Bayesian Learning. Dissertation, Duke University. Retrieved from https://hdl.handle.net/10161/18207.
Collections
  • Duke Dissertations
More Info
Show full item record
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.

Rights for Collection: Duke Dissertations


Works are deposited here by their authors, and represent their research and opinions, not that of Duke University. Some materials and descriptions may include offensive content. More info

Make Your Work Available Here

How to Deposit

Browse

All of DukeSpaceCommunities & CollectionsAuthorsTitlesTypesBy Issue DateDepartmentsAffiliations of Duke Author(s)SubjectsBy Submit DateThis CollectionAuthorsTitlesTypesBy Issue DateDepartmentsAffiliations of Duke Author(s)SubjectsBy Submit Date

My Account

LoginRegister

Statistics

View Usage Statistics
Duke University Libraries

Contact Us

411 Chapel Drive
Durham, NC 27708
(919) 660-5870
Perkins Library Service Desk

Digital Repositories at Duke

  • Report a problem with the repositories
  • About digital repositories at Duke
  • Accessibility Policy
  • Deaccession and DMCA Takedown Policy

TwitterFacebookYouTubeFlickrInstagramBlogs

Sign Up for Our Newsletter
  • Re-use & Attribution / Privacy
  • Harmful Language Statement
  • Support the Libraries
Duke University