Disjoint Label Space Transfer Learning with Common Factorised Space

Xiaobin Chang, Yongxin Yang, Tao Xiang, Timothy Hospedales

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper, a unified approach is presented to transfer learning that addresses several source and target domain labelspace and annotation assumptions with a single model. It is particularly effective in handling a challenging case, where source and target label-spaces are disjoint, and outperforms alternatives in both unsupervised and semi-supervised settings. The key ingredient is a common representation termed Common Factorised Space. It is shared between source and target domains, and trained with an unsupervised factorisation loss and a graph-based loss. With a wide range of experiments, we demonstrate the flexibility, relevance and efficacy of our method, both in the challenging cases with disjoint label spaces, and in the more conventional cases such as unsupervised domain adaptation, where the source and target domains share the same label-sets.
Original languageEnglish
Title of host publicationProceedings of the Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19)
Subtitle of host publicationThirty-First Conference on Innovative Applications of Artificial Intelligence The Ninth Symposium on Educational Advances in Artificial Intelligence - AAAI Technical Track: Machine Learning
Place of PublicationHonolulu, Hawaii, United States
PublisherAAAI Press
Pages3288-3295
Number of pages8
Volume33
ISBN (Print)978-1-57735-809-1
DOIs
Publication statusPublished - 23 Jul 2019
EventThe Thirty-Third AAAI Conference on Artificial Intelligence - Hilton Hawaiian Village, Honolulu, Hawaii, United States
Duration: 27 Jan 20191 Feb 2019
https://aaai.org/Conferences/AAAI-19/

Publication series

NameProceedings of the Thirty-Third AAAI Conference on Artificial Intelligence
Publisher AAAI Press
Number1
Volume33
ISSN (Print)2159-5399
ISSN (Electronic)2374-3468

Conference

ConferenceThe Thirty-Third AAAI Conference on Artificial Intelligence
Abbreviated titleAAAI 2019
CountryUnited States
CityHonolulu, Hawaii
Period27/01/191/02/19
Internet address

Fingerprint Dive into the research topics of 'Disjoint Label Space Transfer Learning with Common Factorised Space'. Together they form a unique fingerprint.

Cite this