Learning Expected Hitting Time Distance
Most distance metric learning (DML) approaches focus on learning a Mahalanobis metric for measuring distances between examples. However, for particular feature representations, e.g., histogram features like BOW and SPM, Mahalanobis metric could not model the correlations between these features well....
Published in: | Proceedings of the AAAI Conference on Artificial Intelligence |
---|---|
Main Authors: | , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Association for the Advancement of Artificial Intelligence
2016
|
Subjects: | |
Online Access: | https://ojs.aaai.org/index.php/AAAI/article/view/10277 https://doi.org/10.1609/aaai.v30i1.10277 |
id |
ftjaaai:oai:ojs.aaai.org:article/10277 |
---|---|
record_format |
openpolar |
spelling |
ftjaaai:oai:ojs.aaai.org:article/10277 2023-05-15T16:01:51+02:00 Learning Expected Hitting Time Distance Zhan, De-Chuan Hu, Peng Chu, Zui Zhou, Zhi-Hua 2016-03-02 application/pdf https://ojs.aaai.org/index.php/AAAI/article/view/10277 https://doi.org/10.1609/aaai.v30i1.10277 eng eng Association for the Advancement of Artificial Intelligence https://ojs.aaai.org/index.php/AAAI/article/view/10277/10136 https://ojs.aaai.org/index.php/AAAI/article/view/10277 doi:10.1609/aaai.v30i1.10277 Proceedings of the AAAI Conference on Artificial Intelligence; Vol. 30 No. 1 (2016): Thirtieth AAAI Conference on Artificial Intelligence 2374-3468 2159-5399 Expected Hitting Time Metric Learning info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion Peer-reviewed Article 2016 ftjaaai https://doi.org/10.1609/aaai.v30i1.10277 2022-07-02T23:40:45Z Most distance metric learning (DML) approaches focus on learning a Mahalanobis metric for measuring distances between examples. However, for particular feature representations, e.g., histogram features like BOW and SPM, Mahalanobis metric could not model the correlations between these features well. In this work, we define a non-Mahalanobis distance for histogram features, via Expected Hitting Time (EHT) of Markov Chain, which implicitly considers the high-order feature relationships between different histogram features. The EHT based distance is parameterized by transition probabilities of Markov Chain, we consequently propose a novel type of distance learning approach (LED, Learning Expected hitting time Distance) to learn appropriate transition probabilities for EHT based distance. We validate the effectiveness of LED on a series of real-world datasets. Moreover, experiments show that the learned transition probabilities are with good comprehensibility. Article in Journal/Newspaper DML AAAI Publications (Association for the Advancement of Artificial Intelligence) Proceedings of the AAAI Conference on Artificial Intelligence 30 1 |
institution |
Open Polar |
collection |
AAAI Publications (Association for the Advancement of Artificial Intelligence) |
op_collection_id |
ftjaaai |
language |
English |
topic |
Expected Hitting Time Metric Learning |
spellingShingle |
Expected Hitting Time Metric Learning Zhan, De-Chuan Hu, Peng Chu, Zui Zhou, Zhi-Hua Learning Expected Hitting Time Distance |
topic_facet |
Expected Hitting Time Metric Learning |
description |
Most distance metric learning (DML) approaches focus on learning a Mahalanobis metric for measuring distances between examples. However, for particular feature representations, e.g., histogram features like BOW and SPM, Mahalanobis metric could not model the correlations between these features well. In this work, we define a non-Mahalanobis distance for histogram features, via Expected Hitting Time (EHT) of Markov Chain, which implicitly considers the high-order feature relationships between different histogram features. The EHT based distance is parameterized by transition probabilities of Markov Chain, we consequently propose a novel type of distance learning approach (LED, Learning Expected hitting time Distance) to learn appropriate transition probabilities for EHT based distance. We validate the effectiveness of LED on a series of real-world datasets. Moreover, experiments show that the learned transition probabilities are with good comprehensibility. |
format |
Article in Journal/Newspaper |
author |
Zhan, De-Chuan Hu, Peng Chu, Zui Zhou, Zhi-Hua |
author_facet |
Zhan, De-Chuan Hu, Peng Chu, Zui Zhou, Zhi-Hua |
author_sort |
Zhan, De-Chuan |
title |
Learning Expected Hitting Time Distance |
title_short |
Learning Expected Hitting Time Distance |
title_full |
Learning Expected Hitting Time Distance |
title_fullStr |
Learning Expected Hitting Time Distance |
title_full_unstemmed |
Learning Expected Hitting Time Distance |
title_sort |
learning expected hitting time distance |
publisher |
Association for the Advancement of Artificial Intelligence |
publishDate |
2016 |
url |
https://ojs.aaai.org/index.php/AAAI/article/view/10277 https://doi.org/10.1609/aaai.v30i1.10277 |
genre |
DML |
genre_facet |
DML |
op_source |
Proceedings of the AAAI Conference on Artificial Intelligence; Vol. 30 No. 1 (2016): Thirtieth AAAI Conference on Artificial Intelligence 2374-3468 2159-5399 |
op_relation |
https://ojs.aaai.org/index.php/AAAI/article/view/10277/10136 https://ojs.aaai.org/index.php/AAAI/article/view/10277 doi:10.1609/aaai.v30i1.10277 |
op_doi |
https://doi.org/10.1609/aaai.v30i1.10277 |
container_title |
Proceedings of the AAAI Conference on Artificial Intelligence |
container_volume |
30 |
container_issue |
1 |
_version_ |
1766397553750835200 |