Rank priors for continuous non-linear dimensionality reduction
Discovering the underlying low-dimensional latent structure in high-dimensional perceptual observations (e.g., images, video) can, in many cases, greatly improve performance in recognition and tracking. However, non-linear dimensionality reduction methods are often susceptible to local minima and pe...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | en_US |
| Published: |
Institute of Electrical and Electronics Engineers
2010
|
| Online Access: | http://hdl.handle.net/1721.1/59287 |
| _version_ | 1826195416826249216 |
|---|---|
| author | Darrell, Trevor J. Urtasun, Raquel Geiger, Andreas |
| author2 | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory |
| author_facet | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Darrell, Trevor J. Urtasun, Raquel Geiger, Andreas |
| author_sort | Darrell, Trevor J. |
| collection | MIT |
| description | Discovering the underlying low-dimensional latent structure in high-dimensional perceptual observations (e.g., images, video) can, in many cases, greatly improve performance in recognition and tracking. However, non-linear dimensionality reduction methods are often susceptible to local minima and perform poorly when initialized far from the global optimum, even when the intrinsic dimensionality is known a priori. In this work we introduce a prior over the dimensionality of the latent space that penalizes high dimensional spaces, and simultaneously optimize both the latent space and its intrinsic dimensionality in a continuous fashion. Ad-hoc initialization schemes are unnecessary with our approach; we initialize the latent space to the observation space and automatically infer the latent dimensionality. We report results applying our prior to various probabilistic non-linear dimensionality reduction tasks, and show that our method can outperform graph-based dimensionality reduction techniques as well as previously suggested initialization strategies. We demonstrate the effectiveness of our approach when tracking and classifying human motion. |
| first_indexed | 2024-09-23T10:12:19Z |
| format | Article |
| id | mit-1721.1/59287 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T10:12:19Z |
| publishDate | 2010 |
| publisher | Institute of Electrical and Electronics Engineers |
| record_format | dspace |
| spelling | mit-1721.1/592872022-09-30T19:35:58Z Rank priors for continuous non-linear dimensionality reduction Darrell, Trevor J. Urtasun, Raquel Geiger, Andreas Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Darrell, Trevor J. Darrell, Trevor J. Urtasun, Raquel Discovering the underlying low-dimensional latent structure in high-dimensional perceptual observations (e.g., images, video) can, in many cases, greatly improve performance in recognition and tracking. However, non-linear dimensionality reduction methods are often susceptible to local minima and perform poorly when initialized far from the global optimum, even when the intrinsic dimensionality is known a priori. In this work we introduce a prior over the dimensionality of the latent space that penalizes high dimensional spaces, and simultaneously optimize both the latent space and its intrinsic dimensionality in a continuous fashion. Ad-hoc initialization schemes are unnecessary with our approach; we initialize the latent space to the observation space and automatically infer the latent dimensionality. We report results applying our prior to various probabilistic non-linear dimensionality reduction tasks, and show that our method can outperform graph-based dimensionality reduction techniques as well as previously suggested initialization strategies. We demonstrate the effectiveness of our approach when tracking and classifying human motion. 2010-10-13T18:13:57Z 2010-10-13T18:13:57Z 2009-08 2009-06 Article http://purl.org/eprint/type/JournalArticle 978-1-4244-3992-8 1063-6919 INSPEC Accession Number: 10835871 http://hdl.handle.net/1721.1/59287 Geiger, A., R. Urtasun, and T. Darrell. “Rank priors for continuous non-linear dimensionality reduction.” Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on. 2009. 880-887. © 2009 Institute of Electrical and Electronics Engineers. en_US http://dx.doi.org/10.1109/CVPRW.2009.5206672 IEEE Conference on Computer Vision and Pattern Recognition, 2009. CVPR 2009 Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/pdf Institute of Electrical and Electronics Engineers IEEE |
| spellingShingle | Darrell, Trevor J. Urtasun, Raquel Geiger, Andreas Rank priors for continuous non-linear dimensionality reduction |
| title | Rank priors for continuous non-linear dimensionality reduction |
| title_full | Rank priors for continuous non-linear dimensionality reduction |
| title_fullStr | Rank priors for continuous non-linear dimensionality reduction |
| title_full_unstemmed | Rank priors for continuous non-linear dimensionality reduction |
| title_short | Rank priors for continuous non-linear dimensionality reduction |
| title_sort | rank priors for continuous non linear dimensionality reduction |
| url | http://hdl.handle.net/1721.1/59287 |
| work_keys_str_mv | AT darrelltrevorj rankpriorsforcontinuousnonlineardimensionalityreduction AT urtasunraquel rankpriorsforcontinuousnonlineardimensionalityreduction AT geigerandreas rankpriorsforcontinuousnonlineardimensionalityreduction |