Practical large scale classification with additive kernels
For classification problems with millions of training examples or dimensions, accuracy, training and testing speed and memory usage are the main concerns. Recent advances have allowed linear SVM to tackle problems with moderate time and space cost, but for many tasks in computer vision, additive ker...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Language: | English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/106282 http://hdl.handle.net/10220/24003 http://jmlr.org/proceedings/papers/v25/yang12/yang12.pdf |
| _version_ | 1826118066103123968 |
|---|---|
| author | Yang, Hao Wu, Jianxin |
| author2 | School of Computer Engineering |
| author_facet | School of Computer Engineering Yang, Hao Wu, Jianxin |
| author_sort | Yang, Hao |
| collection | NTU |
| description | For classification problems with millions of training examples or dimensions, accuracy, training and testing speed and memory usage are the main concerns. Recent advances have allowed linear SVM to tackle problems with moderate time and space cost, but for many tasks in computer vision, additive kernels would have higher accuracies. In this paper, we propose the PmSVM-LUT algorithm that employs Look-Up Tables to boost the training and testing speed and save memory usage of additive kernel SVM classification, in order to meet the needs of large scale problems. The PmSVM-LUT algorithm is based on PmSVM (Wu, 2012), which employed polynomial approximation for the gradient function to speedup the dual coordinate descent method. We also analyze the polynomial approximation numerically to demonstrate its validity. Empirically, our algorithm is faster than PmSVM and feature mapping in many datasets with higher classification accuracies and can save up to 60% memory usage as well. |
| first_indexed | 2024-10-01T04:37:39Z |
| format | Conference Paper |
| id | ntu-10356/106282 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T04:37:39Z |
| publishDate | 2014 |
| record_format | dspace |
| spelling | ntu-10356/1062822020-05-28T07:18:31Z Practical large scale classification with additive kernels Yang, Hao Wu, Jianxin School of Computer Engineering Asian Conference on Machine Learning, ACML (4th : 2012) DRNTU::Engineering::Computer science and engineering For classification problems with millions of training examples or dimensions, accuracy, training and testing speed and memory usage are the main concerns. Recent advances have allowed linear SVM to tackle problems with moderate time and space cost, but for many tasks in computer vision, additive kernels would have higher accuracies. In this paper, we propose the PmSVM-LUT algorithm that employs Look-Up Tables to boost the training and testing speed and save memory usage of additive kernel SVM classification, in order to meet the needs of large scale problems. The PmSVM-LUT algorithm is based on PmSVM (Wu, 2012), which employed polynomial approximation for the gradient function to speedup the dual coordinate descent method. We also analyze the polynomial approximation numerically to demonstrate its validity. Empirically, our algorithm is faster than PmSVM and feature mapping in many datasets with higher classification accuracies and can save up to 60% memory usage as well. Published version 2014-10-13T02:28:53Z 2019-12-06T22:08:00Z 2014-10-13T02:28:53Z 2019-12-06T22:08:00Z 2012 2012 Conference Paper Yang, H., & Wu, J. (2012). Practical large scale classification with additive kernels. Journal of machine learning research: workshop and conference proceedings, 25, 523-538. https://hdl.handle.net/10356/106282 http://hdl.handle.net/10220/24003 http://jmlr.org/proceedings/papers/v25/yang12/yang12.pdf en © 2012 The Authors(Journal of Machine Learning Research). This paper was published in Journal of Machine Learning Research and is made available as an electronic reprint (preprint) with permission of The Authors(Journal of Machine Learning Research). The paper can be found at the following official URL: [http://jmlr.org/proceedings/papers/v25/yang12/yang12.pdf]. One print or electronic copy may be made for personal use only. Systematic or multiple reproduction, distribution to multiple locations via electronic or other means, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper is prohibited and is subject to penalties under law. 16 p. application/pdf |
| spellingShingle | DRNTU::Engineering::Computer science and engineering Yang, Hao Wu, Jianxin Practical large scale classification with additive kernels |
| title | Practical large scale classification with additive kernels |
| title_full | Practical large scale classification with additive kernels |
| title_fullStr | Practical large scale classification with additive kernels |
| title_full_unstemmed | Practical large scale classification with additive kernels |
| title_short | Practical large scale classification with additive kernels |
| title_sort | practical large scale classification with additive kernels |
| topic | DRNTU::Engineering::Computer science and engineering |
| url | https://hdl.handle.net/10356/106282 http://hdl.handle.net/10220/24003 http://jmlr.org/proceedings/papers/v25/yang12/yang12.pdf |
| work_keys_str_mv | AT yanghao practicallargescaleclassificationwithadditivekernels AT wujianxin practicallargescaleclassificationwithadditivekernels |