Regularization Through Feature Knock Out
In this paper, we present and analyze a novel regularization technique based on enhancing our dataset with corrupted copies of the original data. The motivation is that since the learning algorithm lacks information about which parts of thedata are reliable, it has to produce more robust classificat...
| Main Authors: | , |
|---|---|
| Language: | en_US |
| Published: |
2005
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/1721.1/30502 |
| _version_ | 1826216783011381248 |
|---|---|
| author | Wolf, Lior Martin, Ian |
| author_facet | Wolf, Lior Martin, Ian |
| author_sort | Wolf, Lior |
| collection | MIT |
| description | In this paper, we present and analyze a novel regularization technique based on enhancing our dataset with corrupted copies of the original data. The motivation is that since the learning algorithm lacks information about which parts of thedata are reliable, it has to produce more robust classification functions. We then demonstrate how this regularization leads to redundancy in the resulting classifiers, which is somewhat in contrast to the common interpretations of the OccamÂs razor principle. Using this framework, we propose a simple addition to the gentle boosting algorithm which enables it to work with only a few examples. We test this new algorithm on a variety of datasets and show convincing results. |
| first_indexed | 2024-09-23T16:53:21Z |
| id | mit-1721.1/30502 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T16:53:21Z |
| publishDate | 2005 |
| record_format | dspace |
| spelling | mit-1721.1/305022019-04-12T08:37:51Z Regularization Through Feature Knock Out Wolf, Lior Martin, Ian AI In this paper, we present and analyze a novel regularization technique based on enhancing our dataset with corrupted copies of the original data. The motivation is that since the learning algorithm lacks information about which parts of thedata are reliable, it has to produce more robust classification functions. We then demonstrate how this regularization leads to redundancy in the resulting classifiers, which is somewhat in contrast to the common interpretations of the OccamÂs razor principle. Using this framework, we propose a simple addition to the gentle boosting algorithm which enables it to work with only a few examples. We test this new algorithm on a variety of datasets and show convincing results. 2005-12-22T02:15:29Z 2005-12-22T02:15:29Z 2004-11-12 MIT-CSAIL-TR-2004-072 AIM-2004-025 CBCL-242 http://hdl.handle.net/1721.1/30502 en_US Massachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory 0 p. 16224097 bytes 656543 bytes application/postscript application/pdf application/postscript application/pdf |
| spellingShingle | AI Wolf, Lior Martin, Ian Regularization Through Feature Knock Out |
| title | Regularization Through Feature Knock Out |
| title_full | Regularization Through Feature Knock Out |
| title_fullStr | Regularization Through Feature Knock Out |
| title_full_unstemmed | Regularization Through Feature Knock Out |
| title_short | Regularization Through Feature Knock Out |
| title_sort | regularization through feature knock out |
| topic | AI |
| url | http://hdl.handle.net/1721.1/30502 |
| work_keys_str_mv | AT wolflior regularizationthroughfeatureknockout AT martinian regularizationthroughfeatureknockout |