Learning Linear, Sparse, Factorial Codes
In previous work (Olshausen & Field 1996), an algorithm was described for learning linear sparse codes which, when trained on natural images, produces a set of basis functions that are spatially localized, oriented, and bandpass (i.e., wavelet-like). This note shows how the algorithm may b...
| Main Author: | |
|---|---|
| Language: | en_US |
| Published: |
2004
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/1721.1/7184 |
| _version_ | 1826216887043751936 |
|---|---|
| author | Olshausen, Bruno A. |
| author_facet | Olshausen, Bruno A. |
| author_sort | Olshausen, Bruno A. |
| collection | MIT |
| description | In previous work (Olshausen & Field 1996), an algorithm was described for learning linear sparse codes which, when trained on natural images, produces a set of basis functions that are spatially localized, oriented, and bandpass (i.e., wavelet-like). This note shows how the algorithm may be interpreted within a maximum-likelihood framework. Several useful insights emerge from this connection: it makes explicit the relation to statistical independence (i.e., factorial coding), it shows a formal relationship to the algorithm of Bell and Sejnowski (1995), and it suggests how to adapt parameters that were previously fixed. |
| first_indexed | 2024-09-23T16:54:43Z |
| id | mit-1721.1/7184 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T16:54:43Z |
| publishDate | 2004 |
| record_format | dspace |
| spelling | mit-1721.1/71842019-04-10T11:52:41Z Learning Linear, Sparse, Factorial Codes Olshausen, Bruno A. unsupervised learning factorial coding sparse coding MIT In previous work (Olshausen & Field 1996), an algorithm was described for learning linear sparse codes which, when trained on natural images, produces a set of basis functions that are spatially localized, oriented, and bandpass (i.e., wavelet-like). This note shows how the algorithm may be interpreted within a maximum-likelihood framework. Several useful insights emerge from this connection: it makes explicit the relation to statistical independence (i.e., factorial coding), it shows a formal relationship to the algorithm of Bell and Sejnowski (1995), and it suggests how to adapt parameters that were previously fixed. 2004-10-20T20:49:08Z 2004-10-20T20:49:08Z 1996-12-01 AIM-1580 CBCL-138 http://hdl.handle.net/1721.1/7184 en_US AIM-1580 CBCL-138 5 p. 233466 bytes 268006 bytes application/postscript application/pdf application/postscript application/pdf |
| spellingShingle | unsupervised learning factorial coding sparse coding MIT Olshausen, Bruno A. Learning Linear, Sparse, Factorial Codes |
| title | Learning Linear, Sparse, Factorial Codes |
| title_full | Learning Linear, Sparse, Factorial Codes |
| title_fullStr | Learning Linear, Sparse, Factorial Codes |
| title_full_unstemmed | Learning Linear, Sparse, Factorial Codes |
| title_short | Learning Linear, Sparse, Factorial Codes |
| title_sort | learning linear sparse factorial codes |
| topic | unsupervised learning factorial coding sparse coding MIT |
| url | http://hdl.handle.net/1721.1/7184 |
| work_keys_str_mv | AT olshausenbrunoa learninglinearsparsefactorialcodes |