| Presentation | 2014-11-17 Feature Extraction for Image Classification using Restricted Boltzmann Machines Reiki SUDA, Koujin TAKEDA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Learning restricted Boltzmann machines (RBMs) for high-dimensional data using maximum likelihood estimation had been faced with computational complexity. However, after the development of approximation algorithms based on Markov chain Monte Carlo methods, RBMs have been applicable even to high-dimensional data, and playing a central role as feature extractors. In this study, we conduct experiments of feature extraction from handwritten digit images using RBMs and of classification with the extracted features. As classifiers, fast and simple linear classifiers based on confidence-weighted learning are used. Finally, we show that classifiers trained with features extracted by RBMs outperform ones trained with raw data, and from the result we also evaluate the learning performance of the RBMs. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | restricted Boltzmann machines / image recognition / feature extraction / Markov chain Monte Carlo methods / linear classifiers / confidence-weighted learning |
| Paper # | IBISML2014-36 |
| Date of Issue |
| Conference Information | |
| Committee | IBISML |
|---|---|
| Conference Date | 2014/11/10(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Information-Based Induction Sciences and Machine Learning (IBISML) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Feature Extraction for Image Classification using Restricted Boltzmann Machines |
| Sub Title (in English) | |
| Keyword(1) | restricted Boltzmann machines |
| Keyword(2) | image recognition |
| Keyword(3) | feature extraction |
| Keyword(4) | Markov chain Monte Carlo methods |
| Keyword(5) | linear classifiers |
| Keyword(6) | confidence-weighted learning |
| 1st Author's Name | Reiki SUDA |
| 1st Author's Affiliation | Department of Intelligent Systems Engineering, Ibaraki University() |
| 2nd Author's Name | Koujin TAKEDA |
| 2nd Author's Affiliation | Department of Intelligent Systems Engineering, Ibaraki University |
| Date | 2014-11-17 |
| Paper # | IBISML2014-36 |
| Volume (vol) | vol.114 |
| Number (no) | 306 |
| Page | pp.pp.- |
| #Pages | 7 |
| Date of Issue | |