| Presentation | 2011-07-25 Belief Propagation for Slow Feature Analysis Tomoki SEKIGUCHI, Toshiaki OMORI, Masato OKADA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | The slow feature analysis (SFA) is a data analysis method that extracts slowly varying features from time series data. In this study, we propose an estimation algorithm of the probabilistic model of SFA that takes into account an effect of observation noise. Recently, a probabilistic model of the SFA has been proposed. However, the conventional algorithm using the probabilistic SFA model approximately evaluates the likelihood function by assuming observation noise to be zero. This means that observed data is assumed not to include noise in the conventional algorithm. Therefore it is not clear whether we can obtain a suitable result when we apply this conventional algorithm to real noisy data. In this study, we derived a likelihood function exactly in the probabilistic SFA model, that takes into account an effect of observation noise. From the graphical structure of the probabilistic SFA model, we showed that it is possible to derive the likelihood function exactly by using belief propagation. We applied our algorithm to artificial data, and showed that it could accurately estimate the slow feature from data including observation noise. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Slow Feature Analysis / state-space model / belief propagation / graphical structure / probabilistic time series analysis |
| Paper # | NC2011-26 |
| Date of Issue |
| Conference Information | |
| Committee | NC |
|---|---|
| Conference Date | 2011/7/18(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 | Neurocomputing (NC) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Belief Propagation for Slow Feature Analysis |
| Sub Title (in English) | |
| Keyword(1) | Slow Feature Analysis |
| Keyword(2) | state-space model |
| Keyword(3) | belief propagation |
| Keyword(4) | graphical structure |
| Keyword(5) | probabilistic time series analysis |
| 1st Author's Name | Tomoki SEKIGUCHI |
| 1st Author's Affiliation | Graduate School of Frontier Sciences, The University of Tokyo() |
| 2nd Author's Name | Toshiaki OMORI |
| 2nd Author's Affiliation | Graduate School of Frontier Sciences, The University of Tokyo:RIKEN Brain Science Institute |
| 3rd Author's Name | Masato OKADA |
| 3rd Author's Affiliation | Graduate School of Frontier Sciences, The University of Tokyo:RIKEN Brain Science Institute |
| Date | 2011-07-25 |
| Paper # | NC2011-26 |
| Volume (vol) | vol.111 |
| Number (no) | 157 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |