| Presentation | 2012-11-17 Composite likelihood estimation for bipartite Boltzmann machines Takashi ASARI, Muneki YASUDA, Yuji WAIZUMI, Kazuyuki TANAKA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | The recent development of information technology has enabled us to obtain and to storage huge information data. Because of this, effective usages of the information data have been one of the central points of information sciences. A Boltzmann machine is a model of the machine learning theory and can be a powerful tool for the aim. In this paper, we focus on a bipartite Boltzmann machine (BBM) consisted of two different layers: one layer is the input layer and the other is the output layer. In this model, once the input layer is clamped by given data, inference of the output layer is quite easy because of the property of conditional independence. Therefore, this model is expected to be applied to expert systems such as medical test system. However, the learning of BBM using the maximum likelihood estimation is still intractable, Because the computational complexity of learning exponentially grows with the increase in the size of system, In this paper, we propose a new learning algorithm for BBM by using the composite likelihood estimation (CLE) which is a statistical mathematical technique to approximate the MLE. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | probabilistic information processing / statistical machine learning theory / Boltzmann machine / maximum likelihood estimation / composite likelihood estimation |
| Paper # | NC2012-68 |
| Date of Issue |
| Conference Information | |
| Committee | NC |
|---|---|
| Conference Date | 2012/11/9(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) | Composite likelihood estimation for bipartite Boltzmann machines |
| Sub Title (in English) | |
| Keyword(1) | probabilistic information processing |
| Keyword(2) | statistical machine learning theory |
| Keyword(3) | Boltzmann machine |
| Keyword(4) | maximum likelihood estimation |
| Keyword(5) | composite likelihood estimation |
| 1st Author's Name | Takashi ASARI |
| 1st Author's Affiliation | Graduate School of Information Sciences, Tohoku University() |
| 2nd Author's Name | Muneki YASUDA |
| 2nd Author's Affiliation | Graduate School of Information Sciences, Tohoku University |
| 3rd Author's Name | Yuji WAIZUMI |
| 3rd Author's Affiliation | Graduate School of Information Sciences, Tohoku University |
| 4th Author's Name | Kazuyuki TANAKA |
| 4th Author's Affiliation | Graduate School of Information Sciences, Tohoku University |
| Date | 2012-11-17 |
| Paper # | NC2012-68 |
| Volume (vol) | vol.112 |
| Number (no) | 298 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |