| Presentation | 2022-03-27 An Improvement of Prediction Performance of Reservoir Computing using Deep FORCE Learning Kazuki Nakada, Eiji Suzuki, Keita Suda, Yukio Terasaki, Tetsuya Asai, Tomoyuki Sasaki, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | The physical implementation has become increasingly important in the recent machine learning trends. Reservoir Computing (RC) is a machine learning framework for time series processing, and the research on RC has progressed rapidly because of the suitability for physical implementation. For implementing RC as edge AI devices, sequential online learning is an essential requirement. Previously, FORCE (First-Order, Reduced and Controlled Error) learning having fast convergence property has been proposed and its advantages over the conventional Stochastic Gradient Descent algorithm were demonstrated. However, RC is not suitable for incremental learning of multi-class data due to the restriction of a single output per layer since it is based on RLS (Recursive Least Squares) algorithm. In this report, we propose a natural extension of FORCE learning, Deep FORCE, based on Kalman filtering. We demonstrated that Deep FORCE can improve the prediction performance of real trajectory of a double pendulum as a nonlinear transformation task as compared to conventional FORCE learning. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Reservoir Computing / Online leaning / FORCE learning |
| Paper # | CCS2021-40 |
| Date of Issue | 2022-03-20 (CCS) |
| Conference Information | |
| Committee | CCS |
|---|---|
| Conference Date | 2022/3/27(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | RUSUTSU RESORT HOTEL & CONVENTION |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Tetsuya Asai(Hokkaido Univ.) |
| Vice Chair | Megumi Akai(Hokkaido Univ.) / Masaki Aida(TMU) |
| Secretary | Megumi Akai(TDK) / Masaki Aida(Mie Univ.) |
| Assistant | Hidehiro Nakano(Tokyo City Univ.) / Hiroyasu Ando(Tsukuba Univ.) / Takashi Matsubara(Osaka Univ.) / Sumiko Miyata(Shibaura Inst. of Tech.) |
| Paper Information | |
| Registration To | Technical Committee on Complex Communication Sciences |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | An Improvement of Prediction Performance of Reservoir Computing using Deep FORCE Learning |
| Sub Title (in English) | |
| Keyword(1) | Reservoir Computing |
| Keyword(2) | Online leaning |
| Keyword(3) | FORCE learning |
| 1st Author's Name | Kazuki Nakada |
| 1st Author's Affiliation | TDK Corporation(TDK) |
| 2nd Author's Name | Eiji Suzuki |
| 2nd Author's Affiliation | TDK Corporation(TDK) |
| 3rd Author's Name | Keita Suda |
| 3rd Author's Affiliation | TDK Corporation(TDK) |
| 4th Author's Name | Yukio Terasaki |
| 4th Author's Affiliation | TDK Corporation(TDK) |
| 5th Author's Name | Tetsuya Asai |
| 5th Author's Affiliation | Hokkaido University(Hokkaido Univ.) |
| 6th Author's Name | Tomoyuki Sasaki |
| 6th Author's Affiliation | TDK Corporation(TDK) |
| Date | 2022-03-27 |
| Paper # | CCS2021-40 |
| Volume (vol) | vol.121 |
| Number (no) | CCS-442 |
| Page | pp.pp.25-30(CCS), |
| #Pages | 6 |
| Date of Issue | 2022-03-20 (CCS) |