Presentation | 2024-03-11 Time-series Forecasting Coding: Proposal of A New Processing Method Developed from Predictive Coding for Recurrent Neural Networks Yuto Wakui, Junya Kato, Ryo Natsuaki, Akira Hirose, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | This paper proposes a new method for processing time-series data in recurrent neural networks (RNNs), namely, time-series forecasting coding (TFC). We developed them based on the idea of predictive coding (PC). We consider multi-step ahead forecasting task with the following two schemes, that is, smallest-delay feedback TFC and synchronous (temporally matching) feedback TFC. The latter is a natural extension of conventional PC. In experimental results, we show that the smallest-delay TFC presents the highest performance, while synchronous TFC shows comparable or lower performance, and then a conventional RNN method indicates the lowest. This result suggests that the concept of PC is important even in time-series data processing, and that small delay is more important than synchronicity. This may have significant implications to the conceptual fundamentals of the PC and thus of the FEP. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Reservoir computing / Free-energy principle / Predictive coding |
Paper # | NC2023-46 |
Date of Issue | 2024-03-04 (NC) |
Conference Information | |
Committee | NC / MBE |
---|---|
Conference Date | 2024/3/11(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | The Univ. of Tokyo |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Brain architecture, General |
Chair | Hirokazu Tanaka(Tokyo City Univ.) / Hisashi Yoshida(Kinki Univ.) |
Vice Chair | Jun Izawa(Univ. of Tsukub) / Akinori Ueno(Tokyo Denki Univ.) |
Secretary | Jun Izawa(NTT) / Akinori Ueno(NAIST) |
Assistant | Yoshimasa Tawatsuji(Waseda Univ.) / Takato Horii(Osaka Univ.) / Akihiko Tsukahara(Tokyo Denki Univ.) / Miki Kaneko(Osaka Univ.) |
Paper Information | |
Registration To | Technical Committee on Neurocomputing / Technical Committee on ME and Bio Cybernetics |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Time-series Forecasting Coding: Proposal of A New Processing Method Developed from Predictive Coding for Recurrent Neural Networks |
Sub Title (in English) | |
Keyword(1) | Reservoir computing |
Keyword(2) | Free-energy principle |
Keyword(3) | Predictive coding |
1st Author's Name | Yuto Wakui |
1st Author's Affiliation | The University of Tokyo(UTokyo.) |
2nd Author's Name | Junya Kato |
2nd Author's Affiliation | The University of Tokyo(UTokyo.) |
3rd Author's Name | Ryo Natsuaki |
3rd Author's Affiliation | The University of Tokyo(UTokyo.) |
4th Author's Name | Akira Hirose |
4th Author's Affiliation | The University of Tokyo(UTokyo.) |
Date | 2024-03-11 |
Paper # | NC2023-46 |
Volume (vol) | vol.123 |
Number (no) | NC-418 |
Page | pp.pp.19-24(NC), |
#Pages | 6 |
Date of Issue | 2024-03-04 (NC) |