Presentation | 1997/3/17 Adaptive Annealed RNN Learning of Finite State Automata Ken-ichi Arai, Ryohei Nakano, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In the learning of finite state automata (FSA) by a recurrent neural network (RNN), we report a new annealed learning method that adaptively adjusts the neuro gain (β) and a learning rate (η). In the previous report, we proved that an BNN can get stable transitions among clusters when β is larger than the critical value and proposed the annealed learning method with a fixed scheduling of β. In the method, however, an error sometimes rapidly grows after a long term learning. Thus, we propose an adaptive annealed learning method to overcome this problem and to get stable cluster transitions faster and stably. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | recurrent neural network / finite state automata / stability / adaptive annealing |
Paper # | NC96-130 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 1997/3/17(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) | Adaptive Annealed RNN Learning of Finite State Automata |
Sub Title (in English) | |
Keyword(1) | recurrent neural network |
Keyword(2) | finite state automata |
Keyword(3) | stability |
Keyword(4) | adaptive annealing |
1st Author's Name | Ken-ichi Arai |
1st Author's Affiliation | NTT Communication Science Laboratories() |
2nd Author's Name | Ryohei Nakano |
2nd Author's Affiliation | NTT Communication Science Laboratories |
Date | 1997/3/17 |
Paper # | NC96-130 |
Volume (vol) | vol.96 |
Number (no) | 583 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |