Presentation 2008-03-28
Action Learning by An Artificial Neural Network
Risa SHIMADA, Kenya JIN'NO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We would like to understand a learning feature of human. To realize such purpose, we try to learn an action by artificial neural networks. In this article, we consider an operation to load a building block into Tetris as a simple problem which is two dimensional problem. In order to learn an action, we use a supervised learning algorithm and a reinforcement learning algorith. For the supervised learning, namely, backpropergation algorithm, our numerical simulation result indicates that we cannot get a desired learning reslut. However, we confirm that the reinforcement learning algorithm conduces to a desired learning result which the block can be moved toward to the lowest location. Based on the result of reinforcement learning, we modify the backpropagation algorithm. Consequently, we confirm that the backpropagation algorithm endows the neural network with a generalized ability.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neural network / back propagation / reinforcement learning / Tetris
Paper # NLP2007-170
Date of Issue

Conference Information
Committee NLP
Conference Date 2008/3/21(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 Nonlinear Problems (NLP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Action Learning by An Artificial Neural Network
Sub Title (in English)
Keyword(1) neural network
Keyword(2) back propagation
Keyword(3) reinforcement learning
Keyword(4) Tetris
1st Author's Name Risa SHIMADA
1st Author's Affiliation Department of Network and Multi-Media Engineering, Faculty of Engineering, Kanto Gakuin University()
2nd Author's Name Kenya JIN'NO
2nd Author's Affiliation Department of Network and Multi-Media Engineering, Faculty of Engineering, Kanto Gakuin University
Date 2008-03-28
Paper # NLP2007-170
Volume (vol) vol.107
Number (no) 561
Page pp.pp.-
#Pages 6
Date of Issue