Presentation 2011-03-10
Statistical Analysis on Chaotic Neural Networks with Dynamical Noise Injection
Takayuki SUZUKI, Tohru IKEGUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We have already proposed an approximate algorithm for solving QAP by injecting dynamical noise to internal states of a chaotic neural network. As a result, in case of small amount of dynamical noise, the chaotic neural network shows higher performance than without dynamical noise. To investigate its reason, we analyzed the temporal changes of firing pattern generated from the chaotic neural network with additive dynamical noise. The firing patterns and obtained solutions are sensitive to the amount of the noise. In addition, good solutions close to the optimal solutions are obtained with high probability when small noise is added. Therefore, we focus on the temporal change of firing patterns. In this paper, we investigate the temporal intervals of firing. As a result, the frequency distribution conform to the rule of power law. In particular, when small level of noise leads to good performance for solving QAPs, the power-law distribution shows its exponents -1.5.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Quadratic Assignment Problem / Chaotic Neural Network / Dynamical noise / Self-Organized Criticality
Paper # NLP2010-165
Date of Issue

Conference Information
Committee NLP
Conference Date 2011/3/3(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Statistical Analysis on Chaotic Neural Networks with Dynamical Noise Injection
Sub Title (in English)
Keyword(1) Quadratic Assignment Problem
Keyword(2) Chaotic Neural Network
Keyword(3) Dynamical noise
Keyword(4) Self-Organized Criticality
1st Author's Name Takayuki SUZUKI
1st Author's Affiliation Graduate School of Science and Engineering, Saitama University()
2nd Author's Name Tohru IKEGUCHI
2nd Author's Affiliation Graduate School of Science and Engineering, Saitama University
Date 2011-03-10
Paper # NLP2010-165
Volume (vol) vol.110
Number (no) 465
Page pp.pp.-
#Pages 5
Date of Issue