Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
OPTIMUM FITTING PARAMETER DETERMINATION METHOD AND DEVICE THEREFOR, AND OPTIMUM FITTING PARAMETER DETERMINATION PROGRAM
Document Type and Number:
Japanese Patent JP3404532
Kind Code:
B2
Abstract:

PROBLEM TO BE SOLVED: To effectively optimize the fitting parameter of a physical model in a short calculation time without depending on an initial value or without sticking to a local peak.
SOLUTION: The parameter of the physical model is expressed with a real number vector as chromosomal information in genetic algorithm, a plurality of individuals expressed by the chromosomes are prepared, and measurement data points of a physical quantity are fitting-optimized by the genetic algorithm. Child individuals of a crossing- over process of the genetic algorithm are generated according to normal distribution probability decided from three parent individuals. In a substitution process, an individual excellent in all standards is set to an individual of a next generation group among the selected parent individuals and the plurality of child individuals. The optimum fitting parameter is found in each combination of measurement data points in a plurality of experimental conditions, and the measurement data points are fitted such that the fitting parameters smoothly change among the experimental conditions. Thereby, a physical phenomenon can be simulated with the physical model without an error even in the experimental conditions under which experimentation is not executed.


Inventors:
Masahiro Murakawa
Tetsuya Higuchi
Application Number:
JP2002160809A
Publication Date:
May 12, 2003
Filing Date:
May 31, 2002
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
National Institute of Advanced Industrial Science and Technology
International Classes:
G06F17/17; G06N3/00; G06N3/12; (IPC1-7): G06N3/00; G06F17/17
Other References:
北野宏明・編,「遺伝的アルゴリズム」,日本,産業図書株式会社,1993年 6月 3日,初版,pp.61−88
又吉光邦,「Tree−染色体構造をもったGAでの関数同定」,電子情報通信学会論文誌,日本,社団法人電子情報通信学会,1999年11月25日,Vol.J82−D−I,No.11,pp.1327−1335
吉本富士市・他,「実数を遺伝子とした遺伝的アルゴリズムによるデータあてはめ」,情報処理学会論文誌,日本,社団法人情報処理学会,2000年 1月15日,Vol.41,No.1,pp.70−82
小野功・他,「変数間の依存関係を考慮した正規分布交叉の提案と性能評価」,1996年度人工知能学会全国大会(第10回)論文集,日本,社団法人人工知能学会,1996年12月 9日,pp.609−612,JST資料番号:X0580A、前記日付はJST受入日
小野功・他,「実数値GAとその応用」,人工知能学会誌,日本,社団法人人工知能学会,2001年 3月 1日,Vol.15,No.2,pp.259−266
村川正宏・他,「近傍モデル遺伝的アルゴリズムによる多目的最適化」,情報処理学会論文誌,日本,社団法人情報処理学会,1999年 4月15日,Vol.40,No.4,pp.1792−1800