| 摘要 |
PROBLEM TO BE SOLVED: To efficiently calculate optimization of a variety of evaluating functions including multi-peak functions. SOLUTION: When setting initial values, a processing part reads the initial values including an initial solution x0 , an initial temperature T0 , a number of areas r, a count value c0 and a step reduction rate L from a storage part. When creating an area set, the processing part acquires a set H of areas hi according to the initial values read from the storage part. In a loop in the areas hi , the processing part executes random walk retrieval up to the count value c0 by the areas hi according to the initial values (symbol 10). In an evaluation (symbol 11), the processing part acquires a resolution yi optimizing the evaluating functions from predetermined whole areas hi , to store the resolution as yk into the storage part. When applying an annealing method (symbol 12), the processing part determines a next retrieval point xk+1 based on a metropolis standard according to a temperature Tk determined by the initial values. When updating the temperature Tk (symbol 13), the processing part updates the temperature Tk to a temperature Tk+1 . When repeating decisions, the processing part repeats the above described processing to a resolution xk+1 according to the updated temperature Tk+1 and performs a convergence decision.
|
