AcceptIVMetropolis {xegaPopulation}

Individually Adaptive Metropolis Acceptance Rule.

Description

Change the gene by a genetic operator pipeline. Always accept new genes with fitness improvement. For maximizing fitness accept genes with lower fitness with probability (runif(1)<exp(-(fitness-newfitness)*beta/Temperature) and reduce temperature with a cooling schedule. For each gene, the temperature is corrected upward by a term whose size is proportional to the difference between the fitness of the current best gene in the population and the fitness of the gene.

Usage

AcceptIVMetropolis(OperatorPipeline, gene, lF)

Arguments

OperatorPipeline

Genetic operator pipeline.

gene

Gene.

lF

Local configuration.

Details

The temperature is updated at the end of each generation in the main loop of the genetic algorithm.

Value

The new gene if it has at least equal performance as the old gene else the old gene.

References

Locatelli, M. (2000): Convergence of a Simulated Annealing Algorithm for Continuous Global Optimization. Journal of Global Optimization, 18:219-233. <doi:10.1023/A:1008339019740>

The-Crankshaft Publishing (2023): A Comparison of Cooling Schedules for Simulated Annealing. <URL:https://what-when-how.com/artificial-intelligence/a-comparison-of-cooling-schedules-for-simulated-annealing-artificial-intelligence/>

See Also

Other Acceptance Rule: AcceptBest(), AcceptMetropolis(), AcceptNewGene()

Examples

parm<-function(x){function() {return(x)}}
lFxegaGaGene$Beta<-parm(1)
lFxegaGaGene$TempK<-parm(10)
set.seed(2)
OPpipe1<-function(g, lF){InitGene(lF)}
g1<-lFxegaGaGene$EvalGene(InitGene(lFxegaGaGene), lFxegaGaGene)
lFxegaGaGene$CBestFitness<-parm(g1$fit)
g2<-AcceptMetropolis(OPpipe1, g1, lFxegaGaGene)