source: branches/uq/packages/optimizer/src/pgapack/pgapack/examples/c/sampleOptimization.c @ 5679

/*
 * Ganesh Hegde: GENES Learning
 * This is a sample optimization written for simple parameter values
 * The aim is to fit a given I-V for a MOSFET
 * The variables that are entering into the optimization problem are Vgs - Gate to Source Voltage and W/L-
 * The Width of the oxide layer divided by its length
 * using given Vgs -- 2V and W/L == 10, an I-V is obtained and fet as the function to be optimized
 * 
 */
 
#include <pgapack.h>
#include <stdlib.h>

double evaluationFunction(PGAContext *, int, int);


int main(int argc, char **argv){
        PGAContext *ctx;
        int strLen;
        double low[]={1.5,5} ,high[]={5,12.0};

    ctx = PGACreate(&argc, argv, PGA_DATATYPE_REAL, 2, PGA_MINIMIZE); 
    PGASetRealInitRange(ctx,low,high);
    PGASetPopSize(ctx,1000);
    PGASetCrossoverType(ctx,PGA_CROSSOVER_ONEPT);
    PGASetRandomSeed(ctx, 1);
    PGASetStoppingRuleType(ctx, PGA_STOP_NOCHANGE);
    PGASetUp(ctx);
    PGARun(ctx, evaluationFunction);
    PGADestroy(ctx);
        return 0;
}

/*
 * Sample Evaluation function for evaluating the parameters returned by the PGARun function on one cycle
 * ctx--> context variable
 * index--> chromosome index in the population
 * populationNumber --> Which population to look at
 */

double evaluationFunction(PGAContext *ctx, int index, int populationNumber){
        int i, stringLen;
        double realAllele,vGS=0.0,widthToLengthRatio=0.0;
        double Vds[21]={0}, Ids[21]={0};
        double mu = 300 ;/*In cm^2/V-s*/
        int Vt=1,j=0;
        double tox= 20e-7;
        double Cox= (3.9*8.85e-14)/tox;
        double k,slope,targetslope,fitness=0;
        double targetIds[] = {0,0.0005,0.0010,0.0016,0.0021,0.0026,0.0031,0.0036,0.0041,0.0047,0.0052,0.0057,0.0062,0.0067,0.0072,0.0078,0.0083,0.0088,0.0093,0.0098,0.0104};
        stringLen = PGAGetStringLength(ctx);
        for(i=0;i<stringLen;i++){
                realAllele = PGAGetRealAllele(ctx, index, populationNumber, i);
                if(i==0){
                        vGS=realAllele;
                }else{
                        widthToLengthRatio = realAllele;
                }
                
        }
        /*
         * Initing the Vds array
         */
        k=mu*Cox*widthToLengthRatio;
        
        for(i=0;i<21;i++){
                if(i>0){
                Vds[i]=Vds[i-1]+0.5;
                }else{
                        Vds[i]=0;
                }               
                Ids[i]=k*((vGS-Vt)*Vds[i]);
        }
        slope = (Ids[20]-Ids[0])/(Vds[20]-Vds[0]);
        targetslope = (targetIds[20]-targetIds[0])/(Vds[20]-Vds[0]);
        for(j=0;j<21;j++){
//              printf("%lf ",Ids[j]);
        }
        return (double)abs(slope-targetslope);
        
}