Genetic Algorithm Example from Igor Book
Wed, 06/13/2018 - 08:45 am
I post this example here, because several people wrote to me that it is a bit lengthy and that it takes a long time to copy it from the book. So, here it is.... Because this example is implemented as a module, you have to copy it to a new, separate procedure window. If you try to use it in Igor's standard procedure window, the #pragma moduleName statement will give you trouble. Try various combinations of mutation rate and population size and see how long it takes until the genetic routine has solved the puzzle. The first argument gives the maximum number of generations after which the program terminates. Compare for example:
to explore how a deviation from the standard values changes the ability of the algorithm to solve the puzzle and find the secret word. More information can be found here.
#pragma moduleName = FindSecret //---- //--There are other (better) methods for solving the task of //--getting a 'secret' word. The genetic approach //--is, however, very instructive from a //--programmer's point of view. That is why it is done here. static function main(generations,[mutationProb,individuals]) variable generations variable mutationProb variable individuals mutationProb = (ParamIsDefault(mutationProb) ? 0.05 : mutationProb) individuals = (ParamIsDefault(individuals) ? 1000 : individuals) string secret = "Santa Claus is coming!" variable lengthSecret = strlen(secret) variable i,k,j variable parentIndex string gene ="" string parentGenes = "" // make a two-dimensional text-wave to store the "genepool" // the first column stores a string with the individual's genes, // the second column stores the number of correct characters // "Santa Clerx" "8" // if the number of correct characters == lengthSecret // the solution was found! print "----------------------------------- Starting search!" print " " //set up a genepool as a free (transient) 2D-textwave make /FREE /T /N=(individuals,2) GenePool wave /T gP = GenePool // // make the first generation // for (i=0; i<individuals; i++) gP[i]="" gP[i]="0" for (k=0; k<lengthSecret; k++) //get a random number between 30 and 127 //and interpret it as ASCII code by using num2char //(0.5 + enoise(0.5)) generates a random number between 0 and 1 //use floor for rounding to integer values gene = num2char(floor(30+97*(0.5+enoise(0.5)))) gP[i] += gene //test if this newest gene is correct, if yes: increase score by one if(char2num(gene) == char2num(secret[k])) gP[i] = num2str( str2num( gP[i] ) + 1 ) endif endfor endfor SortColumns /A /R /KNDX=1 sortWaves=gP //sort by score, i.e., sort by number of correct characters //without '/A' 1,13,9 is sorted as 9,13,1 //(sorting by the first character) // //now do the further generations // for (j=0; j<generations; j++) //leave the best 10% of the previous generation untouched //start the for loop at a higher i for (i=floor(0.1*individuals); i<individuals; i++) gP[i]="" gP[i]="0" for (k=0; k<lengthSecret; k++) //allow a mutation with a certain probability if ( (0.5+enoise(0.5)) < mutationProb) //draw again a random ASCII character gene = num2char(floor(30+97*(0.5+enoise(0.5)))) gP[i] += gene if (GrepString(gene,secret[k])) gP[i] = num2str( str2num( gP[i] ) + 1 ) endif else //no mutation //pick a random individual from the best 10% parentIndex = floor(0.1*individuals*(0.5+enoise(0.5))) parentGenes = gP[parentIndex] gP[i] += parentGenes[k] if (char2num(parentGenes[k])==char2num(secret[k])) gP[i] = num2str(str2num(gP[i])+1) endif endif //end mutation-choice endfor //end loop over individual letters (genes) endfor //end loop over individuals (rows in genePool) SortColumns /A /R /KNDX=1 sortWaves=gP print "Generation",j+1,":", gP if (str2num(gP)==lengthSecret) print " " print "Puzzle solved! " break endif endfor //end loop over further generations end
Igor Pro 8
Igor XOP Toolkit
Igor NIDAQ Tools MX