/* * sGA.java version 1.00 8 August 1996 * version 1.01 22 August 1996 -- split into packages * version 1.02 9 May 1997 -- JDK 1.1 mods * version 1.03 20 June 1997 -- removed last deprecation * * A simple genetic algorithm where the fitness function takes * non-negative values only: generate initial population, then for each * generation select a new population, apply crossover and mutation, until * the maximum number of generations is exceeded or, if known, the best * fitness is attained. Make sure the most fit member survives to the * next generation (elitism). * This Java code was derived from the C code in the Appendix of "Genetic * Algorithms + Data Structures = Evolution Programs," by Zbigniew * Michalewicz, Second Extended Edition, New York: Springer-Verlag (1994). * Other ideas and code were drawn from AGAC by Bill Spears (12 June 1991). * Available by anonymous ftp from ftp.aic.nrl.navy.mil in file GAC.shar.Z * in directory /pub/galist/src. * * (C) 1996 Stephen J. Hartley. All rights reserved. * Permission to use, copy, modify, and distribute this software for * non-commercial uses is hereby granted provided this notice is kept * intact within the source file. * * mailto:shartley@mcs.drexel.edu http://www.mcs.drexel.edu/~shartley * Drexel University, Math and Computer Science Department * Philadelphia, PA 19104 USA telephone: +1-215-895-2678 */ import java.io.*; import Utilities.*; import GAutilities.*; import Chromosomes.*; import Selections.*; import Crossovers.*; class sGA { // give fundamental parameters controlling the GA default values private static int populationSize = 100; private static int numXoverPoints = 1; private static boolean doElitism = false; private static double crossoverRate = .7; private static double mutationRate = .001; // these parameters control the running of the program private static int printPerGens = 100; // print every this often private static int maxGenerations = 1000; // quit when reached private static String logFileName = null; // record the time the program started private static final long startTime = System.currentTimeMillis(); private static long age() { return System.currentTimeMillis() - startTime; } public static void main(String[] args) { // process command line arguments, overriding GA parameter defaults GetOpt go = new GetOpt(args, "Udp:x:c:m:P:G:EF:"); go.optErr = true; String usage = "Usage: -d -p populationSize -x numXoverPoints\n" + " -E -c crossoverRate -m mutationRate\n" + " -P printPerGens -G maxGenerations -F logFileName"; int ch = -1; while ((ch = go.getopt()) != go.optEOF) { if ((char)ch == 'U') { System.err.println(usage); System.exit(0); } else if ((char)ch == 'd') Debug.flag = true; else if ((char)ch == 'p') populationSize = go.processArg(go.optArgGet(), populationSize); else if ((char)ch == 'x') numXoverPoints = go.processArg(go.optArgGet(), numXoverPoints); else if ((char)ch == 'c') crossoverRate = go.processArg(go.optArgGet(), crossoverRate); else if ((char)ch == 'm') mutationRate = go.processArg(go.optArgGet(), mutationRate); else if ((char)ch == 'E') doElitism = true; else if ((char)ch == 'P') printPerGens = go.processArg(go.optArgGet(), printPerGens); else if ((char)ch == 'G') maxGenerations = go.processArg(go.optArgGet(), maxGenerations); else if ((char)ch == 'F') logFileName = go.optArgGet(); else { // undefined option System.err.println(usage); System.exit(1); } } Defaults.logFileName = logFileName; Globals.stdout.println ("The GA parameters are:\n" + " populationSize = " + populationSize + "\n" + " numXoverPoints = " + numXoverPoints + "\n" + " crossoverRate = " + crossoverRate + "\n" + " mutationRate = " + mutationRate + "\n" + " doElitism = " + doElitism + "\n" + " printPerGens = " + printPerGens + "\n" + " maxGenerations = " + maxGenerations + "\n" + " Debug.flag = " + Debug.flag + "\n" + " logFileName = " + logFileName); // force loading class MyChromosome and setting static variables new MyChromosome(); Globals.stdout.println("GA: chromosome length = " + Chromosome.getChromosomeLength()); sGA theGA = new sGA(); theGA.mainLoop(); } private sGA() { super(); } // only GA.main() can create GA private void printChromosome(String name, Chromosome c) { Globals.stdout.println(name + ":\n " + c.toGenotype() + "\n " + c.toPhenotype() + "\n fitness= " + c.getFitness()); } private final Chromosome theBest = new MyChromosome(); private int theBestGeneration; // generation theBest first showed up // 0..popSize-1 holds population and theBest holds the most fit private Chromosome[] population = new Chromosome[populationSize]; private Chromosome[] newPopulation = new Chromosome[populationSize]; private int generationNum; // current generation number private int numCrossovers, numMutations; // tabulate for report() private Selection theSelection = null; private Crossover theCrossover = null; private void mainLoop() { Globals.stdout.println("GA: mainLoop"); if (Chromosome.isSolutionFitnessKnown()) { Globals.stdout.println("Known solution fitness is " + Chromosome.getSolutionFitness()); } initialize(); findTheBest(); report("Initial population"); while (!terminated()) { generationNum++; try { theSelection.select(population, newPopulation, populationSize); swapPopulationArrays(); } catch (FitnessSumZeroException e) { // replace current population with a new random one; // other possibilities are to do nothing and hope mutation // fixes the problem eventually if (Debug.flag) Globals.stdout.println(e + " (randomizing)"); for (int i = 0; i < populationSize; i++) { population[i].initializeChromosomeRandom(); } } catch (SelectionException e) { if (Debug.flag) Globals.stdout.println(e); } if (Debug.flag) report("Selection"); crossover(); if (Debug.flag) report("Crossover"); mutate(); if (Debug.flag) report("Mutation"); if (doElitism) elitism(); else justUpdateTheBest(); if (Debug.flag || (generationNum % printPerGens) == 0) { report("Report"); } } Globals.stdout.println("Simulation completed in " + generationNum + " generations and " + age()/1000.0 + " seconds"); printChromosome("Best member (generation="+theBestGeneration+")", theBest); System.exit(0); } private void initialize() { generationNum = 0; numCrossovers = 0; numMutations = 0; theSelection = new ProportionalSelection(); if (numXoverPoints == 0) theCrossover = new UniformCrossover(); else if (numXoverPoints == 1) theCrossover = new OnePointCrossover(); else theCrossover = new NPointCrossover(numXoverPoints); for (int j = 0; j < populationSize; j++) { population[j] = new MyChromosome(); population[j].initializeChromosomeRandom(); newPopulation[j] = new MyChromosome(); } if (Debug.flag) { for (int j=0; j currentBestFitness ) { currentBest = j; currentBestFitness = next; } } // once the best member in the population is found, copy the genes population[currentBest].copyChromosome(theBest); if (Debug.flag) { printChromosome("currentBest (generation="+theBestGeneration+")", theBest); } } private boolean terminated() { return (theBest.isSolutionFitnessKnown() && theBest.getFitness() == theBest.getSolutionFitness()) || generationNum >= maxGenerations; } private void swapPopulationArrays() { Chromosome[] temp = population; population = newPopulation; newPopulation = temp; } /***************************************************************/ /* Crossover selection: selects two parents that take part in */ /* the crossover. */ /***************************************************************/ private void crossover() { int one = -1; // compiler complains not being initialized int first = 0; // count of the number of members chosen for (int mem = 0; mem < populationSize; ++mem) { if (MyRandom.dblRandom() < crossoverRate) { ++first; if (first % 2 == 0) { numCrossovers++; if (Debug.flag) { Globals.stdout.println("crossing " + one + " and " + mem); } theCrossover.xOver(population[one], population[mem]); } else one = mem; } } } /**************************************************************/ /* Mutation: Random uniform mutation. A variable selected for */ /* mutation is replaced by a random value between lower and */ /* upper bounds of this variable */ /**************************************************************/ private void mutate() { int chromosomeLength = Chromosome.getChromosomeLength(); for (int i = 0; i < populationSize; i++) { for (int j = 0; j < chromosomeLength; j++) { if (MyRandom.dblRandom() < mutationRate) { numMutations++; population[i].mutateGene(j); if (Debug.flag) { printChromosome("mutation, i=" + i + ", gene=" + j, population[i]); } } } } } /***************************************************************/ /* Report function: Reports progress of the simulation. /***************************************************************/ private void report(String title) { double best_val; // best population fitness double avg; // avg population fitness double stddev; // std. deviation of population fitness double sum_square; // sum of square for std. calc double square_sum; // square of sum for std. calc double sum; // total population fitness double fitness; sum = 0.0; sum_square = 0.0; for (int i = 0; i < populationSize; i++) { fitness = population[i].getFitness(); sum += fitness; sum_square += fitness * fitness; } avg = sum/(double)populationSize; square_sum = sum * sum/(double)populationSize; stddev = Math.sqrt((1.0/(double)(populationSize - 1)) *(sum_square - square_sum)); best_val = theBest.getFitness(); Globals.stdout.println(title + ": generation=" + generationNum + " most fit=" + best_val + " avg=" + avg + " stddev=" + stddev); printChromosome("most fit (generation="+theBestGeneration+")", theBest); Globals.stdout.println("number of crossovers and mutations: " + numCrossovers + " and " + numMutations); if (Debug.flag) { for (int j=0; j population[i+1].getFitness()) { if (population[i].getFitness() > best) { best = population[i].getFitness(); bestMember = i; } if (population[i+1].getFitness() < worst) { worst = population[i+1].getFitness(); worstMember = i + 1; } } else { if (population[i].getFitness() < worst) { worst = population[i].getFitness(); worstMember = i; } if (population[i+1].getFitness() > best) { best = population[i+1].getFitness(); bestMember = i + 1; } } } // if best individual from the new population is better than // the best individual from the previous population, then // copy the best from the new population; else replace the // worst individual from the current population with the // best one from the previous generation if (best > theBest.getFitness()) { population[bestMember].copyChromosome(theBest); theBestGeneration = generationNum; } else { theBest.copyChromosome(population[worstMember]); } if (Debug.flag) { printChromosome("elitism, best (index="+bestMember+")", population[bestMember]); printChromosome("elitism, worst (index="+worstMember+")", population[worstMember]); } } private void justUpdateTheBest() { double currentBestFitness = population[0].getFitness(); double next = -1; int currentBest = 0; // index of the current best individual for (int j = 1; j < populationSize; j++) { if ((next = population[j].getFitness()) > currentBestFitness ) { currentBest = j; currentBestFitness = next; } } if (currentBestFitness > theBest.getFitness()) { population[currentBest].copyChromosome(theBest); theBestGeneration = generationNum; } if (Debug.flag) { printChromosome("theBest (generation="+theBestGeneration+")", theBest); } } } /* ............... Example compile and run(s) D:\>javac sGA.java D:\>java sGA -p100 -m0.01 -P10 -G100 -F bitCount.output D:\>type bitCount.output The GA parameters are: populationSize = 100 numXoverPoints = 1 crossoverRate = 0.7 mutationRate = 0.01 doElitism = false printPerGens = 10 maxGenerations = 100 Debug.flag = false logFileName = bitCount.output BitCountChromosome: chromosome length is 32 GA: chromosome length = 32 GA: mainLoop Known solution fitness is 32 Initial population: generation=0 most fit=23 avg=16.72 stddev=2.77099 most fit (generation=0): 01111110111111011110101000111011 this chromosome has 23 bits set fitness= 23 number of crossovers and mutations: 0 and 0 Report: generation=10 most fit=26 avg=19.17 stddev=2.39129 most fit (generation=10): 10111111110111011111100111111101 this chromosome has 26 bits set fitness= 26 number of crossovers and mutations: 341 and 304 Report: generation=20 most fit=27 avg=21.41 stddev=2.31855 most fit (generation=18): 01111111111111011111001111111101 this chromosome has 27 bits set fitness= 27 number of crossovers and mutations: 690 and 633 Report: generation=30 most fit=28 avg=21 stddev=2.3741 most fit (generation=26): 11111111111111011111101111110110 this chromosome has 28 bits set fitness= 28 number of crossovers and mutations: 1033 and 954 Report: generation=40 most fit=29 avg=23.01 stddev=2.32464 most fit (generation=34): 11111111111111111101011111110111 this chromosome has 29 bits set fitness= 29 number of crossovers and mutations: 1384 and 1269 Report: generation=50 most fit=30 avg=23.67 stddev=2.01537 most fit (generation=45): 11111111111111111111011011111111 this chromosome has 30 bits set fitness= 30 number of crossovers and mutations: 1736 and 1620 Report: generation=60 most fit=30 avg=23.53 stddev=2.10077 most fit (generation=45): 11111111111111111111011011111111 this chromosome has 30 bits set fitness= 30 number of crossovers and mutations: 2081 and 1919 Report: generation=70 most fit=30 avg=23.57 stddev=1.83267 most fit (generation=45): 11111111111111111111011011111111 this chromosome has 30 bits set fitness= 30 number of crossovers and mutations: 2437 and 2231 Report: generation=80 most fit=30 avg=24.24 stddev=1.99555 most fit (generation=45): 11111111111111111111011011111111 this chromosome has 30 bits set fitness= 30 number of crossovers and mutations: 2789 and 2520 Report: generation=90 most fit=30 avg=24.59 stddev=2.59796 most fit (generation=45): 11111111111111111111011011111111 this chromosome has 30 bits set fitness= 30 number of crossovers and mutations: 3142 and 2873 Report: generation=100 most fit=30 avg=24.94 stddev=2.0192 most fit (generation=45): 11111111111111111111011011111111 this chromosome has 30 bits set fitness= 30 number of crossovers and mutations: 3481 and 3169 Simulation completed in 100 generations and 26.58 seconds Best member (generation=45): 11111111111111111111011011111111 this chromosome has 30 bits set fitness= 30 D:\>edit sGA.java D:\>javac sGA.java D:\>java sGA -E -p100 -m0.01 -P10 -G100 The GA parameters are: populationSize = 100 numXoverPoints = 1 crossoverRate = 0.7 mutationRate = 0.01 doElitism = true printPerGens = 10 maxGenerations = 100 Debug.flag = false logFileName = null XtoTenthChromosome: chromosome length is 32 GA: chromosome length = 32 GA: mainLoop Known solution fitness is 1 Initial population: generation=0 most fit=0.919291 avg=0.122947 stddev=0.216728 most fit (generation=0): 01100010000010110101101110111111 x=0.99162 fitness= 0.919291 number of crossovers and mutations: 0 and 0 Report: generation=10 most fit=0.999597 avg=0.827711 stddev=0.206224 most fit (generation=9): 01111001110110101011111111111111 x=0.99996 fitness= 0.999597 number of crossovers and mutations: 353 and 320 Report: generation=20 most fit=0.999599 avg=0.870255 stddev=0.196344 most fit (generation=16): 01001111111110101011111111111111 x=0.99996 fitness= 0.999599 number of crossovers and mutations: 694 and 632 Report: generation=30 most fit=0.999997 avg=0.891283 stddev=0.159036 most fit (generation=29): 10110111010111111111111111111111 x=1 fitness= 0.999997 number of crossovers and mutations: 1039 and 990 Report: generation=40 most fit=1 avg=0.894416 stddev=0.183714 most fit (generation=40): 00000111111111111111111111111111 x=1 fitness= 1 number of crossovers and mutations: 1380 and 1368 Report: generation=50 most fit=1 avg=0.907259 stddev=0.169297 most fit (generation=48): 00110111111111111111111111111111 x=1 fitness= 1 number of crossovers and mutations: 1728 and 1691 Report: generation=60 most fit=1 avg=0.941364 stddev=0.142952 most fit (generation=48): 00110111111111111111111111111111 x=1 fitness= 1 number of crossovers and mutations: 2083 and 2017 Report: generation=70 most fit=1 avg=0.92843 stddev=0.140908 most fit (generation=67): 10111111111111111111111111111111 x=1 fitness= 1 number of crossovers and mutations: 2432 and 2328 Report: generation=80 most fit=1 avg=0.918706 stddev=0.162607 most fit (generation=67): 10111111111111111111111111111111 x=1 fitness= 1 number of crossovers and mutations: 2776 and 2666 Report: generation=90 most fit=1 avg=0.927445 stddev=0.169067 most fit (generation=67): 10111111111111111111111111111111 x=1 fitness= 1 number of crossovers and mutations: 3137 and 2978 Report: generation=100 most fit=1 avg=0.926531 stddev=0.184471 most fit (generation=67): 10111111111111111111111111111111 x=1 fitness= 1 number of crossovers and mutations: 3481 and 3287 Simulation completed in 100 generations and 27.74 seconds Best member (generation=67): 10111111111111111111111111111111 x=1 fitness= 1 D:\>edit sGA.java D:\>javac sGA.java D:\>java sGA -p100 -m0.1 -P10 -G100 The GA parameters are: populationSize = 100 numXoverPoints = 1 crossoverRate = 0.7 mutationRate = 0.1 doElitism = false printPerGens = 10 maxGenerations = 100 Debug.flag = false logFileName = null SinesChromosome: chromosome length is 3 GA: chromosome length = 3 GA: mainLoop Initial population: generation=0 most fit=36.8932 avg=21.3019 stddev=5.27425 most fit (generation=0): gene[0]=9.62878 gene[1]=5.21988 gene[2]=0.823022 21.5+9.62878*sin(4*PI*9.62878)+5.21988*sin(20*PI*5.21988)+0.823022 fitness= 36.8932 number of crossovers and mutations: 0 and 0 Report: generation=10 most fit=37.4763 avg=28.3356 stddev=5.45076 most fit (generation=8): gene[0]=11.6685 gene[1]=5.12788 gene[2]=0.963809 21.5+11.6685*sin(4*PI*11.6685)+5.12788*sin(20*PI*5.12788)+0.963809 fitness= 37.4763 number of crossovers and mutations: 346 and 317 Report: generation=20 most fit=37.4763 avg=29.5948 stddev=4.84377 most fit (generation=8): gene[0]=11.6685 gene[1]=5.12788 gene[2]=0.963809 21.5+11.6685*sin(4*PI*11.6685)+5.12788*sin(20*PI*5.12788)+0.963809 fitness= 37.4763 number of crossovers and mutations: 684 and 603 Report: generation=30 most fit=37.4763 avg=28.6652 stddev=4.5108 most fit (generation=8): gene[0]=11.6685 gene[1]=5.12788 gene[2]=0.963809 21.5+11.6685*sin(4*PI*11.6685)+5.12788*sin(20*PI*5.12788)+0.963809 fitness= 37.4763 number of crossovers and mutations: 1015 and 934 Report: generation=40 most fit=38.1848 avg=29.4046 stddev=6.39036 most fit (generation=39): gene[0]=11.6621 gene[1]=5.32681 gene[2]=0.973592 21.5+11.6621*sin(4*PI*11.6621)+5.32681*sin(20*PI*5.32681)+0.973592 fitness= 38.1848 number of crossovers and mutations: 1384 and 1251 Report: generation=50 most fit=38.2355 avg=31.7354 stddev=6.47591 most fit (generation=44): gene[0]=11.1241 gene[1]=5.12562 gene[2]=0.49033 21.5+11.1241*sin(4*PI*11.1241)+5.12562*sin(20*PI*5.12562)+0.49033 fitness= 38.2355 number of crossovers and mutations: 1740 and 1555 Report: generation=60 most fit=38.2355 avg=30.5727 stddev=5.91608 most fit (generation=44): gene[0]=11.1241 gene[1]=5.12562 gene[2]=0.49033 21.5+11.1241*sin(4*PI*11.1241)+5.12562*sin(20*PI*5.12562)+0.49033 fitness= 38.2355 number of crossovers and mutations: 2098 and 1851 Report: generation=70 most fit=38.4794 avg=32.0127 stddev=5.7576 most fit (generation=63): gene[0]=11.6621 gene[1]=5.62342 gene[2]=0.964925 21.5+11.6621*sin(4*PI*11.6621)+5.62342*sin(20*PI*5.62342)+0.964925 fitness= 38.4794 number of crossovers and mutations: 2437 and 2124 Report: generation=80 most fit=38.621 avg=33.0924 stddev=5.99235 most fit (generation=75): gene[0]=11.6621 gene[1]=5.72433 gene[2]=0.982909 21.5+11.6621*sin(4*PI*11.6621)+5.72433*sin(20*PI*5.72433)+0.982909 fitness= 38.621 number of crossovers and mutations: 2776 and 2406 Report: generation=90 most fit=39.0127 avg=32.0723 stddev=6.0964 most fit (generation=82): gene[0]=12.0983 gene[1]=5.32681 gene[2]=0.794961 21.5+12.0983*sin(4*PI*12.0983)+5.32681*sin(20*PI*5.32681)+0.794961 fitness= 39.0127 number of crossovers and mutations: 3120 and 2740 Report: generation=100 most fit=39.0127 avg=32.7011 stddev=6.00134 most fit (generation=82): gene[0]=12.0983 gene[1]=5.32681 gene[2]=0.794961 21.5+12.0983*sin(4*PI*12.0983)+5.32681*sin(20*PI*5.32681)+0.794961 fitness= 39.0127 number of crossovers and mutations: 3478 and 3029 Simulation completed in 100 generations and 9.77 seconds Best member (generation=82): gene[0]=12.0983 gene[1]=5.32681 gene[2]=0.794961 21.5+12.0983*sin(4*PI*12.0983)+5.32681*sin(20*PI*5.32681)+0.794961 fitness= 39.0127 D:\>edit sGA.java D:\>javac sGA.java D:\>java sGA -E -p100 -m0.01 -P10 -G100 The GA parameters are: populationSize = 100 numXoverPoints = 1 crossoverRate = 0.7 mutationRate = 0.01 doElitism = true printPerGens = 10 maxGenerations = 100 Debug.flag = false logFileName = null ByteCountChromosome: chromosome length is 64 GA: chromosome length = 64 GA: mainLoop Known solution fitness is 64 Initial population: generation=0 most fit=8 avg=0.48 stddev=1.90947 most fit (generation=0): 1101001111111111011101001100100100110011101001010001010000000001 this chromosome has 1 bytes set fitness= 8 number of crossovers and mutations: 0 and 0 Report: generation=10 most fit=24 avg=14.48 stddev=4.35885 most fit (generation=7): 1111111111111111000001101100100111110000011101100011110011111111 this chromosome has 3 bytes set fitness= 24 number of crossovers and mutations: 343 and 593 Report: generation=20 most fit=24 avg=17.6 stddev=5.57048 most fit (generation=7): 1111111111111111000001101100100111110000011101100011110011111111 this chromosome has 3 bytes set fitness= 24 number of crossovers and mutations: 696 and 1270 Report: generation=30 most fit=24 avg=20.64 stddev=4.9798 most fit (generation=7): 1111111111111111000001101100100111110000011101100011110011111111 this chromosome has 3 bytes set fitness= 24 number of crossovers and mutations: 1056 and 1926 Report: generation=40 most fit=24 avg=17.52 stddev=5.53498 most fit (generation=7): 1111111111111111000001101100100111110000011101100011110011111111 this chromosome has 3 bytes set fitness= 24 number of crossovers and mutations: 1414 and 2590 Report: generation=50 most fit=24 avg=18.56 stddev=5.55654 most fit (generation=7): 1111111111111111000001101100100111110000011101100011110011111111 this chromosome has 3 bytes set fitness= 24 number of crossovers and mutations: 1753 and 3229 Report: generation=60 most fit=32 avg=21.52 stddev=6.09103 most fit (generation=51): 1111111111111111010011111111111111110011011101011000110111111111 this chromosome has 4 bytes set fitness= 32 number of crossovers and mutations: 2092 and 3844 Report: generation=70 most fit=32 avg=23.6 stddev=5.72783 most fit (generation=51): 1111111111111111010011111111111111110011011101011000110111111111 this chromosome has 4 bytes set fitness= 32 number of crossovers and mutations: 2449 and 4511 Report: generation=80 most fit=32 avg=23.44 stddev=6.84654 most fit (generation=51): 1111111111111111010011111111111111110011011101011000110111111111 this chromosome has 4 bytes set fitness= 32 number of crossovers and mutations: 2789 and 5156 Report: generation=90 most fit=32 avg=23.52 stddev=5.66485 most fit (generation=51): 1111111111111111010011111111111111110011011101011000110111111111 this chromosome has 4 bytes set fitness= 32 number of crossovers and mutations: 3147 and 5815 Report: generation=100 most fit=40 avg=26.72 stddev=7.02963 most fit (generation=92): 1111111111111111001011101111111111110011111101011111111111111111 this chromosome has 5 bytes set fitness= 40 number of crossovers and mutations: 3496 and 6469 Simulation completed in 100 generations and 46.13 seconds Best member (generation=92): 1111111111111111001011101111111111110011111101011111111111111111 this chromosome has 5 bytes set fitness= 40 ... end of example run(s) */