| 1 |
"""
|
| 2 |
A simple Genetic Algorithm in Python
|
| 3 |
"""
|
| 4 |
|
| 5 |
__author__ = "Douglas Blank <dblank@brynmawr.edu>"
|
| 6 |
__version__ = "$Revision$"
|
| 7 |
|
| 8 |
import Numeric, math, random, time, sys, string
|
| 9 |
from copy import deepcopy
|
| 10 |
|
| 11 |
def display(v):
|
| 12 |
print v,
|
| 13 |
|
| 14 |
def sum(a):
|
| 15 |
sum = 0
|
| 16 |
for n in a:
|
| 17 |
sum += n
|
| 18 |
return sum
|
| 19 |
|
| 20 |
def flip(probability):
|
| 21 |
"""
|
| 22 |
Flip a biased coin
|
| 23 |
"""
|
| 24 |
return random.random() <= probability
|
| 25 |
|
| 26 |
|
| 27 |
# The classes:
|
| 28 |
# Gene - specifics of gene representation
|
| 29 |
# Population - collection of Genes
|
| 30 |
# GA - the parameters of evolution
|
| 31 |
|
| 32 |
class Gene:
|
| 33 |
def __init__(self, **args):
|
| 34 |
self.verbose = 0
|
| 35 |
self.genotype = []
|
| 36 |
self.fitness = 0.0
|
| 37 |
self.mode = 'float'
|
| 38 |
self.crossoverPoints = 1
|
| 39 |
self.bias = 0.5
|
| 40 |
self.min = -1 # inclusive
|
| 41 |
self.max = 1 # inclusive
|
| 42 |
self.imin = -1 # inclusive, initial
|
| 43 |
self.imax = 1 # inclusive, initial
|
| 44 |
self.maxStep = 1
|
| 45 |
self.args = args
|
| 46 |
self.alphabet = "abcdefghijklmnopqrstuvwxyz "
|
| 47 |
if args.has_key('verbose'):
|
| 48 |
self.verbose = args['verbose']
|
| 49 |
if args.has_key('min'):
|
| 50 |
self.min = args['min']
|
| 51 |
self.imin = args['min']
|
| 52 |
if args.has_key('max'):
|
| 53 |
self.max = args['max']
|
| 54 |
self.imax = args['max']
|
| 55 |
if args.has_key('imin'):
|
| 56 |
self.imin = args['imin']
|
| 57 |
if args.has_key('imax'):
|
| 58 |
self.imax = args['imax']
|
| 59 |
if args.has_key('maxStep'):
|
| 60 |
self.maxStep = args['maxStep']
|
| 61 |
if args.has_key('crossoverPoints'):
|
| 62 |
self.crossoverPoints = args['crossoverPoints']
|
| 63 |
if args.has_key('mode'):
|
| 64 |
self.mode = args['mode']
|
| 65 |
if args.has_key('bias'):
|
| 66 |
self.bias = args['bias']
|
| 67 |
for i in range(args['size']):
|
| 68 |
if self.mode == 'bit':
|
| 69 |
self.genotype.append( random.random() < self.bias)
|
| 70 |
elif self.mode == 'integer':
|
| 71 |
self.genotype.append( math.floor(random.random() *
|
| 72 |
(self.imax - self.imin + 1)) + self.imin)
|
| 73 |
elif self.mode == 'float':
|
| 74 |
self.genotype.append( (random.random() * (self.imax - self.imin)) + self.imin)
|
| 75 |
elif self.mode == 'char':
|
| 76 |
self.genotype.append( self.alphabet[int(random.random() * len(self.alphabet)) ] )
|
| 77 |
else:
|
| 78 |
raise "unknownMode", self.mode
|
| 79 |
|
| 80 |
def copy(self):
|
| 81 |
return deepcopy(self)
|
| 82 |
|
| 83 |
def __getitem__(self, val):
|
| 84 |
return self.genotype[val]
|
| 85 |
|
| 86 |
def __len__(self):
|
| 87 |
return len(self.genotype)
|
| 88 |
|
| 89 |
def display(self):
|
| 90 |
if self.mode == 'bit' or self.mode == 'integer':
|
| 91 |
print string.join(map(lambda v: `int(v)`, self.genotype), "")
|
| 92 |
elif self.mode == 'float':
|
| 93 |
map(lambda v: display("%3.2f" % v), self.genotype)
|
| 94 |
elif self.mode == 'char':
|
| 95 |
print string.join(self.genotype, '')
|
| 96 |
else:
|
| 97 |
raise "unknownMode", self.mode
|
| 98 |
|
| 99 |
def mutate(self, mutationRate):
|
| 100 |
"""
|
| 101 |
Depending on the mutationRate, will mutate the genotype.
|
| 102 |
"""
|
| 103 |
for i in range(len(self.genotype)):
|
| 104 |
if flip(mutationRate):
|
| 105 |
if self.verbose > 2:
|
| 106 |
print "mutating at position", i
|
| 107 |
if self.mode == 'bit':
|
| 108 |
self.genotype[i] = not self.genotype[i]
|
| 109 |
elif self.mode == 'integer':
|
| 110 |
r = random.random()
|
| 111 |
if (r < .33):
|
| 112 |
self.genotype[i] += round(random.random() * self.maxStep)
|
| 113 |
self.genotype[i] = min(self.genotype[i], self.max)
|
| 114 |
elif (r < .67):
|
| 115 |
self.genotype[i] -= round(random.random() * self.maxStep)
|
| 116 |
self.genotype[i] = max(self.genotype[i], self.min)
|
| 117 |
else:
|
| 118 |
self.genotype[i] = round(random.random() * (self.max - self.min + 1)) + self.min
|
| 119 |
elif self.mode == 'float':
|
| 120 |
r = random.random()
|
| 121 |
if (r < .33):
|
| 122 |
self.genotype[i] += (random.random() * self.maxStep)
|
| 123 |
self.genotype[i] = min(self.genotype[i], self.max)
|
| 124 |
elif (r < .67):
|
| 125 |
self.genotype[i] -= (random.random() * self.maxStep)
|
| 126 |
self.genotype[i] = max(self.genotype[i], self.min)
|
| 127 |
else:
|
| 128 |
self.genotype[i] = (random.random() * (self.max - self.min)) + self.min
|
| 129 |
elif self.mode == 'char':
|
| 130 |
self.genotype[i] = self.alphabet[ int(random.random() * len(self.alphabet)) ]
|
| 131 |
else:
|
| 132 |
raise "unknownMode", self.mode
|
| 133 |
|
| 134 |
def crossover(self, parent2, crossoverRate):
|
| 135 |
"""
|
| 136 |
Depending on the crossoverRate, will return two new children
|
| 137 |
created by crossing over the given parents at a single point,
|
| 138 |
or will return copies of the parents.
|
| 139 |
"""
|
| 140 |
parent1 = self
|
| 141 |
geneLength = len(parent1.genotype)
|
| 142 |
if flip(crossoverRate):
|
| 143 |
p1 = parent1.genotype[:]
|
| 144 |
p2 = parent2.genotype[:]
|
| 145 |
child1 = [0] * geneLength
|
| 146 |
child2 = [0] * geneLength
|
| 147 |
# go through and pick the crossoverpoints:
|
| 148 |
if self.crossoverPoints == -3:
|
| 149 |
# one right in middle
|
| 150 |
crossPoints = [0] * geneLength
|
| 151 |
crossPoints[int(geneLength/2)] = 1
|
| 152 |
elif self.crossoverPoints == -2:
|
| 153 |
# no crossoverpoints; I know, this should be zero
|
| 154 |
crossPoints = [0] * geneLength
|
| 155 |
elif self.crossoverPoints == -1:
|
| 156 |
# shuffle: every other one
|
| 157 |
crossPoints = [1] * geneLength
|
| 158 |
elif self.crossoverPoints > 0:
|
| 159 |
# number of cross points:
|
| 160 |
# NOTE: not guaranteed to be exactly that many;
|
| 161 |
# could duplicate randomly
|
| 162 |
crossPoints = [0] * geneLength
|
| 163 |
for i in range(self.crossoverPoints):
|
| 164 |
crossPoints[(int)(random.random() * geneLength)] = 1
|
| 165 |
elif self.crossoverPoints == 0:
|
| 166 |
# uniform crossover when crossoverPoints = 0
|
| 167 |
crossPoints = [0] * geneLength
|
| 168 |
for i in range(geneLength):
|
| 169 |
# flip coin for each position:
|
| 170 |
if random.random() < .5:
|
| 171 |
crossPoints[i] = 1
|
| 172 |
else:
|
| 173 |
raise "unknownCrossoverType", self.crossoverPoints
|
| 174 |
# now, each time there is a cross point, swap parents
|
| 175 |
for i in range(geneLength):
|
| 176 |
if crossPoints[i]:
|
| 177 |
if self.verbose > 2:
|
| 178 |
print "crossing over at point", i
|
| 179 |
p1, p2 = p2, p1
|
| 180 |
child1[i] = p1[i]
|
| 181 |
child2[i] = p2[i]
|
| 182 |
new_child1 = self.__class__(**self.args)
|
| 183 |
new_child2 = self.__class__(**self.args)
|
| 184 |
new_child1.genotype = child1
|
| 185 |
new_child2.genotype = child2
|
| 186 |
return new_child1, new_child2
|
| 187 |
else:
|
| 188 |
if self.verbose > 2:
|
| 189 |
print "no crossover"
|
| 190 |
return parent1.copy(), parent2.copy()
|
| 191 |
|
| 192 |
class Population:
|
| 193 |
def __init__(self, cnt, geneConstructor, **args):
|
| 194 |
self.sumFitness = 0
|
| 195 |
self.avgFitness = 0
|
| 196 |
self.individuals = []
|
| 197 |
self.eliteMembers = []
|
| 198 |
self.elitePercent = 0.0
|
| 199 |
self.bestMember = -1
|
| 200 |
self.size = cnt
|
| 201 |
self.verbose = 0
|
| 202 |
self.args = args
|
| 203 |
self.geneConstructor = geneConstructor
|
| 204 |
if args.has_key('elitePercent'):
|
| 205 |
self.elitePercent = args['elitePercent']
|
| 206 |
if args.has_key('verbose'):
|
| 207 |
self.verbose = args['verbose']
|
| 208 |
for i in range(cnt):
|
| 209 |
self.individuals.append(geneConstructor(pos = i,
|
| 210 |
popSize = cnt,
|
| 211 |
**args))
|
| 212 |
def copy(self):
|
| 213 |
newPop = self.__class__(0, self.geneConstructor, **self.args)
|
| 214 |
newPop.size = self.size
|
| 215 |
for i in range(self.size):
|
| 216 |
newPop.individuals.append( self.individuals[i].copy() )
|
| 217 |
return newPop
|
| 218 |
|
| 219 |
def __getitem__(self, val):
|
| 220 |
return self.individuals[val]
|
| 221 |
|
| 222 |
def __len__(self):
|
| 223 |
return len(self.individuals)
|
| 224 |
|
| 225 |
def select(self):
|
| 226 |
"""
|
| 227 |
Select a single individual via the roulette wheel method.
|
| 228 |
Algorithm from Goldberg's book, page 63. NOTE: fitness
|
| 229 |
function must return positive values to use this method
|
| 230 |
of selection.
|
| 231 |
"""
|
| 232 |
index = 0
|
| 233 |
partsum = 0.0
|
| 234 |
if self.sumFitness == 0:
|
| 235 |
raise "Population has a total of zero fitness"
|
| 236 |
spin = random.random() * self.sumFitness
|
| 237 |
while index < self.size-1:
|
| 238 |
fitness = self.individuals[index].fitness
|
| 239 |
if fitness < 0:
|
| 240 |
raise "Negative fitness in select", fitness
|
| 241 |
partsum += self.individuals[index].fitness
|
| 242 |
if partsum >= spin:
|
| 243 |
break
|
| 244 |
index += 1
|
| 245 |
if self.verbose > 2:
|
| 246 |
print "selected",
|
| 247 |
self.individuals[index].display(),
|
| 248 |
print "fitness", self.individuals[index].fitness
|
| 249 |
return self.individuals[index].copy()
|
| 250 |
|
| 251 |
def statistics(self):
|
| 252 |
"""
|
| 253 |
Maintains important statistics about the current population.
|
| 254 |
It calculates total fitness, average fitness, best fitness,
|
| 255 |
and worst fitness. Stores the best individual in the variable
|
| 256 |
self.bestMember. When the elitePercent is greater than zero,
|
| 257 |
this method also maintains a list of the elite members of the
|
| 258 |
population so that they can be saved for the next generation.
|
| 259 |
"""
|
| 260 |
self.sumFitness = 0
|
| 261 |
best = self.individuals[0]
|
| 262 |
best.bestPosition = 0
|
| 263 |
worst= self.individuals[0]
|
| 264 |
self.eliteMembers = self.individuals[0:int(self.elitePercent * len(self.individuals))]
|
| 265 |
self.eliteMembers.sort(lambda x, y: cmp( x.fitness, y.fitness))
|
| 266 |
for i in range(self.size):
|
| 267 |
current = self.individuals[i]
|
| 268 |
current.position = i #needed to save the elite members of the population
|
| 269 |
self.sumFitness += current.fitness
|
| 270 |
if current.fitness < worst.fitness:
|
| 271 |
worst = current
|
| 272 |
if current.fitness > best.fitness:
|
| 273 |
best = current
|
| 274 |
best.bestPosition = i
|
| 275 |
if len(self.eliteMembers) > 0 and current.fitness > self.eliteMembers[0].fitness:
|
| 276 |
self.eliteMembers.append( current )
|
| 277 |
self.eliteMembers.sort(lambda x, y: cmp( x.fitness, y.fitness))
|
| 278 |
self.eliteMembers = self.eliteMembers[1:]
|
| 279 |
self.bestMember = best
|
| 280 |
self.avgFitness = (self.sumFitness * 1.0) / self.size
|
| 281 |
if self.verbose > 0:
|
| 282 |
print "Fitness: Total", "%7.2f" % self.sumFitness,
|
| 283 |
print "Best", "%5.2f" % best.fitness,
|
| 284 |
print "Average", "%5.2f" % self.avgFitness,
|
| 285 |
print "Worst", "%5.2f" % worst.fitness
|
| 286 |
print "Elite fitness:", map( lambda x: x.fitness, self.eliteMembers)
|
| 287 |
sys.stdout.flush()
|
| 288 |
|
| 289 |
class GA:
|
| 290 |
"""
|
| 291 |
Class which defines everything needed to run a GA.
|
| 292 |
"""
|
| 293 |
def __init__(self, population, **args):
|
| 294 |
self.averageLog = None
|
| 295 |
self.bestLog = None
|
| 296 |
self.mutationRate = 0.1
|
| 297 |
self.crossoverRate = 0.6
|
| 298 |
self.maxGeneration = 0
|
| 299 |
self.generation = 0
|
| 300 |
self.verbose = 0
|
| 301 |
if args.has_key('verbose'):
|
| 302 |
self.verbose = args['verbose']
|
| 303 |
if args.has_key('mutationRate'):
|
| 304 |
self.mutationRate = args['mutationRate']
|
| 305 |
if args.has_key('crossoverRate'):
|
| 306 |
self.crossoverRate = args['crossoverRate']
|
| 307 |
if args.has_key('maxGeneration'):
|
| 308 |
self.maxGeneration = args['maxGeneration']
|
| 309 |
x = random.random() * 100000 + time.time()
|
| 310 |
self.setSeed(x)
|
| 311 |
self.origPop = population
|
| 312 |
if self.verbose > 0:
|
| 313 |
print "crossoverRate = %.3f" % self.crossoverRate
|
| 314 |
print "mutationRate = %.3f" % self.mutationRate
|
| 315 |
print "populationSize = %d" % self.origPop.size
|
| 316 |
print "elitePercent = %.3f" % self.origPop.elitePercent
|
| 317 |
print "maxGeneration = %d" % self.maxGeneration
|
| 318 |
print "================================================================================"
|
| 319 |
self.setup(**args)
|
| 320 |
self.reInitialize()
|
| 321 |
|
| 322 |
def setup(self, **args):
|
| 323 |
pass
|
| 324 |
|
| 325 |
def reInitialize(self):
|
| 326 |
self.pop = self.origPop.copy()
|
| 327 |
self.initialize()
|
| 328 |
|
| 329 |
def initialize(self):
|
| 330 |
self.applyFitnessFunction()
|
| 331 |
if self.verbose > 0:
|
| 332 |
print "-" * 60
|
| 333 |
print "Initial population"
|
| 334 |
self.pop.statistics()
|
| 335 |
if self.verbose > 1:
|
| 336 |
self.display()
|
| 337 |
|
| 338 |
def logAverageFitness(self, filename="GAAvgFitness"):
|
| 339 |
self.averageLog = open(filename, 'w')
|
| 340 |
|
| 341 |
def logBestFitness(self, filename="GABestFitness"):
|
| 342 |
self.bestLog = open(filename, 'w')
|
| 343 |
|
| 344 |
def isDone(self):
|
| 345 |
# Override this
|
| 346 |
pass
|
| 347 |
|
| 348 |
def fitnessFunction(self, genePosition, **args):
|
| 349 |
# Override this
|
| 350 |
pass
|
| 351 |
|
| 352 |
def applyFitnessFunction(self):
|
| 353 |
for i in range( len(self.pop.individuals) ):
|
| 354 |
self.pop.individuals[i].fitness = self.fitnessFunction(i)
|
| 355 |
|
| 356 |
def setSeed(self, value):
|
| 357 |
self.seed = value
|
| 358 |
random.seed(self.seed)
|
| 359 |
|
| 360 |
def display_one(self, p):
|
| 361 |
self.pop.individuals[p].display()
|
| 362 |
print "Fitness:", self.pop.individuals[p].fitness
|
| 363 |
|
| 364 |
def display(self):
|
| 365 |
print "Population:"
|
| 366 |
for p in range(len(self.pop.individuals)):
|
| 367 |
self.display_one(p)
|
| 368 |
|
| 369 |
def generate(self):
|
| 370 |
"""
|
| 371 |
Iteratively creates a new population from the current population.
|
| 372 |
Selects two parents, attempts to cross them, and then attempts to
|
| 373 |
mutate the resulting children. The probability of these operations
|
| 374 |
occurring is determined by the crossoverRate and the mutationRate.
|
| 375 |
Overwrites the old population with the new population.
|
| 376 |
"""
|
| 377 |
newpop = range(self.pop.size)
|
| 378 |
i = 0
|
| 379 |
while i < self.pop.size - 1:
|
| 380 |
parent1 = self.pop.select()
|
| 381 |
parent2 = self.pop.select()
|
| 382 |
newpop[i], newpop[i+1] = parent1.crossover(parent2, self.crossoverRate)
|
| 383 |
newpop[i].mutate(self.mutationRate)
|
| 384 |
newpop[i+1].mutate(self.mutationRate)
|
| 385 |
i += 2
|
| 386 |
# For odd sized populations, need to create the last child
|
| 387 |
if self.pop.size % 2 == 1:
|
| 388 |
newpop[self.pop.size-1] = self.pop.select()
|
| 389 |
newpop[self.pop.size-1].mutate(self.mutationRate)
|
| 390 |
# Copy new generation into population
|
| 391 |
elitePositions = map( lambda x: x.position, self.pop.eliteMembers)
|
| 392 |
for i in range(self.pop.size):
|
| 393 |
if i not in elitePositions:
|
| 394 |
self.pop.individuals[i] = newpop[i]
|
| 395 |
|
| 396 |
def evolve(self, cont = 0):
|
| 397 |
if not cont:
|
| 398 |
self.generation = 0
|
| 399 |
else:
|
| 400 |
if self.generation == self.maxGeneration:
|
| 401 |
self.maxGeneration = self.generation + 100
|
| 402 |
while self.generation < self.maxGeneration or self.maxGeneration == 0:
|
| 403 |
self.generation += 1
|
| 404 |
if self.verbose > 0:
|
| 405 |
print "-" * 60
|
| 406 |
print "Generation", self.generation
|
| 407 |
self.generate()
|
| 408 |
self.applyFitnessFunction()
|
| 409 |
self.pop.statistics()
|
| 410 |
if self.bestLog != None:
|
| 411 |
self.bestLog.write("%d %5.2f\n" %
|
| 412 |
(self.generation,
|
| 413 |
self.pop.bestMember.fitness))
|
| 414 |
if self.averageLog != None:
|
| 415 |
self.averageLog.write("%d %5.2f\n" %
|
| 416 |
(self.generation,
|
| 417 |
self.pop.avgFitness))
|
| 418 |
if self.verbose > 1:
|
| 419 |
self.display()
|
| 420 |
if self.isDone():
|
| 421 |
break
|
| 422 |
print "-" * 60
|
| 423 |
print "Done evolving at generation", self.generation
|
| 424 |
print "Current best individual [#%d]" % self.pop.bestMember.bestPosition,
|
| 425 |
self.pop.bestMember.display()
|
| 426 |
print "Fitness", self.pop.bestMember.fitness
|
| 427 |
|
| 428 |
def saveToFile(self, filename):
|
| 429 |
import pickle
|
| 430 |
fp = open(filename, "w")
|
| 431 |
if self.verbose > 0:
|
| 432 |
print "Saving GA to '%s'..." % (filename,)
|
| 433 |
pickle.dump(self, fp)
|
| 434 |
fp.close()
|
| 435 |
|
| 436 |
def loadFromFile(self, filename):
|
| 437 |
# probably just copy this... no need to create an entire object
|
| 438 |
# to load another one.
|
| 439 |
import pickle
|
| 440 |
fp = open(filename, "w")
|
| 441 |
if self.verbose > 0:
|
| 442 |
print "Loading GA from '%s'..." % (filename,)
|
| 443 |
fp.close()
|
| 444 |
return pickle.load(fp)
|
| 445 |
|
| 446 |
def saveGenesToFile(self, filename, listOfPositions = None):
|
| 447 |
import pickle
|
| 448 |
if listOfPositions == None:
|
| 449 |
listOfPositions = range(len(self.pop.individuals))
|
| 450 |
fp = open(filename, "w")
|
| 451 |
if self.verbose > 0:
|
| 452 |
print "Saving %d genes to '%s'..." % (len(listOfPositions), filename)
|
| 453 |
pickle.dump( len(listOfPositions), fp)
|
| 454 |
for i in listOfPositions:
|
| 455 |
pickle.dump(self.pop.individuals[i], fp)
|
| 456 |
fp.close()
|
| 457 |
|
| 458 |
def getGenesFromFile(self, filename):
|
| 459 |
import pickle
|
| 460 |
fp = open(filename, "r")
|
| 461 |
geneCount = pickle.load(fp)
|
| 462 |
if self.verbose > 0:
|
| 463 |
print "Loading %d genes from '%s'..." % (geneCount, filename)
|
| 464 |
individuals = []
|
| 465 |
for i in range(geneCount):
|
| 466 |
individuals.append(pickle.load(fp))
|
| 467 |
fp.close()
|
| 468 |
return individuals
|
| 469 |
|
| 470 |
def loadGenesFromFile(self, filename):
|
| 471 |
self.pop.individuals = self.getGenesFromFile(filename)
|
| 472 |
|
| 473 |
def initGenesFromFile(self, filename, sampleSize = 0,mutate = 1,full = 0):
|
| 474 |
# sampleSize = how many to get from saved pop?
|
| 475 |
# mutate = should I mutate them?
|
| 476 |
# full = should I create a full pop, or just replace sampleSize?
|
| 477 |
oldGenes = self.getGenesFromFile(filename)
|
| 478 |
if sampleSize == 0:
|
| 479 |
sampleSize = len(oldGenes)
|
| 480 |
if self.verbose > 0:
|
| 481 |
print "oldGenes had %d individuals" % len(oldGenes)
|
| 482 |
print "current has %d individuals" % len(self.pop.individuals)
|
| 483 |
print "Loading %d..." % sampleSize
|
| 484 |
if full:
|
| 485 |
currentOld = 0
|
| 486 |
for i in range(len(self.pop.individuals)):
|
| 487 |
currentOld = currentOld % len(oldGenes)
|
| 488 |
self.pop.individuals[i] = oldGenes[currentOld]
|
| 489 |
currentOld += 1
|
| 490 |
if mutate:
|
| 491 |
self.pop.individuals[i].mutate(self.mutationRate)
|
| 492 |
else:
|
| 493 |
for i in range(sampleSize):
|
| 494 |
self.pop.individuals[i] = oldGenes[i]
|
| 495 |
if mutate:
|
| 496 |
self.pop.individuals[i].mutate(self.mutationRate)
|
| 497 |
|
| 498 |
if __name__ == '__main__':
|
| 499 |
# Here is a test to evolve a list of integers to maximize their sum:
|
| 500 |
|
| 501 |
class MaxSumGA(GA):
|
| 502 |
def fitnessFunction(self, i):
|
| 503 |
return max(sum(self.pop.individuals[i].genotype), 0)
|
| 504 |
def isDone(self):
|
| 505 |
print "Best:",
|
| 506 |
self.pop.bestMember.display()
|
| 507 |
print
|
| 508 |
return self.pop.bestMember.fitness > 30
|
| 509 |
|
| 510 |
print "Do you want to evolve a list of integers to maximize their sum? ",
|
| 511 |
if sys.stdin.readline().lower()[0] == 'y':
|
| 512 |
print
|
| 513 |
ga = MaxSumGA(Population(20, Gene, size=10, mode='integer',
|
| 514 |
verbose=1, elitePercent = .1,
|
| 515 |
max = 3, maxStep = 2, min = 0,
|
| 516 |
crossoverPoints = 1),
|
| 517 |
mutationRate=0.1, crossoverRate=0.5, verbose=1,
|
| 518 |
maxGeneration=50)
|
| 519 |
ga.evolve()
|
| 520 |
print "Testing loading/saving..."
|
| 521 |
ga.saveGenesToFile("maxsumga.genes")
|
| 522 |
print "Deleting genes..."
|
| 523 |
ga.pop.individuals = []
|
| 524 |
ga.loadGenesFromFile("maxsumga.genes")
|
| 525 |
print "Press enter to continue evolving...",
|
| 526 |
sys.stdin.readline()
|
| 527 |
ga.evolve()
|
| 528 |
print "Press enter to Test init from file (load all with mutate)...",
|
| 529 |
sys.stdin.readline()
|
| 530 |
print "reInitialize pop..."
|
| 531 |
ga.reInitialize()
|
| 532 |
ga.initGenesFromFile("maxsumga.genes")
|
| 533 |
ga.evolve()
|
| 534 |
print "Press enter to Test init from file (load 1 no mutate)...",
|
| 535 |
sys.stdin.readline()
|
| 536 |
ga.saveGenesToFile("bestsumga.genes", (ga.pop.bestMember.position,))
|
| 537 |
ga.reInitialize()
|
| 538 |
ga.initGenesFromFile("bestsumga.genes", 1, 0)
|
| 539 |
ga.evolve()
|
| 540 |
print "Press enter to Test init from file (load 1, with mutate, full)...",
|
| 541 |
sys.stdin.readline()
|
| 542 |
ga.reInitialize()
|
| 543 |
ga.initGenesFromFile("bestsumga.genes", mutate = 1, full = 1)
|
| 544 |
ga.evolve()
|
| 545 |
print
|
| 546 |
|
| 547 |
# Here is a test to evolve the weights/biases in a neural network
|
| 548 |
# that solves the XOR problem:
|
| 549 |
|
| 550 |
from pyrobot.brain.conx import *
|
| 551 |
class NNGA(GA):
|
| 552 |
def __init__(self, cnt):
|
| 553 |
n = Network()
|
| 554 |
n.add( Layer('input', 2) )
|
| 555 |
n.add( Layer('hidden', 3) )
|
| 556 |
n.add( Layer('output', 1) )
|
| 557 |
n.connect('input', 'hidden')
|
| 558 |
n.connect('hidden', 'output')
|
| 559 |
n.setInputs([[0.0, 0.0],
|
| 560 |
[0.0, 1.0],
|
| 561 |
[1.0, 0.0],
|
| 562 |
[1.0, 1.0]])
|
| 563 |
n.setOutputs([[0.0],
|
| 564 |
[1.0],
|
| 565 |
[1.0],
|
| 566 |
[0.0]])
|
| 567 |
n.setVerbosity(0)
|
| 568 |
n.setTolerance(.4)
|
| 569 |
n.setLearning(0)
|
| 570 |
g = n.arrayify()
|
| 571 |
self.network = n
|
| 572 |
GA.__init__(self,
|
| 573 |
Population(cnt, Gene, size=len(g), verbose=1,
|
| 574 |
min=-10, max=10, maxStep = 1,
|
| 575 |
imin=-10, imax=10,
|
| 576 |
elitePercent = .01),
|
| 577 |
mutationRate=0.05, crossoverRate=0.6,
|
| 578 |
maxGeneration=400, verbose=1)
|
| 579 |
def fitnessFunction(self, genePos):
|
| 580 |
self.network.unArrayify(self.pop.individuals[genePos].genotype)
|
| 581 |
error, correct, count, pcorrect = self.network.sweep()
|
| 582 |
return 4 - error
|
| 583 |
def isDone(self):
|
| 584 |
self.network.unArrayify(self.pop.bestMember.genotype)
|
| 585 |
error, correct, count, pcorrect = self.network.sweep()
|
| 586 |
print "Correct:", correct
|
| 587 |
return correct == 4
|
| 588 |
|
| 589 |
print "Do you want to evolve a neural network that can do XOR? ",
|
| 590 |
if sys.stdin.readline().lower()[0] == 'y':
|
| 591 |
ga = NNGA(300)
|
| 592 |
ga.evolve()
|
| 593 |
ga.network.unArrayify(ga.pop.bestMember.genotype)
|
| 594 |
ga.network.setInteractive(1)
|
| 595 |
ga.network.sweep()
|
| 596 |
ga.saveGenesToFile("gann.pop")
|
| 597 |
ga.initGenesFromFile("gann.pop")
|
| 598 |
|
| 599 |
print "Do you want to evolve a phrase? ",
|
| 600 |
if sys.stdin.readline().lower()[0] == 'y':
|
| 601 |
phrase = "evolution is one cool search mechanism"
|
| 602 |
size = len(phrase)
|
| 603 |
print
|
| 604 |
class PhraseGA(GA):
|
| 605 |
def fitnessFunction(self, i):
|
| 606 |
sum = 0
|
| 607 |
for c in range(len(self.pop.individuals[i].genotype)):
|
| 608 |
if self.pop.individuals[i].genotype[c] == phrase[c]:
|
| 609 |
sum += 1
|
| 610 |
return float(sum) / len(self.pop.individuals[i].genotype)
|
| 611 |
def isDone(self):
|
| 612 |
print "Best:",
|
| 613 |
self.pop.bestMember.display()
|
| 614 |
return (phrase == string.join(self.pop.bestMember.genotype, ""))
|
| 615 |
|
| 616 |
ga = PhraseGA(Population(300, Gene, size=size, mode='char',
|
| 617 |
verbose=1, elitePercent = .1,
|
| 618 |
crossoverPoints = 2),
|
| 619 |
mutationRate=0.06, crossoverRate=0.6, verbose=1,
|
| 620 |
maxGeneration=0)
|
| 621 |
ga.evolve()
|
| 622 |
|
| 623 |
print "Do you want to play mastermind? ",
|
| 624 |
if sys.stdin.readline().lower()[0] == 'y':
|
| 625 |
# composed of N colors, M places (usually 6 and 4)
|
| 626 |
# feedback is # in correct place, # of correct color
|
| 627 |
phrase = "abcdefghijklmnopqrstuvwxyz"
|
| 628 |
size = len(phrase)
|
| 629 |
primer = 0
|
| 630 |
print
|
| 631 |
class MasterMindGA(GA):
|
| 632 |
def fitnessFunction(self, i):
|
| 633 |
sumPosition = 0
|
| 634 |
sumColor = 0
|
| 635 |
guessed = []
|
| 636 |
correct = []
|
| 637 |
for c in range(len(self.pop.individuals[i].genotype)):
|
| 638 |
if self.pop.individuals[i].genotype[c] == phrase[c]:
|
| 639 |
sumPosition += 1
|
| 640 |
else:
|
| 641 |
guessed.append(self.pop.individuals[i].genotype[c])
|
| 642 |
correct.append(phrase[c])
|
| 643 |
for g in guessed:
|
| 644 |
if g in correct:
|
| 645 |
correct.remove(g)
|
| 646 |
sumColor += 1
|
| 647 |
if primer:
|
| 648 |
if (sumColor + sumPosition > size/2):
|
| 649 |
goodPosition = 10
|
| 650 |
goodColor = 1
|
| 651 |
else:
|
| 652 |
goodPosition = 0
|
| 653 |
goodColor = 1
|
| 654 |
else:
|
| 655 |
goodPosition = 100
|
| 656 |
goodColor = 1
|
| 657 |
return sumColor * goodColor + sumPosition * goodPosition
|
| 658 |
def isDone(self):
|
| 659 |
print "Best:",
|
| 660 |
self.pop.bestMember.display()
|
| 661 |
return (phrase == string.join(self.pop.bestMember.genotype, ""))
|
| 662 |
|
| 663 |
ga = MasterMindGA(Population(300, Gene, size=size, mode='char',
|
| 664 |
verbose=1, elitePercent = .1,
|
| 665 |
crossoverPoints = 2),
|
| 666 |
mutationRate=0.06, crossoverRate=0.6, verbose=1,
|
| 667 |
maxGeneration=0)
|
| 668 |
ga.evolve()
|
| 669 |
|
| 670 |
# 26 ** 26 = 6,156,119,580,207,157,310,796,674,288,400,203,776 6x10^36
|
| 671 |
# 60 * 300 = 18,000
|
| 672 |
# 224 * 300 = 67200
|
| 673 |
|
| 674 |
# generations
|
| 675 |
# 224
|
| 676 |
# 171
|
| 677 |
# 60
|
| 678 |
# 167
|
| 679 |
# 404
|
| 680 |
|
| 681 |
# With priming
|
| 682 |
# 73
|
| 683 |
# 126
|
| 684 |
# 260
|
| 685 |
|
| 686 |
# no overlap
|
| 687 |
# 98
|
| 688 |
# 158
|
Parent Directory
| 
Revision Log