def __init__(self):
self.mutationRate = 50
self.crossoverRate = 30
self.elitismRate = 10
self.newRate = 10
self.maxEpoch = 100
self.populationSize = 100
self.generation = [] #list
self.random = Go_Random(randomFunction)
self.chromosomeLength = 50
self.minCodon = 0
self.maxCodon = 1
self.calculatedFitness = {}
class Go_GeneticsAlgorithm(object):
def __init__(self):
self.mutationRate = 50
self.crossoverRate = 30
self.elitismRate = 10
self.newRate = 10
self.maxEpoch = 100
self.populationSize = 100
self.generation = [] #list
self.random = Go_Random(randomFunction)
self.chromosomeLength = 50
self.minCodon = 0
self.maxCodon = 1
self.calculatedFitness = {}
def __str__(self):
string = ''
for i in xrange(len(self.generation)):
string+= '\nGeneration %d\n' % i
for j in xrange(self.populationSize):
fitness = str(j)
string+= 'fitness #%s : %5.3f, ' % (fitness.rjust(3), self.generation[i][j]['fitness'])
string += '\n'
return string
def __currentGenerationIndex(self):
return len(self.generation)-1
def _mutation(self, chromosome):
#return new chromosome
i=0
rnd_count = int(round(self.random.get(0,self.chromosomeLength-1)/2))
while i<rnd_count:
rnd_addition = int(round(self.random.get(self.minCodon,self.maxCodon)))
rnd_index = int(round(self.random.get(0,self.chromosomeLength-1)))
newCodon = self.minCodon+((chromosome[rnd_index]-self.minCodon+rnd_addition) % (self.maxCodon+1-self.minCodon))
chromosome[rnd_index] = newCodon
i+=1
return chromosome
def _crossover(self, chromosome1, chromosome2):
#return new chromosome
random = int(round(self.random.get(0,self.chromosomeLength-1)))
segment1 = chromosome1[0:random]
segment2 = chromosome2[random:self.chromosomeLength]
chromosome = segment1+segment2
return chromosome
def _new(self):
#return new chromosome
chromosome = []
i=0
while i< self.chromosomeLength:
random = int(round(self.random.get(self.minCodon,self.maxCodon)))
chromosome.append(random);
i+=1
return chromosome
def _pick(self, index=None):
#return any chromosome from current generation
currentGenerationIndex = self.__currentGenerationIndex()
currentGeneration = self.generation[currentGenerationIndex]
if index!=None:
return currentGeneration[index]['chromosome'][:]
else:
maxCumSum = float(currentGeneration[self.populationSize-1]['cumsum'])
random = float(self.random.get(0,maxCumSum))
chromosome = None
i=0
while random > currentGeneration[i]['cumsum'] and i<len(currentGeneration)-1:
i += 1
chromosome = currentGeneration[i]['chromosome'][:]
return chromosome
def _fitnessFunction(self, chromosome):
#you surely need to override this
return sum(chromosome)
def _makeNextGeneration(self):
newGeneration = []
#the ratio below are related to the decision created
elitismLimit = float(self.elitismRate)
mutationLimit = elitismLimit+float(self.mutationRate)
crossoverLimit = mutationLimit+float(self.crossoverRate)
allLimit = crossoverLimit+float(self.newRate)
elitismRatio = elitismLimit/allLimit
mutationRatio = mutationLimit/allLimit
crossoverRatio = crossoverLimit/allLimit
for i in xrange(self.populationSize):
newChromosome = []
if(self.__currentGenerationIndex()==-1):
newChromosome = self._new()
else:
ratio=float(i)/float(self.populationSize)
newChromosome = []
if ratio <= elitismRatio:
newChromosome = self._pick(i)
elif ratio <= mutationRatio:
newChromosome = self._mutation(self._pick())
elif ratio <= crossoverRatio:
newChromosome = self._crossover(self._pick(), self._pick())
else:
newChromosome = self._new()
#calculate fitness if needed or take from self.calculatedFitness
#if calculation has been performed before
fitness=0
chromosomeKey = ' '.join(map(str,newChromosome))
if not(chromosomeKey in self.calculatedFitness):
fitness = self._fitnessFunction(newChromosome) #calculate...
self.calculatedFitness[chromosomeKey] = fitness #and save
else:
fitness = self.calculatedFitness[chromosomeKey] #take from calculated
#new Individu
newIndividu = {'chromosome' : newChromosome,
'fitness' : fitness,
'cumsum' : 0}
newGeneration.append(newIndividu)
#sort the new generation based on their fitness
newGeneration_dec = map(lambda x: (x["fitness"],x), newGeneration)
newGeneration_dec.sort(reverse=True)
newGeneration = map(lambda x: x[1],newGeneration_dec)
#calculate cumsum
cumsum=0
for i in xrange(self.populationSize):
cumsum += newGeneration[i]['fitness']
newGeneration[i]['cumsum'] = cumsum
self.generation.append(newGeneration)
self._printProgress()
def _printProgress(self):
currentGenerationIndex = self.__currentGenerationIndex()
currentGeneration = self.generation[currentGenerationIndex]
bestFitness = currentGeneration[0]['fitness']
print('Generation %d, best fitness : %5.5f'%(currentGenerationIndex, bestFitness))
def train(self):
i=0
while i<self.maxEpoch:
self._makeNextGeneration()
i+=1
