Another python GA implementation

05-genetic-algorithms/python/ga.py

@@ -0,0 +1,89 @@
+#!/usr/bin/env python
+import random
+import string
+import sys
+from collections import namedtuple
+
+TARGET="I LOVE GENETIC ALGORITHMS"
+MUTATION_RATE = 0.75
+Candidate = namedtuple("Candidate", "string fitness")
+Population = namedtuple("Population", "average min max size best")
+
+def generate_ran_string():
+    return "".join([random.choice(string.ascii_letters + " ") for i in xrange(random.randint(1, 100))])
+
+def generate_population():
+   candidates = []
+   for i in xrange(500):
+        string = generate_ran_string()
+        fitness = calculate_fitness(string)
+        candidate = Candidate(string=string, fitness=fitness)
+        candidates.append(candidate)
+   return candidates
+
+def calculate_fitness(string):
+    fitness = 0
+    for (a,b) in zip(string, TARGET):
+        if a==b:
+            fitness +=1
+    diff_length = abs(len(string) - len(TARGET))
+    fitness -= (diff_length*1.1)
+    return fitness
+
+def calc_population_stats(pop):
+    pop = sorted(pop, key = lambda x: -x.fitness)
+    fitnesses = [cand.fitness for cand in pop]
+    average = sum(fitnesses) / len(pop)
+    min_fit = min(fitnesses)
+    max_fit = max(fitnesses)
+    length = len(fitnesses)
+    best = pop[0]
+    population = Population(average=average, min=min_fit, max=max_fit, size=length, best=best.string)
+    return population
+
+def select_candidates(pop):
+    ordered_pop = sorted(pop, key=lambda x: x.fitness)
+    return ordered_pop[len(ordered_pop)/2:]
+
+def breed_population(candidates):
+    shuffled_cands = sorted(candidates, key=lambda x: random.randint(1,100))
+    pairs = zip(candidates, shuffled_cands)
+    cands = []
+    for pa, pb in pairs:
+        c= breed_candidates(pa, pb)
+        cands.extend([pa, c])
+    return cands
+
+def breed_candidates(candA, candB):
+    (a1, a2) = split_cand(candA)
+    (b1, b2) = split_cand(candB)
+    string = mutate_child(a1 + b2)
+    child = Candidate(string=string, fitness=calculate_fitness(string))
+    return child
+
+def split_cand(cand):
+     return (cand.string[:len(cand.string)/2], cand.string[len(cand.string)/2:])
+
+def mutate_child(dna):
+    new_dna = dna
+    if random.random() < MUTATION_RATE:
+        cut_point = random.randint(0, len(dna))
+        new_dna = dna[:cut_point-1] + random.choice(string.ascii_letters + " ") + dna[cut_point:]
+    return new_dna
+
+def main():
+    candidates = generate_population()
+    num_iter = 0
+    while True:
+        print calc_population_stats(candidates)
+        for cand in candidates:
+            if cand.fitness == calculate_fitness(TARGET):
+                print cand.string
+                print num_iter
+                sys.exit(0)
+        candidates = select_candidates(candidates)
+        candidates = breed_population(candidates)
+        num_iter += 1
+
+if __name__ == "__main__":
+    main()