creeky belly
Posts: 205
Joined: June 2006
|
I went for the numerical alphabet for convenience and wrote it up in python. It has a variable alphabet length, phrase length, number of generations, population size, and flat mutation rate per letter (plus a random 'phrase' every time). Anyone enjoy python? (insert Randall Munroe joke here)
| Quote | import pylab as p
import numpy as N
import copy as c
# main function - runs WEASEL simulation
def main(let=27,lin=20,pop=50,mut=0.05,gen=100):
# 'let' = length of alphabet
# 'lin' = 'phrase' length
# 'pop' = number of offspring
# 'mut' = probability of mutation of any letter in offspring
# 'gen' = number of generations to simulate
#create an array of 'lin' elements from a 'let' length alphabet
target = N.array([int(let*p.rand()) for x in N.arange(0,lin,1)])
#create initial population of completely random lines (offspring)
offsp = []
for i in N.arange(0,pop,1):
offsp.append(N.array([int(let*p.rand()) for x in N.arange(0,lin,1)]))
#store best fit
curbest=-1
##loop over generations
for i in N.arange(0,gen,1):
#initialize best fit offspring parameters
bestfit=0
bestoff=offsp[int(p.rand()*(pop-1))]
newoff=[]
##loop over offspring
for off in offsp:
#initialize fitness
currentfit = 0
##loop over 'letters' in 'phrase'
for pos in N.arange(0,lin,1):
#letters match => (fitness+1)
if target[pos]==off[pos]:
currentfit=currentfit+1
##end letter loop
#found a more fit individual, use it to populate next generation
if currentfit>bestfit:
bestoff=off
bestfit=currentfit
##end offspring loop
#only output if there's a change in fitness between generations
if curbest != bestfit:
print 'Best fit candidate %s out of %s in generation %s'% (bestfit,lin,i)
curbest = bestfit
##create next generation from the most fit candidate from previous generation
for k in N.arange(0,pop,1):
#make a copy of 'fittest' offspring
coff = c.copy(bestoff)
#create an array of uniform random numbers [0,1) of the 'phrase' length
probs = p.rand(lin)
##loop over random number array
for z in N.arange(0,lin,1):
#mutate corresponding letter if random number is less than the mutation rate
if probs[z]<mut:
coff[z]=int(let*p.rand())
##end random array loop
#add offspring to next generation
newoff.append(coff)
#set new generation
offsp=newoff
##end 'reproduction' loop
##end generation loop
print 'Target',target
print 'Best ',bestoff |
Here's some sample output:
| Quote | In [106]: we.main(lin=25,let=27,pop=100,mut=0.05,gen=50)
Best fit candidate 4 out of 25 in generation 0
Best fit candidate 5 out of 25 in generation 1
Best fit candidate 6 out of 25 in generation 2
Best fit candidate 8 out of 25 in generation 3
Best fit candidate 9 out of 25 in generation 4
Best fit candidate 10 out of 25 in generation 5
Best fit candidate 11 out of 25 in generation 6
Best fit candidate 12 out of 25 in generation 7
Best fit candidate 13 out of 25 in generation 8
Best fit candidate 14 out of 25 in generation 9
Best fit candidate 15 out of 25 in generation 10
Best fit candidate 16 out of 25 in generation 13
Best fit candidate 17 out of 25 in generation 14
Best fit candidate 18 out of 25 in generation 16
Best fit candidate 19 out of 25 in generation 22
Best fit candidate 20 out of 25 in generation 27
Best fit candidate 21 out of 25 in generation 28
Best fit candidate 22 out of 25 in generation 34
Best fit candidate 23 out of 25 in generation 35
Target [ 2 1 25 14 6 10 3 10 21 16 17 10 5 19 5 10 11 1 18 23 24 3 21 8 8]
Best [ 2 1 25 14 6 10 3 10 21 16 12 15 5 19 5 10 11 1 18 23 24 3 21 8 8] |
|
