1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
|
#genetic algorithm scaling routines
#based on galib.
#
from ga_util import *
import scipy.stats as stats
from numpy import *
# if a score is less the 2 standard deviations below, the average, its score
# is arbitrarily set to zero
class sigma_truncation_scaling:
def __init__(self,scaling = 2):
self.scaling = scaling
def scale(self,pop):
sc = pop.scores()
avg = my_mean(sc)
if len(sc) > 1: dev = my_std(sc)
else: dev = 0
f = sc - avg + self.scaling * dev
f=choose(less_equal(f,0.),(f,0.))
for i in range(len(pop)): pop[i].fitness(f[i])
return pop
class no_scaling:
def scale(self,pop):
for ind in pop: ind.fitness(ind.score())
return pop
class linear_scaling:
def __init__(self,mult = 1.2):
self.mult = mult
def scale(self,pop):
sc = pop.scores()
pmin = min(sc)
if pmin < 0: raise GAError, 'linear scaling does not work with objective scores < 0'
pmax = max(sc)
pavg = my_mean(sc)
if(pavg == pmax):
a = 1.
b = 0.
elif pmin > (self.mult * pavg - pmax)/(self.mult - 1.):
delta = pmax - pavg
a = (self.mult - 1.) * pavg / delta
b = pavg * (pmax - self.mult * pavg) / delta
else:
delta = pavg - pmin
a = pavg / delta
b = -pmin * pavg / delta
f = sc * a + b
f=choose(less_equal(f,0.),(f,0.))
for i in range(len(pop)): pop[i].fitness(f[i])
|
