-
Notifications
You must be signed in to change notification settings - Fork 0
/
NaturalSelection.java
121 lines (104 loc) · 4.76 KB
/
NaturalSelection.java
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
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
import java.util.*;
import java.util.Random;
class NaturalSelection {
Network[] networks;
public static final float MUTATION_CHANCE = 0.05f;
public static final int MUTATION_FORCE = 20;
NaturalSelection (Network[] networks) {
this.networks = networks;
}
//performs natural selection by only allowing the 50% most fit to continue
//to the next generation. the original population is restored through
//breeding. Genes will be split and possibly mutated to create the new children
public Network[] newGeneration () {
//get the top 50%, rounded up
Network[] mostFit = getMostFit();
//next, create a new generation through crossing genes and mutations
Network[] nextGen = createNewNetworks(mostFit, networks.length);
return nextGen;
}
private Network[] getMostFit () {
//sort the network array by the greatest fitness value
sort();
//now only take the top 50%. round up if the number is odd
int finalLength = (int) Math.ceil(networks.length / 2.0f);
Network[] newNetworks = new Network[finalLength];
for (int index = 0; index < finalLength; index++) {
newNetworks[index] = networks[index];
}
return newNetworks;
}
//makes more networks based on two parents and mutations. networkCount is how many
//should exist by the end of this function
private Network[] createNewNetworks (Network[] mostFit, int networkTotal) {
//keep making new networks until networkTotal is reached
Network[] newGen = new Network[0];
while (newGen.length < networkTotal) {
//use probability based on their fitness scores to choose random parents
float fitnessTotal = 0;
float[] probabilities = new float[mostFit.length];
for (int i = 0; i < mostFit.length; i++) {
probabilities[i] = mostFit[i].getFitness();
fitnessTotal += mostFit[i].getFitness();
}
//if fitnessTotal is 0, everyone failed. give everyone equal probability to reproduce
if (fitnessTotal == 0) {
for (int i = 0; i < mostFit.length; i++) {
mostFit[i].setFitness(1);
probabilities[i] = 1;
fitnessTotal += 1;
}
}
//normalize all the values so they are in percentages
for (int i = 0; i < mostFit.length; i++) {
probabilities[i] /= fitnessTotal;
}
//pick two parents based on this probability distribution given to each network
//get two random parents together
int firstParentIndex = pickIndex(probabilities);
int secondParentIndex = pickIndex(probabilities);
//let parents allow to be the same in the case where one parent gets all the points and has
//100% chance to be picked; that will always be picked anyway
Gene firstParentGene = new Gene(mostFit[firstParentIndex]);
Gene secondParentGene = new Gene(mostFit[secondParentIndex]);
Gene[] children = firstParentGene.breed(secondParentGene);
Network offspring1 = children[0].toNetwork();
Network offspring2 = children[1].toNetwork();
int finalLength = newGen.length + 2;
if (finalLength > networkTotal) {
finalLength = networkTotal;
}
Network[] newGenAddTwo = new Network[finalLength];
for (int index = 0; index < newGen.length; index++) {
newGenAddTwo[index] = newGen[index];
}
//add the two new offspring
if (newGen.length < networkTotal)
newGenAddTwo[newGen.length] = offspring1;
if (newGen.length + 1 < networkTotal)
newGenAddTwo[newGen.length + 1] = offspring2;
newGen = newGenAddTwo; //newGen has two more offspring on it
}
return newGen;
}
public int pickIndex (float[] probabilities) {
float prob = (float) Math.random();
float cumulativeProbability = 0.0f;
for (int i = 0; i < probabilities.length; i++) {
cumulativeProbability += probabilities[i];
if (prob <= cumulativeProbability) {
return i;
}
}
return probabilities.length - 1; //if the for loop went all the way through
//(theoretically impossible) then just return the last index
}
//sort the network array, largest to smallest
private void sort () {
Arrays.sort(networks, new Comparator<Network>() {
public int compare(Network n1, Network n2) {
return Float.compare(n2.getFitness(), n1.getFitness());
}
});
}
}