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
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154


use super::traits::*;
use std::collections::HashSet;
use rand::random;

fn randint(n: usize) -> usize {
    ((n as f64)*random::<f64>()) as usize
}

fn pick_one(list: &[f64]) -> usize {
    let total: f64 = list.iter().sum();
    let mut r = random::<f64>();

    for (index, &item) in list.iter().enumerate() {
        let prob = item / total;
        r -= prob;
        if (r < 0.) {
            return index;
        }
    }

    list.len() - 1
}

/// Neat struct that takes care of evolving the population based on the fitness scores
#[derive(Debug)]
pub struct Neat<T: Gene> {
    nodes: usize,
    connections: HashSet<(usize, usize)>,
    genomes: Vec<T>,
    mutation_rate: f64
}

impl<T: Gene> Neat<T> {
    /// Creates a new population
    pub fn new(inputs: usize, outputs: usize, size: usize, mutation_rate: f64) -> Self {
        Self {
            nodes: 1 + inputs + outputs,
            connections: HashSet::new(),
            genomes: (0..size).map(|_| Gene::empty(inputs, outputs)).collect(),
            mutation_rate
        }
    }

    fn speciate(&self) -> Vec<Vec<usize>> {
        let mut speciess = Vec::new();
        speciess.push(vec![0]);

        for i in 1..self.genomes.len() {
            let mut add_to = 0;
            for species in &mut speciess {
                let index = species[randint(species.len())];
                if self.genomes[i].is_same_species_as(&self.genomes[index]) {
                    species.push(i);
                    break;
                }
                add_to += 1;
            }
            if add_to >= speciess.len() {
                speciess.push(vec![i]);
            }
        }
        speciess
    }

    fn calculate_fitness(&self, calculate: fn(&T, bool) -> f64) -> (Vec<f64>, f64) {
        let mut total_score = 0.;
        let mut best_score = 0.;
        let mut fittest = &self.genomes[0];
        let mut scores = Vec::new();

        for genome in &self.genomes {
            let score = calculate(genome, false);
            if score > best_score {
                best_score = score;
                fittest = genome;
            }
            scores.push(score);
            total_score += score;
        }
        calculate(fittest, true);
        (scores, total_score)
    }

    /// Takes care of evaluation, speciation, selection and mutation and creates the
    /// new population
    pub fn next_generation(&mut self, calculate: fn(&T, bool) -> f64) {
        let (scores, total_score) = self.calculate_fitness(calculate);

        let total_mean = total_score / (scores.len() as f64);

        let mut speciess = self.speciate();

        println!("Number of species = {}", speciess.len());
        println!("Number of genomes = {}", self.genomes.len());

        let mut next_generation = Vec::new();

        for species in &mut speciess {
            // Sort species based on fitness
            species.sort_by(|&a, &b| scores[b].partial_cmp(&scores[a]).unwrap());

            let top = (0.6 * species.len() as f64).round() as usize;

            let species_scores: Vec<_> = species
                .iter()
                .map(|&i| scores[i])
                .collect();

            let num: f64 = species_scores.iter().sum();
            let num = (num / total_mean).round() as usize;

            for _ in 0..num {
                let index1 = pick_one(&species_scores[..top]);
                let index2 = pick_one(&species_scores[..top]);

                let one = &self.genomes[index1];
                let two = &self.genomes[index2];

                let mut child = if scores[index1] > scores[index2] {
                    one.cross(two)
                } else {
                    two.cross(one)
                };

                if random::<f64>() < self.mutation_rate {
                    child = child.mutate(self);
                }

                next_generation.push(child);
            }
        }

        self.genomes = next_generation;
    }
}

impl<T: Gene> GlobalNeatCounter for Neat<T> {
    fn try_adding_connection(&mut self, from: usize, to: usize) -> Option<usize> {
        let innov_num = self.connections.len();

        if self.connections.insert((from, to)) {
            Some(innov_num)
        } else {
            None
        }
    }

    fn get_new_node(&mut self) -> usize {
        let new_node = self.nodes;
        self.nodes += 1;
        new_node
    }
}