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

//! This module contains the definition of the PopulationRefitness trait and the provided
//! population refitness functions

use genotype::Genotype;
use niches_beta_rate::*;
use rayon::prelude::*;
use std::cmp::PartialEq;
use IndWithFitness;
use PopulationRefitnessFunctions::*;

/// This trait defines the population_refitness function that permits to modify the
/// individuals fitness comparing them to the others individuals in the population.
/// This function is called when all individuals fitness have been computed and just
/// before the survival pressure kill.
pub trait PopulationRefitness<T: PartialEq + Send + Sync, G: Genotype<T>>: Send + Sync {
    /// Modify the individual fitness comparing it with the other individuals in the
    /// population. Called just before survival pressure kill.
    ///
    /// # Parameters:
    /// - `individual_index`: The individual index inside the population.
    /// - `population`: The full population. It includes the individual that is being
    /// evaluated (`individual_index`) that is probably wanted to be excluded of the
    /// comparison with population individuals. The `Option<Fitness>` is always
    /// `Some(Fitness)` when this function is called.
    /// - `generation`: The current generation.
    /// - `progress`: The progress in the last iterations.
    /// - `n_solutions`: The number of found solutions in the last generation.
    ///
    /// # Returns
    /// The new fitness of the individual.
    fn population_refitness(
        &self,
        individual_index: usize,
        population: &[IndWithFitness<T, G>],
        generation: u64,
        progress: f64,
        n_solutions: usize,
    ) -> f64;
}

/// Importance given to distance among individuals. It is usually a number between 0 and 1.
#[derive(Debug)]
pub struct NichesAlpha(pub f64);

/// Threshold used in niches to divide distance by it if the distance is lower than sigma.
#[derive(Debug)]
pub struct NichesSigma(pub f64);

/// Provided PopulationRefitness functions.
pub enum PopulationRefitnessFunctions {
    /// Remains the fitness unmodified.
    None,
    /// Scales the fitness using the formula `f' = f^beta / m`, where `m` is
    /// the sum of (1 - distance / sigma) for each individual in the population.
    Niches(NichesAlpha, Box<NichesBetaRates>, NichesSigma),
}

impl<T: PartialEq + Send + Sync, G: Genotype<T>> PopulationRefitness<T, G>
    for PopulationRefitnessFunctions
{
    fn population_refitness(
        &self,
        individual_index: usize,
        population: &[IndWithFitness<T, G>],
        generation: u64,
        progress: f64,
        n_solutions: usize,
    ) -> f64 {
        match self {
            None => {
                population[individual_index]
                    .fitness
                    .unwrap()
                    .original_fitness
            }
            Niches(alfa, beta, sigma) => {
                let current_ind = &population[individual_index].ind;
                let current_fitness = &population[individual_index].fitness;
                let mut current_fitness = current_fitness.unwrap().original_fitness;
                if current_fitness > 0.0 {
                    let mut m = population
                        .par_iter()
                        .enumerate()
                        .filter(|(i, _ind)| *i != individual_index)
                        .map(|(_i, ind)| current_ind.distance(&ind.ind))
                        .map(|d| {
                            if d >= sigma.0 {
                                0.0
                            } else {
                                1.0 - (d / sigma.0).powf(alfa.0)
                            }
                        })
                        .sum::<f64>();
                    current_fitness =
                        current_fitness.powf(beta.rate(generation, progress, n_solutions));
                    if m == 0.0 {
                        m = f64::EPSILON;
                    }
                    current_fitness / m
                } else {
                    current_fitness
                }
            }
        }
    }
}