2 unstable releases

0.2.0 Aug 17, 2023
0.1.0 Aug 16, 2023

#2432 in Algorithms

MIT license

29KB
549 lines

genomic.rs

A small crate to help implement genetical algorithms in Rust.

This crate defines two traits:

  • Chromosome represents a single trainable parameter. The Chromosome trait is implemented for a few built-in types, and you can implement it on your own.
  • Genome represents a set of Chromosomes.

A few functions are also provided, notably:

  • mutate, to mutate an individual's genome
  • crossover, to perform the crossover operation on two individuals
  • reproduce, to perform sexuated reproduction on two individuals

Implementing the Genome trait is done in a declarative fashion:

use genomic::prelude::*;
use genomic::chromosome::UniformCh;

pub struct Triple {
    first: i32,
    second: i8,
    third: f32,
}

impl Genome for Triple {
    fn mutate(&mut self, mutator: &mut Mutator<impl rand::Rng>) {
        mutator
            .chromosome(&mut self.first)
            .chromosome(&mut self.second)
            // For floats, we need to choose a method and bounds for mutating them:
            .wrap_ch(UniformCh::new(self.third, 0.0, 1.0), &mut self.third);
    }

    fn crossover(&mut self, other: &mut Self, crossover: &mut Crossover<impl rand::Rng>) {
        crossover
            .chromosome(&mut self.first, &mut other.first)
            .chromosome(&mut self.second, &mut other.second)
            .chromosome(&mut self.third, &mut other.third);
    }

    fn size_hint(&self) -> usize {
        // We have three chromosomes
        3
    }
}

let mut triple = Triple {
    first: 0,
    second: 0,
    third: 0.0
};

genomic::mutate(
    // The genome to mutate
    &mut triple,
    // The mutation rate - a value of 1.0 means that the chromosomes will be fully scrambled
    1.0,
    // An RNG
    &mut rand::thread_rng()
);

Dependencies

~310KB