52 releases (29 stable)

Uses old Rust 2015

1.8.1 Jan 21, 2021
1.8.0 Jul 20, 2018
1.7.14 Feb 23, 2018
1.7.9 Oct 14, 2017
0.9.1 Dec 29, 2015

#473 in Algorithms

27 downloads per month

MIT/Apache

58KB
908 lines

RsGenetic

Rust Report Card Build Status Crates Version License MIT License Apache

Summary and Features

RsGenetic is a framework for executing genetic algorithms in Rust. It is designed to have a simple but modular API.

Examples and Documentation

Documentation is available here.

Implementing the Fitness trait

Note that, if your fitness type is an integer type, you do not need to write a wrapper struct around this integer. See the types module documentation for more details.

use rsgenetic::pheno::*;
use std::cmp::Ordering;

#[derive(Eq, PartialEq, PartialOrd, Ord)]
struct MyFitness {
    value: i32,
}

impl Fitness for MyFitness {
    // The zero value for our custom type
    fn zero() -> MyFitness {
        MyFitness { value: 0 }
    }

    // The absolute difference between two instances
    fn abs_diff(&self, other: &MyFitness) -> MyFitness {
        MyFitness {
            value: (self.value - other.value).abs()
        }
    }
}

Implementing the Phenotype trait

Note that we use an integer type as the fitness type parameter to make this example more simple. Replace it with your custom type if needed. In this example, we try to find individuals with two integer components that sum to a target value.

This example is far-fetched, but simplified to show how easy it is to define new individuals and implement the Phenotype trait.

use rsgenetic::pheno::*;

const TARGET: i32 = 100;

#[derive(Copy, Clone)]
struct MyPheno {
    x: i32,
    y: i32,
}

impl Phenotype<i32> for MyPheno {
    // How fit is this individual?
    fn fitness(&self) -> i32 {
        TARGET - (self.x + self.y)
    }

    // Have two individuals create a new individual
    fn crossover(&self, other: &MyPheno) -> MyPheno {
        MyPheno {
            x: self.x,
            y: other.y,
        }
    }

    // Mutate an individual, changing its state
    fn mutate(&self) -> MyPheno {
        MyPheno {
            x: self.x + 1,
            y: self.y - 1,
        }
    }
}

Creating and running a Simulator

use rsgenetic::pheno::*;
use rsgenetic::sim::*;
use rsgenetic::sim::seq::Simulator;
use rsgenetic::sim::select::*;

const TARGET: i32 = 100;

#[derive(Copy, Clone)]
struct MyPheno {
    x: i32,
    y: i32,
}

impl Phenotype<i32> for MyPheno {
    // How fit is this individual?
    fn fitness(&self) -> i32 {
        TARGET - (self.x + self.y)
    }

    // Have two individuals create a new individual
    fn crossover(&self, other: &MyPheno) -> MyPheno {
        MyPheno {
            x: self.x,
            y: other.y,
        }
    }

    // Mutate an individual, changing its state
    fn mutate(&self) -> MyPheno {
        MyPheno {
            x: self.x + 1,
            y: self.y - 1,
        }
    }
}

fn main() {
    let mut population = (0..100).map(|i| MyPheno { x: i, y: 100 - i }).collect();
    let mut s = Simulator::builder(&mut population)
                    .set_selector(Box::new(StochasticSelector::new(10)))
                    .set_max_iters(50)
                    .build();
    s.run();
    let result = s.get().unwrap(); // The best individual
}

See the examples directory in the repository for more elaborate examples.

Note

This library is currently in maintenance mode. There have been some indications that the API needs an update to be more flexible, which would require an incrementation of the major version number (#23, #30). Unfortunately, I currently do not have the time to implement such a redesign. I will however continue to reply to issues and merge pull requests, but features might not be implemented by me, depending on their size.

License

Licensed under either of

at your option.

Contribution

Contributions are always welcome. Take a look at the issues for any enhancements that need to be done or bugs that need to be fixed. If you encounter any bugs while using the library, feel free to open an issue and/or fix the bug, and submit pull requests.

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.

Dependencies

~2MB
~31K SLoC