#optimization #pure #algorithm #scalar #bounded #evaluate

swoop

Simple, lightweight optimisation algorithms in pure Rust

1 unstable release

0.1.0 May 1, 2022

#1657 in Algorithms

MIT license

24KB
603 lines

CircleCI MSRV version

swoop

Simple, lightweight optimisation algorithms in pure Rust

Motivation

This crate aims to mimic the scipy.optimize module in pure Rust.

Example

This crate has an asynchronous API and all examples use Tokio. To start your Cargo.toml should at least include

[dependencies]
swoop = { "git" = "https://github.com/benjaminjellis/swoop" }
tokio = { version = "1", features = ["full"] }

To minimise the function f(x) = 3x^2 + 4x + 50 in the bound -10 <= x <= 10 you can use the bounded optimiser

use swoop::minimise_scalar::{bounded, ScalarObjectiveFunction};
use swoop::SwoopErrors;

struct MyObjectiveFunction {
    a: f64,
    b: f64,
    c: f64,
}

impl MyObjectiveFunction {
    fn new(a: f64, b: f64, c: f64) -> Self {
        Self { a, b, c }
    }
}

impl ScalarObjectiveFunction for MyObjectiveFunction {
    fn evaluate(&self, x: f64) -> f64 {
        self.a * x.powf(2f64) + self.b * x + self.c
    }
}

#[tokio::main]
async fn main() -> Result<(), SwoopErrors> {
    let objective_function = MyObjectiveFunction::new(3f64, 4f64, 50f64);
    let result = bounded(objective_function, (-10f64, 10f64), 500usize).await?;
    println!("{:?}", result);
    Ok(())
}

Dependencies

~0.4–0.9MB
~20K SLoC