#auto-diff #floating-point #rootsearch

rusty-rootsearch

This package implements general one-dimensional root-finding algorithms built on the shoulders of the num_dual crate

9 releases

0.4.0 Apr 6, 2024
0.3.1 Jun 11, 2023
0.2.1 Jun 11, 2023
0.1.4 Jun 11, 2023
0.1.2 May 4, 2023

#983 in Algorithms

Download history 1/week @ 2024-02-22 8/week @ 2024-02-29 49/week @ 2024-03-07 5/week @ 2024-03-14 4/week @ 2024-03-28 140/week @ 2024-04-04 9/week @ 2024-04-11

153 downloads per month

MIT license

12KB
263 lines

rusty-rootsearch

This is a small Rust-package that performs Newton's method for finding the roots of a function using the method of bisections.

It stands completely on the shoulders of the num_dual crate that implements auto-differentation for floating point numbers.

In an effort to generalise the implementation, the traits Derivable and Coercable are defined that allow the user to define their own types that can be used within the root_search, newton, and find_bisections functions. The Derivable trait is used to define the derivative of a number, which in the current implementation works well with the num_dual crate. The Coercable trait is used to convert a floating point number into the type that is derivable. This is useful for example when using the f64 type, which is not derivable, but can be converted to a Dual type exposed by the num_dual crate.

Example

use rusty_rootsearch::{root_search};
use num_dual::*;

fn find_sine_roots() {
    fn sine<D: DualNum<f32>>(x: D) -> D {
        x.sin()
    }
    let roots = root_search::<_,Dual32,f32>(&sine, -5.0, 5.0, 2000, 1000, 0.0001);
    for root in &roots.0 {
        println!("root: {}", root);
    }
    assert_eq!(roots.0.len(), 3);
    assert!(roots.0.contains(&std::f32::consts::PI));
    assert!(roots.0.contains(&(-std::f32::consts::PI)));
    assert!(roots.0.contains(&0.0));
}

Dependencies

~3MB
~62K SLoC