simple_optimization

Some simple multi-threaded optimizers.

You could do a random search like:

use std::sync::Arc;
use simple_optimization::{random_search, Polling};
// Our evaluation function takes 3 `f64`s and no additional data `()`.
fn simple_function(list: &[f64; 3], _: Option<Arc::<()>>) -> f64 { list.iter().sum() }
let best = random_search!(
    (0f64..10f64, 5u32..15u32, 10i16..20i16), // Value ranges.
    simple_function, // Evaluation function.
    None, // No additional evaluation data.
    // Every `10ms` we print progress and exit early if `simple_function` return a value less than `19.`.
    Some(Polling::new(true,Some(19.))),
    None, // We don't specify the number of threads to run.
    1000, // Take `1000` samples.
);
assert!(simple_function(&best, None) < 19.);

Which during execution will give an output like:

1000
 500 (50.00%) 00:00:11 / 00:00:47 [25.600657363049734]

Representing:

<Total number of samples>
<Current number of samples> (<Percentage of samples taken>) <Time running> / <ETA> [<Current best value>]

Support

Optimizer	Status
Random search	✅
Grid search	✅
Simulated annealing	✅
Bayesian optimization	WIP
Gradient descent	See my note here
Genetic algorithms	No plans
Ant colony optimization	No plans
Linear programming	No plans

I made this for my own use since the existing libraries I found felt awkward.