Skip to content

Latest commit

 

History

History
65 lines (43 loc) · 1.96 KB

README.md

File metadata and controls

65 lines (43 loc) · 1.96 KB

stochastic_optimizers

This crate provides implementations of common stochstic gradient optimization algorithms. They are designed to be lightweight, flexible and easy to use.

Currently implemted:

  • Adam
  • SGD
  • AdaGrad

The crate does not provide automatic differentiation, the gradient is given by the user.

Examples

use stochastic_optimizers::{Adam, Optimizer};
//minimise the function (x-4)^2
let start = -3.0;
let mut optimizer = Adam::new(start, 0.1);

for _ in 0..10000 {
   let current_paramter = optimizer.parameters();

   // d/dx (x-4)^2
   let gradient = 2.0 * current_paramter - 8.0;

   optimizer.step(&gradient);
}

assert_eq!(optimizer.into_parameters(), 4.0);

The parameters are owned by the optimizer and a reference can be optained by parameters(). After optimization they can be optained by into_parameters().

What types can be optimized

All types which impement the Parameters trait can be optimized. Implementations for the standart types f32, f64, Vec<T : Parameters> and [T : Parameters ; N] are provided.

Its realativly easy to implement it for custom types, see Parameters.

ndarray

By enabling the ndarray feature you can use Array as Parameters

Unit tests

The unit tests require libtorch via the tch crate. See github for installation details.

License

Licensed under either of

at your option.

Contribution

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.