stats_claw/optimizers/gradient/mod.rs
1//! First-order optimizers: learning-rate methods (gradient descent, SGD, Adam,
2//! `RMSProp`, `AdaGrad`) and the Fletcher–Reeves conjugate gradient.
3//!
4//! Each optimizer drives an [`Objective`](super::Objective) toward a minimum and
5//! returns an [`OptimizeResult`](super::OptimizeResult). The learning-rate
6//! methods stop when the gradient norm falls below a tolerance; conjugate
7//! gradient additionally restarts every `n` steps.
8
9pub mod adagrad;
10pub mod adam;
11pub mod conjugate_gradient;
12pub mod gradient_descent;
13pub mod rmsprop;
14pub mod sgd;
15
16pub use adagrad::adagrad;
17pub use adam::adam;
18pub use conjugate_gradient::conjugate_gradient;
19pub use gradient_descent::gradient_descent;
20pub use rmsprop::rmsprop;
21pub use sgd::sgd;