Crate reservoir_train 

Training utilities for Echo State Networks (ESN).

reservoir-train provides training-time components for Echo State Networks:

• ridge / lasso solvers for linear readouts
• an ESNBuilder for quickly assembling common ESN configurations
• (optional, std) static code generation for embedded inference models

This crate pairs with [reservoir-infer] (inference-time reservoirs/readouts).

Features

• std (default): Enables std and code generation utilities.
• libm: Enables libm-backed math for no_std targets (also affects dependencies).

High-level workflow

1. Build an ESN (dense or sparse) with ESNBuilder
2. Fit the readout using ESNFitRidge::fit or ESNFitLasso::fit_lasso
3. Run inference using the same model (predict), or export weights for deployment.

Examples

See the examples/ directory:

• esn_henon.rs
• esn_mackey_glass.rs
• esn_narma.rs

Re-exports

pub use codegen::StaticModelGenerator;

pub use esn::ESNBuilder;

pub use esn::ESNFitLasso;

pub use esn::ESNFitRidge;

pub use trainer::RidgeTrainer;

pub use reservoir_infer;

Modules

codegen

esn

float

trainer

Readout training algorithms.

Structs

DenseReservoir

Dense ESN reservoir using dense matrices.

EchoStateNetwork

A lightweight wrapper that wires a Reservoir and a Readout for inference.

LassoReadout

Linear readout backed by a dense matrix (DMatrix).

RidgeReadout

Linear readout backed by a dense matrix (DMatrix).

Type Aliases

RngType

RNG type used internally by ESNBuilder.