EchoStateNetworkOptions

Options & Configuration · AiDotNet.NeuralNetworks.Options

Configuration options for the EchoStateNetwork.

Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → ReservoirOptions → EchoStateNetworkOptions

For Beginners

An Echo State Network keeps a large pool of randomly connected neurons fixed and trains only a linear readout on top, which makes it very fast to train. The values here are the ones it ships with.

Constructors

Constructor Summary
EchoStateNetworkOptions() Initializes a new instance carrying the model’s published defaults.
EchoStateNetworkOptions(EchoStateNetworkOptions other) Initializes a new instance by copying another instance.

Properties

Property Summary
double Regularization { get; set; } Gets or sets the ridge regularization strength for the readout.
double Sparsity { get; set; } Gets or sets the fraction of reservoir connections that are non-zero.
int WarmupPeriod { get; set; } Gets or sets how many initial steps are discarded before readout training.

Methods

Method Summary
void Validate() Throws if a value this model requires has been left unset or is not positive.