GraphIsomorphismNetworkOptions

Options & Configuration · AiDotNet.NeuralNetworks.Options

Configuration options for the GraphIsomorphismNetwork.

Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → GraphEncoderOptions → GraphIsomorphismNetworkOptions

For Beginners

The Graph Isomorphism Network is designed to tell apart graphs that other designs confuse. The values here are the ones the paper publishes, so you can use the model without configuring anything.

Constructors

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

Properties

Property Summary
bool LearnEpsilon { get; set; } Gets or sets whether epsilon is learned rather than held fixed.
double InitialEpsilon { get; set; } Gets or sets the starting value of epsilon.
int MlpHiddenDim { get; set; } Gets or sets the width of the small network inside each layer.

Methods

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