GraphGenerationModelOptions
Options & Configuration · AiDotNet.NeuralNetworks.Options
Configuration options for the GraphGenerationModel.
Inheritance: ModelOptions → NeuralNetworkOptions → ModelHyperparameterOptions → GraphGenerationModelOptions
For Beginners
This model learns to produce new graphs that look like the ones it was trained on — new molecules, new networks. The values here are the ones it ships with, so you can use it without configuring anything.
How It Works
Derives from ModelHyperparameterOptions rather than one of the graph family
bases. It is a generator, not an encoder or a task head: it shares neither
GraphEncoderOptions.NumLayers’s meaning nor GraphModelOptions.HiddenDim’s
required-ness with those families, and its architecture is built from these values rather
than supplied.
Constructors
| Constructor | Summary |
|---|---|
GraphGenerationModelOptions() |
Initializes a new instance carrying the model’s published defaults. |
GraphGenerationModelOptions(GraphGenerationModelOptions other) |
Initializes a new instance by copying another instance. |
Properties
| Property | Summary |
|---|---|
GraphGenerationType GenerationType { get; set; } |
Gets or sets the generation strategy. |
bool UseAMSGrad { get; set; } |
Gets or sets whether the optimizer uses the AMSGrad variant. |
double KlWeight { get; set; } |
Gets or sets the weight on the KL divergence term. |
double LearningRate { get; set; } |
Gets or sets the optimizer learning rate. |
int HiddenDim { get; set; } |
Gets or sets the width of the encoder’s hidden layers. |
int InputFeatures { get; set; } |
Gets or sets the number of features on each input node. |
int LatentDim { get; set; } |
Gets or sets the width of the latent space the model samples new graphs from. |
int MaxNodes { get; set; } |
Gets or sets the largest graph the model will generate, in nodes. |
int NumEncoderLayers { get; set; } |
Gets or sets the number of encoder layers. |
Methods
| Method | Summary |
|---|---|
void Validate() |
Throws if a value this model requires has been left unset or is not positive. |