oracle.pgx.api.mllib

Class GraphWiseConvLayerConfigBuilder

java.lang.Object

oracle.pgx.api.mllib.GraphWiseConvLayerConfigBuilder

public class GraphWiseConvLayerConfigBuilder
extends java.lang.Object

Builder for GraphWiseConvLayerConfig. Currently, only mean aggregation layers are supported.

Since:

19.4

Constructor Summary

Constructors

Constructor and Description
GraphWiseConvLayerConfigBuilder()

Method Summary

Modifier and Type	Method and Description
GraphWiseConvLayerConfig	build() Build the configuration
GraphWiseConvLayerConfigBuilder	setActivationFunction(ActivationFunction activationFunction) Set the activation function for this layer (see ActivationFunction)
GraphWiseConvLayerConfigBuilder	setDropoutRate(double dropoutRate) Set the dropout rate for this layer.
GraphWiseConvLayerConfigBuilder	setNumSampledNeighbors(int numSampledNeighbors) Set the number of neighbors that will be sampled in this layer.
GraphWiseConvLayerConfigBuilder	setWeightedAggregationProperty(java.lang.String neighborWeightProperty) Set the graph property to be used for weighted aggregation.
GraphWiseConvLayerConfigBuilder	setWeightInitScheme(WeightInitScheme weightInitScheme) Set the weight initialization scheme for the weights of this layer.
GraphWiseConvLayerConfigBuilder	useEdgeToEdgeConnection(boolean edgeToEdgeConnection) Use the connection between edges and edges.
GraphWiseConvLayerConfigBuilder	useEdgeToVertexConnection(boolean edgeToVertexConnection) Use the connection between edges and vertices.
GraphWiseConvLayerConfigBuilder	useVertexToEdgeConnection(boolean vertexToEdgeConnection) Use the connection between vertices and edges.
GraphWiseConvLayerConfigBuilder	useVertexToVertexConnection(boolean vertexToVertexConnection) Use the connection between vertices and vertices.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

GraphWiseConvLayerConfigBuilder

public GraphWiseConvLayerConfigBuilder()

Method Detail

build

public GraphWiseConvLayerConfig build()

Build the configuration

Returns:

built configuration

Since:

19.4

setActivationFunction

public GraphWiseConvLayerConfigBuilder setActivationFunction(ActivationFunction activationFunction)

Set the activation function for this layer (see ActivationFunction)

default: GraphWiseConvLayerConfig.DEFAULT_ACTIVATION_FUNCTION

Parameters:

activationFunction - activation function

Returns:

this

Since:

19.4

setDropoutRate

public GraphWiseConvLayerConfigBuilder setDropoutRate(double dropoutRate)

Set the dropout rate for this layer. Represents the probability to drop each neuron. Note that dropoutRate must satisfy 0 <= dropoutRate < 1. If dropoutRate = 0, dropout layer will not be constructed.

default: GraphWiseConvLayerConfig.DEFAULT_DROPOUT_RATE

Parameters:

dropoutRate - dropout rate

Returns:

this

Since:

22.3

setNumSampledNeighbors

public GraphWiseConvLayerConfigBuilder setNumSampledNeighbors(int numSampledNeighbors)

Set the number of neighbors that will be sampled in this layer.

default: GraphWiseConvLayerConfig.DEFAULT_NUM_SAMPLED_NEIGHBORS

Parameters:

numSampledNeighbors - number of neighbors to sample

Returns:

this

Since:

19.4

setWeightedAggregationProperty

public GraphWiseConvLayerConfigBuilder setWeightedAggregationProperty(java.lang.String neighborWeightProperty)

Set the graph property to be used for weighted aggregation. If null, weighted aggregation will be disabled.

Weighted aggregation will perform the mean aggregation of neighbors based on weights given by the specified graph property.

default: GraphWiseConvLayerConfig.DEFAULT_NEIGHBOR_WEIGHT_PROPERTY_NAME

Parameters:

neighborWeightProperty - the property name

Returns:

this

Since:

19.4

setWeightInitScheme

public GraphWiseConvLayerConfigBuilder setWeightInitScheme(WeightInitScheme weightInitScheme)

Set the weight initialization scheme for the weights of this layer. Note that biases are always initialized to zeros. See WeightInitScheme.

default: GraphWiseConvLayerConfig.DEFAULT_WEIGHT_INIT_SCHEME

Parameters:

weightInitScheme - weight initialization scheme

Returns:

this

Since:

19.4

useEdgeToEdgeConnection

public GraphWiseConvLayerConfigBuilder useEdgeToEdgeConnection(boolean edgeToEdgeConnection)

Use the connection between edges and edges. Should be set to True if neighboring edges should be sampled to compute the edge embeddings of that layer.

default: GraphWiseConvLayerConfig#DEFAULT_EDGE_EDGE_CONNECTION

Parameters:

edgeToEdgeConnection - boolean flag to use edge to edge connection

Returns:

this

Since:

22.2

useEdgeToVertexConnection

public GraphWiseConvLayerConfigBuilder useEdgeToVertexConnection(boolean edgeToVertexConnection)

Use the connection between edges and vertices. Should be set to True if neighboring edges should be sampled to compute the vertex embeddings of that layer.

default: GraphWiseConvLayerConfig#DEFAULT_EDGE_VERTEX_CONNECTION

Parameters:

edgeToVertexConnection - boolean flag to use edge to vertex connection

Returns:

this

Since:

22.2

useVertexToEdgeConnection

public GraphWiseConvLayerConfigBuilder useVertexToEdgeConnection(boolean vertexToEdgeConnection)

Use the connection between vertices and edges. Should be set to True if neighboring vertices should be sampled to compute the edges embeddings of that layer.

default: GraphWiseConvLayerConfig#DEFAULT_VERTEX_EDGE_CONNECTION

Parameters:

vertexToEdgeConnection - boolean flag to use vertex to edge connection

Returns:

this

Since:

22.2

useVertexToVertexConnection

public GraphWiseConvLayerConfigBuilder useVertexToVertexConnection(boolean vertexToVertexConnection)

Use the connection between vertices and vertices. Should be set to True if neighboring vertices should be sampled to compute the vertex embeddings of that layer.

default: GraphWiseConvLayerConfig#DEFAULT_VERTEX_VERTEX_CONNECTION

Parameters:

vertexToVertexConnection - boolean flag to use vertex to vertex connection

Returns:

this

Since:

22.2