oracle.pgx.api.mllib

Class GraphWiseModelBuilder<Model extends GraphWiseModel,Config extends GraphWiseModelConfig,Self extends GraphWiseModelBuilder>

java.lang.Object

oracle.pgx.api.mllib.GraphWiseModelBuilder<Model,Config,Self>

Direct Known Subclasses:

SupervisedGraphWiseModelBuilder, UnsupervisedGraphWiseModelBuilder

public abstract class GraphWiseModelBuilder<Model extends GraphWiseModel,Config extends GraphWiseModelConfig,Self extends GraphWiseModelBuilder>
extends java.lang.Object

Abstract builder class for GraphWise models. Use SupervisedGraphWiseModelBuilder instead of this.

Since:

19.4

Constructor Summary

Constructors

Constructor and Description
GraphWiseModelBuilder()

Method Summary

Modifier and Type	Method and Description
abstract Model	build() Builds the GraphWise model with the specified parameters.
Self	setBatchSize(int batchSize) Set the batch size.
Self	setConvLayerConfigs(GraphWiseConvLayerConfig... layerConfigs) Set the convolutional layer configurations (See GraphWiseConvLayerConfig).
Self	setEdgeInputPropertyNames(java.util.List<java.lang.String> edgeInputPropertyNames) Set the edge input features for the algorithm in the form of names of properties on the graph.
Self	setEdgeInputPropertyNames(java.lang.String... edgeInputPropertyNames) Set the edges input features for the algorithm in the form of names of properties on the graph.
Self	setEmbeddingDim(int embeddingDim) Set the embedding dimension of the model.
Self	setLearningRate(double learningRate) Set the learning rate used for the Adam optimizer
Self	setNumEpochs(int numEpochs) Set the number of training epochs.
Self	setSeed(java.lang.Integer seed) Sets the random seed used for fitting the model.
Self	setShuffle(boolean shuffle) If true, the model will shuffle the order of vertices before each epoch.
Self	setStandardize(boolean standardize) Whether to standardize the batch before passing them through a convolutional layer.
Self	setTargetVertexLabels(java.util.List<java.lang.String> targetVertexLabels) Set the target vertex labels for the algorithm.
Self	setTargetVertexLabels(java.lang.String... targetVertexLabels) Set the target vertex labels for the algorithm.
Self	setVertexInputPropertyNames(java.util.List<java.lang.String> vertexInputPropertyNames) Set the input features for the algorithm in the form of names of properties on the graph.
Self	setVertexInputPropertyNames(java.lang.String... vertexInputPropertyNames) Set the input features for the algorithm in the form of names of properties on the graph.
Self	setWeightDecay(double weightDecay) Set the weight decay used for the Adam optimizer

Methods inherited from class java.lang.Object

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

Constructor Detail

GraphWiseModelBuilder

public GraphWiseModelBuilder()

Method Detail

build

public abstract Model build()
                     throws java.lang.InterruptedException,
                            java.util.concurrent.ExecutionException

Builds the GraphWise model with the specified parameters.

Returns:

GraphWise model

Throws:

java.lang.InterruptedException

java.util.concurrent.ExecutionException

Since:

19.4

setBatchSize

public Self setBatchSize(int batchSize)

Set the batch size. This batch size will be used for both the training and inference.

default: GraphWiseModelConfig.DEFAULT_BATCH_SIZE

Parameters:

batchSize - batch size

Returns:

this

Since:

19.4

setConvLayerConfigs

public Self setConvLayerConfigs(GraphWiseConvLayerConfig... layerConfigs)

Set the convolutional layer configurations (See GraphWiseConvLayerConfig).

default: GraphWiseModelConfig.DEFAULT_CONV_LAYER_CONFIGS

Parameters:

layerConfigs - layer configurations

Returns:

this

Since:

19.4

setEdgeInputPropertyNames

public Self setEdgeInputPropertyNames(java.util.List<java.lang.String> edgeInputPropertyNames)

Set the edge input features for the algorithm in the form of names of properties on the graph. You can set multiple input properties.

Supported property types are given by GraphWiseModelConfig.SUPPORTED_INPUT_TYPES. Additionally, float and double vector properties are supported.

Parameters:

edgeInputPropertyNames - list of property names

Returns:

this

Since:

21.2

setEdgeInputPropertyNames

public Self setEdgeInputPropertyNames(java.lang.String... edgeInputPropertyNames)

Set the edges input features for the algorithm in the form of names of properties on the graph. You can set multiple input properties.

Supported graph property types are:

• double
• float
• double vectors
• float vectors

Parameters:

edgeInputPropertyNames - property names

Returns:

this

Since:

21.2

setEmbeddingDim

public Self setEmbeddingDim(int embeddingDim)

Set the embedding dimension of the model. This is the dimension of the embeddings produced by each of the convolutional layers.

default: GraphWiseModelConfig.DEFAULT_EMBEDDING_DIM

Parameters:

embeddingDim - embedding dimension

Returns:

this

Since:

19.4

setLearningRate

public Self setLearningRate(double learningRate)

Set the learning rate used for the Adam optimizer

default: GraphWiseModelConfig.DEFAULT_LEARNING_RATE

Parameters:

learningRate - learning rate

Returns:

this

Since:

19.4

setNumEpochs

public Self setNumEpochs(int numEpochs)

Set the number of training epochs.

default: GraphWiseModelConfig.DEFAULT_NUM_EPOCHS

Parameters:

numEpochs - number of epochs

Returns:

this

Since:

19.4

setSeed

public Self setSeed(java.lang.Integer seed)

Sets the random seed used for fitting the model. If null, the random seed will be chosen randomly.

default: GraphWiseModelConfig.DEFAULT_SEED

Parameters:

seed - random seed

Returns:

this

Since:

19.4

setShuffle

public Self setShuffle(boolean shuffle)

If true, the model will shuffle the order of vertices before each epoch.

default: GraphWiseModelConfig.DEFAULT_SHUFFLE

Parameters:

shuffle - whether to shuffle

Returns:

this

Since:

19.4

setStandardize

public Self setStandardize(boolean standardize)

Whether to standardize the batch before passing them through a convolutional layer.

default: GraphWiseModelConfig.DEFAULT_STANDARDIZE

Parameters:

standardize - whether to standardize

Returns:

this

Since:

19.4

setTargetVertexLabels

public Self setTargetVertexLabels(java.util.List<java.lang.String> targetVertexLabels)

Set the target vertex labels for the algorithm. Only the related vertices need to have the target property. Training and inference will be done on the vertices with those labels

Parameters:

targetVertexLabels - list of label names

Returns:

this

Since:

22.2

setTargetVertexLabels

public Self setTargetVertexLabels(java.lang.String... targetVertexLabels)

Set the target vertex labels for the algorithm. Only the related vertices need to have the target property.

Parameters:

targetVertexLabels - label names

Returns:

this

Since:

22.2

setVertexInputPropertyNames

public Self setVertexInputPropertyNames(java.util.List<java.lang.String> vertexInputPropertyNames)

Set the input features for the algorithm in the form of names of properties on the graph. You can set multiple input properties.

Supported property types are given by GraphWiseModelConfig.SUPPORTED_INPUT_TYPES. Additionally, float and double vector properties are supported.

Parameters:

vertexInputPropertyNames - list of property names

Returns:

this

Since:

19.4

setVertexInputPropertyNames

public Self setVertexInputPropertyNames(java.lang.String... vertexInputPropertyNames)

Set the input features for the algorithm in the form of names of properties on the graph. You can set multiple input properties.

Supported graph property types are:

• double
• float
• double vectors
• float vectors

Parameters:

vertexInputPropertyNames - property names

Returns:

this

Since:

19.4

setWeightDecay

public Self setWeightDecay(double weightDecay)

Set the weight decay used for the Adam optimizer

default: GraphWiseModelConfig.DEFAULT_WEIGHT_DECAY

Parameters:

weightDecay - weight decay

Returns:

this

Since:

22.2