Package oracle.pgx.api.mllib

Class UnsupervisedEdgeWiseModel

java.lang.Object

oracle.pgx.api.mllib.Model<ModelType>

oracle.pgx.api.mllib.EdgeWiseModel<UnsupervisedEdgeWiseModelConfig,oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata,UnsupervisedEdgeWiseModel>

oracle.pgx.api.mllib.UnsupervisedEdgeWiseModel

All Implemented Interfaces:

java.lang.AutoCloseable

public class UnsupervisedEdgeWiseModel
extends EdgeWiseModel<UnsupervisedEdgeWiseModelConfig,oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata,UnsupervisedEdgeWiseModel>

A variant of GraphSAGE working on edges. See UnsupervisedEdgeWiseModelBuilder for documentation of the hyperparameters.

Since:

23.2

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	UnsupervisedEdgeWiseModel.UnsupervisedEdgeWiseInferenceType

Field Summary

Fields

Modifier and Type	Field	Description
static java.lang.String	ALGORITHM_NAME

Constructor Summary

Constructors

Constructor	Description
UnsupervisedEdgeWiseModel(PgxSession session, oracle.pgx.api.internal.Core core, java.util.function.Supplier<java.lang.String> keystorePathSupplier, java.util.function.Supplier<char[]> keystorePasswordSupplier, java.util.function.BiFunction<PgxSession,oracle.pgx.api.internal.Graph,PgxGraph> graphConstructor, oracle.pgx.api.internal.mllib.ModelMetadata modelMetadata)
UnsupervisedEdgeWiseModel(PgxSession session, oracle.pgx.api.internal.Core core, java.util.function.Supplier<java.lang.String> keystorePathSupplier, java.util.function.Supplier<char[]> keystorePasswordSupplier, java.util.function.BiFunction<PgxSession,oracle.pgx.api.internal.Graph,PgxGraph> graphConstructor, oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata modelMetadata)	This constructor should never be used to get a model.

Method Summary

Modifier and Type	Method	Description
PgxFuture<java.lang.Double>	fitAsync(PgxGraph graph)	Trains the EdgeWise model on the input graph.
PgxFuture<java.lang.Double>	fitAsync(PgxGraph trainGraph, PgxGraph valGraph)	Trains the EdgeWise model on the input graph.
GraphWiseDgiLayerConfig	getDgiLayerConfigs()	Gets the configuration object for the dgi layer
UnsupervisedEdgeWiseModelConfig.LossFunction	getLossFunction()	Gets the loss function
java.util.List<java.util.Set<java.lang.String>>	getTargetEdgeLabels()	Gets the target vertex labels
PgxFuture<PgxFrame>	getTrainingLogAsync()	Gets the training log that has evaluation results from validation.
PgxFuture<PgxFrame>	inferEmbeddingsAsync(PgxGraph graph, java.lang.Iterable<PgxEdge> edges)	Infers the embeddings for the specified edges.
void	store(java.lang.String path, java.lang.String key)	Blocking version of storeAsync(String, String).
void	store(java.lang.String path, java.lang.String key, boolean overwrite)	Blocking version of storeAsync(String, String).
PgxFuture<java.lang.Void>	storeAsync(java.lang.String path, java.lang.String key)	Stores the GraphWise model in the specified path, with encryption.
PgxFuture<java.lang.Void>	storeAsync(java.lang.String path, java.lang.String key, boolean overwrite)	Stores the GraphWise model in the specified path, with encryption.

Methods inherited from class oracle.pgx.api.mllib.EdgeWiseModel

destroy, destroyAsync, fit, fit, getBatchSize, getConfig, getConvLayerConfigs, getEdgeCombinationMethod, getEdgeInputFeatureDim, getEdgeInputPropertyNames, getEmbeddingDim, getInputFeatureDim, getLearningRate, getNumEpochs, getSeed, getTrainingLog, getTrainingLoss, getVertexInputPropertyNames, inferEmbeddings, isFitted

Methods inherited from class oracle.pgx.api.mllib.Model

close, export

Methods inherited from class java.lang.Object

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

Field Detail

ALGORITHM_NAME

public static final java.lang.String ALGORITHM_NAME

See Also:

Constant Field Values

Constructor Detail

UnsupervisedEdgeWiseModel

public UnsupervisedEdgeWiseModel(PgxSession session,
                                 oracle.pgx.api.internal.Core core,
                                 java.util.function.Supplier<java.lang.String> keystorePathSupplier,
                                 java.util.function.Supplier<char[]> keystorePasswordSupplier,
                                 java.util.function.BiFunction<PgxSession,oracle.pgx.api.internal.Graph,PgxGraph> graphConstructor,
                                 oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata modelMetadata)

This constructor should never be used to get a model. Use UnsupervisedEdgeWiseModelBuilder instead.

Parameters:

session - PgxSession to which the model is connected

core - Core to which the model is connected

graphConstructor - Constructor for a PgxGraph

modelMetadata - Metadata concerning the different hyper-parameters of the EdgeWise Model

Since:

23.2

UnsupervisedEdgeWiseModel

public UnsupervisedEdgeWiseModel(PgxSession session,
                                 oracle.pgx.api.internal.Core core,
                                 java.util.function.Supplier<java.lang.String> keystorePathSupplier,
                                 java.util.function.Supplier<char[]> keystorePasswordSupplier,
                                 java.util.function.BiFunction<PgxSession,oracle.pgx.api.internal.Graph,PgxGraph> graphConstructor,
                                 oracle.pgx.api.internal.mllib.ModelMetadata modelMetadata)

Method Detail

fitAsync

public PgxFuture<java.lang.Double> fitAsync(PgxGraph graph)

Trains the EdgeWise model on the input graph.

Specified by:

fitAsync in class EdgeWiseModel<UnsupervisedEdgeWiseModelConfig,oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata,UnsupervisedEdgeWiseModel>

Parameters:

graph - input graph to fit on.

Since:

23.2

fitAsync

public PgxFuture<java.lang.Double> fitAsync(PgxGraph trainGraph,
                                            PgxGraph valGraph)

Trains the EdgeWise model on the input graph.

Specified by:

fitAsync in class EdgeWiseModel<UnsupervisedEdgeWiseModelConfig,oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata,UnsupervisedEdgeWiseModel>

Parameters:

trainGraph - input train graph to fit on.

valGraph - input graph to evaluate on for validation.

Since:

24.2

getTrainingLogAsync

public PgxFuture<PgxFrame> getTrainingLogAsync()

Gets the training log that has evaluation results from validation. It is available only after the model was trained with validation.

Specified by:

getTrainingLogAsync in class EdgeWiseModel<UnsupervisedEdgeWiseModelConfig,oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata,UnsupervisedEdgeWiseModel>

Returns:

training log

Since:

24.2

inferEmbeddingsAsync

public PgxFuture<PgxFrame> inferEmbeddingsAsync(PgxGraph graph,
                                                java.lang.Iterable<PgxEdge> edges)

Infers the embeddings for the specified edges.

Specified by:

inferEmbeddingsAsync in class EdgeWiseModel<UnsupervisedEdgeWiseModelConfig,oracle.pgx.api.internal.mllib.UnsupervisedEdgeWiseModelMetadata,UnsupervisedEdgeWiseModel>

Parameters:

graph - the input graph

edges - the edges to produce embeddings to

Returns:

PgxFrame containing the embeddings for each edge.

Since:

23.2

storeAsync

public PgxFuture<java.lang.Void> storeAsync(java.lang.String path,
                                            java.lang.String key)
                                     throws java.util.concurrent.ExecutionException,
                                            java.lang.InterruptedException

Stores the GraphWise model in the specified path, with encryption.

Parameters:

path - path to store the model

key - the encryption key, or null if no encryption should be used.

Throws:

java.util.concurrent.ExecutionException

java.lang.InterruptedException

Since:

23.2

storeAsync

public PgxFuture<java.lang.Void> storeAsync(java.lang.String path,
                                            java.lang.String key,
                                            boolean overwrite)

Stores the GraphWise model in the specified path, with encryption.

Parameters:

path - path to store the model

key - the encryption key, or null if no encryption should be used.

Since:

23.2

store

public void store(java.lang.String path,
                  java.lang.String key)
           throws java.util.concurrent.ExecutionException,
                  java.lang.InterruptedException

Blocking version of storeAsync(String, String). Calls storeAsync(String, String) and waits for the returned PgxFuture to complete.

Throws:

java.lang.InterruptedException - if the caller thread gets interrupted while waiting for completion.

java.util.concurrent.ExecutionException - if any exception occurred during asynchronous execution. The actual exception will be nested.

Since:

23.2

store

public void store(java.lang.String path,
                  java.lang.String key,
                  boolean overwrite)
           throws java.util.concurrent.ExecutionException,
                  java.lang.InterruptedException

Blocking version of storeAsync(String, String). Calls storeAsync(String, String) and waits for the returned PgxFuture to complete.

Throws:

java.lang.InterruptedException - if the caller thread gets interrupted while waiting for completion.

java.util.concurrent.ExecutionException - if any exception occurred during asynchronous execution. The actual exception will be nested.

Since:

23.2

getLossFunction

public UnsupervisedEdgeWiseModelConfig.LossFunction getLossFunction()

Gets the loss function

Returns:

loss function

Since:

23.2

getTargetEdgeLabels

public java.util.List<java.util.Set<java.lang.String>> getTargetEdgeLabels()

Gets the target vertex labels

Returns:

target vertex labels

Since:

23.2

getDgiLayerConfigs

public GraphWiseDgiLayerConfig getDgiLayerConfigs()

Gets the configuration object for the dgi layer

Returns:

configuration

Since:

23.2