Package oracle.pgx.api.mllib

Class EdgeWiseModel<Config extends EdgeWiseModelConfig,Metadata extends oracle.pgx.api.internal.mllib.EdgeWiseModelMetadata<Config>,ModelType extends EdgeWiseModel<Config,Metadata,ModelType>>

java.lang.Object

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

oracle.pgx.api.mllib.EdgeWiseModel<Config,Metadata,ModelType>

All Implemented Interfaces:

java.lang.AutoCloseable

Direct Known Subclasses:

SupervisedEdgeWiseModel, UnsupervisedEdgeWiseModel

public abstract class EdgeWiseModel<Config extends EdgeWiseModelConfig,Metadata extends oracle.pgx.api.internal.mllib.EdgeWiseModelMetadata<Config>,ModelType extends EdgeWiseModel<Config,Metadata,ModelType>>
extends Model<ModelType>

Base class for EdgeWiseModels

Since:

23.1

Constructor Summary

Constructors

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

Method Summary

Modifier and Type	Method	Description
void	destroy()	Blocking version of destroyAsync().
PgxFuture<java.lang.Void>	destroyAsync()	Destroys an EdgeWise model
double	fit(PgxGraph graph)	Blocking version of fitAsync(PgxGraph).
double	fit(PgxGraph trainGraph, PgxGraph valGraph)	Blocking version of fitAsync(PgxGraph, PgxGraph).
abstract PgxFuture<java.lang.Double>	fitAsync(PgxGraph graph)	Trains the EdgeWise model on the input graph.
abstract PgxFuture<java.lang.Double>	fitAsync(PgxGraph trainGraph, PgxGraph valGraph)	Trains the EdgeWise model on the input trainGraph and evaluate on the input valGraph.
int	getBatchSize()	Gets the batch size
Config	getConfig()	Gets the model configuration object
GraphWiseBaseConvLayerConfig[]	getConvLayerConfigs()	Gets the configuration objects for the convolutional layers
EdgeCombinationMethod	getEdgeCombinationMethod()	Gets the edge embedding method used to produce the embedding
int	getEdgeInputFeatureDim()	Gets the edges input feature dimension, that is, the dimension of all the input edge properties when concatenated
java.util.List<java.lang.String>	getEdgeInputPropertyNames()	Gets the edges input feature names
int	getEmbeddingDim()	Gets the dimension of the embeddings
int	getInputFeatureDim()	Gets the input feature dimension, that is, the dimension of all the input vertex properties when concatenated
double	getLearningRate()	Gets the initial learning rate
int	getNumEpochs()	Gets the number of epochs to train the model
java.lang.Integer	getSeed()	Gets the random seed
PgxFrame	getTrainingLog()	Blocking version of getTrainingLogAsync().
abstract PgxFuture<PgxFrame>	getTrainingLogAsync()	Get the training log that has evaluation results from validation.
double	getTrainingLoss()	Gets the final training loss
java.util.List<java.lang.String>	getVertexInputPropertyNames()	Gets the vertices input feature names
PgxFrame	inferEmbeddings(PgxGraph graph, java.lang.Iterable<PgxEdge> edges)	Blocking version of inferEmbeddingsAsync(PgxGraph, Iterable).
abstract PgxFuture<PgxFrame>	inferEmbeddingsAsync(PgxGraph graph, java.lang.Iterable<PgxEdge> edges)	Infers the embeddings for the specified edges.
boolean	isFitted()	Checks if the model is fitted

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

Constructor Detail

EdgeWiseModel

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

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

Parameters:

session - PgxSession to which the model is connected

core - Core to which the model is connected

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

Since:

23.1

Method Detail

destroyAsync

public PgxFuture<java.lang.Void> destroyAsync()

Destroys an EdgeWise model

Specified by:

destroyAsync in class Model<ModelType extends EdgeWiseModel<Config,Metadata,ModelType>>

Returns:

a future which will be completed once the destruction request finishes.

Since:

23.1

destroy

public void destroy()
             throws java.util.concurrent.ExecutionException,
                    java.lang.InterruptedException

Blocking version of destroyAsync(). Calls destroyAsync() 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:

23.1

getNumEpochs

public int getNumEpochs()

Gets the number of epochs to train the model

Returns:

number of epochs to train the model

Since:

23.1

getLearningRate

public double getLearningRate()

Gets the initial learning rate

Returns:

initial learning rate

Since:

23.1

getBatchSize

public int getBatchSize()

Gets the batch size

Returns:

batch size

Since:

23.1

getEmbeddingDim

public int getEmbeddingDim()

Gets the dimension of the embeddings

Returns:

embedding dimension

Since:

23.1

getSeed

public java.lang.Integer getSeed()

Gets the random seed

Returns:

random seed

Since:

23.1

getConvLayerConfigs

public GraphWiseBaseConvLayerConfig[] getConvLayerConfigs()

Gets the configuration objects for the convolutional layers

Returns:

configurations

Since:

23.1

getVertexInputPropertyNames

public java.util.List<java.lang.String> getVertexInputPropertyNames()

Gets the vertices input feature names

Returns:

vertices input feature names

Since:

23.1

getEdgeInputPropertyNames

public java.util.List<java.lang.String> getEdgeInputPropertyNames()

Gets the edges input feature names

Returns:

edges input feature names

Since:

23.1

isFitted

public boolean isFitted()

Checks if the model is fitted

Returns:

true if the model is fitted

Since:

23.1

getTrainingLoss

public double getTrainingLoss()

Gets the final training loss

Returns:

training loss

Since:

23.1

getInputFeatureDim

public int getInputFeatureDim()

Gets the input feature dimension, that is, the dimension of all the input vertex properties when concatenated

Returns:

input feature dimension

Since:

23.1

getEdgeInputFeatureDim

public int getEdgeInputFeatureDim()

Gets the edges input feature dimension, that is, the dimension of all the input edge properties when concatenated

Returns:

edges input feature dimension

Since:

23.1

getEdgeCombinationMethod

public EdgeCombinationMethod getEdgeCombinationMethod()

Gets the edge embedding method used to produce the embedding

Returns:

edge embedding method

Since:

23.1

getConfig

public Config getConfig()

Gets the model configuration object

Returns:

model configuration

Since:

23.1

getTrainingLogAsync

public abstract PgxFuture<PgxFrame> getTrainingLogAsync()

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

Returns:

training log

Since:

24.2

getTrainingLog

public PgxFrame getTrainingLog()
                        throws java.util.concurrent.ExecutionException,
                               java.lang.InterruptedException

Blocking version of getTrainingLogAsync(). Calls getTrainingLogAsync() and waits for the returned PgxFuture to complete.

Returns:

training log

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:

24.2

fitAsync

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

Trains the EdgeWise model on the input graph.

Parameters:

graph - input graph to fit on.

Since:

23.1

fitAsync

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

Trains the EdgeWise model on the input trainGraph and evaluate on the input valGraph.

Parameters:

trainGraph - input graph to fit on.

valGraph - input graph to evaluate on for validation.

Since:

24.2

fit

public double fit(PgxGraph graph)
           throws java.util.concurrent.ExecutionException,
                  java.lang.InterruptedException

Blocking version of fitAsync(PgxGraph). Calls fitAsync(PgxGraph) and waits for the returned PgxFuture to complete.

Parameters:

graph - input graph to fit on.

Returns:

the training loss of the last batch

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:

23.1

fit

public double fit(PgxGraph trainGraph,
                  PgxGraph valGraph)
           throws java.util.concurrent.ExecutionException,
                  java.lang.InterruptedException

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

Parameters:

trainGraph - input graph to fit on.

valGraph - input graph to evaluate on for validation.

Returns:

the training loss of the last batch

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:

24.2

inferEmbeddingsAsync

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

Infers the embeddings for the specified edges.

Returns:

PgxFrame containing the embeddings for each edge.

Since:

23.1

inferEmbeddings

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

Blocking version of inferEmbeddingsAsync(PgxGraph, Iterable). Infers the embeddings for the specified edges.

Returns:

PgxFrame containing the embeddings for each edge.

Since:

23.1