Advanced Hyperparameter Customization

17.6.3 Advanced Hyperparameter Customization

You can build an Unsupervised Anomaly Detection GraphWise model using rich hyperparameter customization.

This is implemented using the sub-config classes, GraphWiseConvLayerConfig and GraphWiseEmbeddingConfig.

Also, you can enable or disable a graphics processing unit (GPU) by using the enable_ml_accelerators graph server (PGX) configuration parameter (see Configuration Parameters for the Graph Server (PGX) Engine for more information). In addition, ensure that your system meets the following prerequisites to use the GPU support:

You must have a GPU device with the CUDA (Compute Unified Device Architecture) toolkit installed.
The following list of CUDA libraries are expected:
- libcuda.so.1
- libnvrtc.so.12
- libnvToolsExt.so.1
- libcudart.so.12

Note that the enable_ml_accelerators option is enabled by default. But if a GPU device is not detected and the CUDA toolkit is not installed, then this feature gets disabled and the CPU will be used for all the PgxML library operations.

The following example specifies a weight decay parameter of 0.001 and dropout with dropping probability 0.5 for the model to counteract overfitting. The Dominant embedding layer's alpha value is specified as 0.6 to slightly increase the importance of the feature reconstruction. Also, note that the setEnableAccelerator method is enabled to use any available GPU.

opg4j> var weightProperty = analyst.pagerank(trainGraph).getName()
opg4j> var convLayerConfig = analyst.graphWiseConvLayerConfigBuilder(). // customize convolutional layer config
         setNumSampledNeighbors(25).
         setActivationFunction(ActivationFunction.TANH).
         setWeightInitScheme(WeightInitScheme.XAVIER).
         setWeightedAggregationProperty(weightProperty).
         setDropoutRate(0.5). // set dropout rate to prevent overfitting
         build()

opg4j> var predictionLayerConfig = analyst.graphWisePredictionLayerConfigBuilder().
        setHiddenDimension(8).
        setActivationFunction(ActivationFunction.RELU).
        build()

opg4j> var dominantConfig = analyst.graphWiseDominantLayerConfigBuilder(). // customize embedding layer config
        setDecoderLayerConfigs(predictionLayerConfig).
        setAlpha(0.6). // increase the importance of feature reconstruction
        build()

opg4j> var model = analyst.unsupervisedAnomalyDetectionGraphWiseModelBuilder().
         setVertexInputPropertyNames("vertex_features").
         setConvLayerConfigs(convLayerConfig).
         setEmbeddingConfig(dominantConfig).
         setWeightDecay(0.001). // set weight decay to prevent overfitting
         setEmbeddingDim(256).
         setLearningRate(0.05).
         setNumEpochs(30).
         setSeed(42).
         setShuffle(false).
         setStandardize(true).
         setBatchSize(64).
         setEnableAccelerator(true). // Enable or disable GPU
         build()

// customize convolutional layer config
String weightProperty = analyst.pagerank(trainGraph).getName()
GraphWiseConvLayerConfig convLayerConfig = analyst.graphWiseConvLayerConfigBuilder()
    .setNumSampledNeighbors(25)
    .setActivationFunction(ActivationFunction.TANH)
    .setWeightInitScheme(WeightInitScheme.XAVIER)
    .setWeightedAggregationProperty(weightProperty)
    .setDropoutRate(0.5) // set dropout rate to prevent overfitting
    .build();

GraphWisePredictionLayerConfig predictionLayerConfig = analyst.graphWisePredictionLayerConfigBuilder()
    .setHiddenDimension(8)
    .setActivationFunction(ActivationFunction.RELU)
    .build();

// customize embedding layer config
GraphWiseEmbeddingConfig dominantConfig = analyst.graphWiseDominantLayerConfigBuilder()
    .setDecoderLayerConfigs(predictionLayerConfig)
    .setAlpha(0.6) // increase the importance of feature reconstruction
    .build();

// build the anomaly detection model
UnsupervisedAnomalyDetectionGraphWiseModel model = analyst.unsupervisedAnomalyDetectionGraphWiseModelBuilder()
    .setVertexInputPropertyNames("vertex_features")
    .setEmbeddingConfig(dominantConfig)
    .setConvLayerConfigs(convLayerConfig)
    .setWeightDecay(0.001) // set weight decay to prevent overfitting
    .setEmbeddingDim(256)
    .setLearningRate(0.05)
    .setNumEpochs(30)
    .setSeed(42)
    .setShuffle(false)
    .setStandardize(true)
    .setBatchSize(64)
    .setEnableAccelerator(true) // Enable or disable GPU
    .build();

# customize convolutional layer config
weightProperty = analyst.pagerank(train_graph).name

conv_layer_config = dict(num_sampled_neighbors=25,
                         activation_fn='tanh',
                         weight_init_scheme='xavier',
                         neighbor_weight_property_name=weightProperty,
                         dropout_rate=0.5) # set dropout rate to prevent overfitting
conv_layer = analyst.graphwise_conv_layer_config(**conv_layer_config)

# customize embedding layer config
dominant_config = dict(alpha=0.6) # increase the importance of feature reconstruction

dominant_layer = analyst.graphwise_dominant_layer_config(**dominant_config)

# build the anomaly detection model
params = dict(conv_layer_config=[conv_layer],
              embedding_config=dominant_layer,
              vertex_input_property_names=["vertex_features"],
              weight_decay=0.001, # set weight decay to prevent overfitting
              layer_size=256,
              learning_rate=0.05,
              num_epochs=30,
              seed=42,
              standardize=true,
              batch_size=64,
              enable_accelerator=True # Enable or disable GPU
)
model = analyst.unsupervised_anomaly_detection_graphwise_builder(**params)

Parent topic: Using the Unsupervised Anomaly Detection GraphWise Algorithm (Vertex Embeddings and Anomaly Scores)