MultiLayerPerceptron-AS node Build Options

Use the Build Options tab to specify build options for the MultiLayerPerceptron-AS node, including performance, modeling building, and expert options. For more information about these options, see http://spark.apache.org/docs/latest/api/java/org/apache/spark/ml/classification/MultilayerPerceptronClassifier.html.¹

Performance

Perceptrons Layer. Use this setting to define the number of perceptron layers to include. This value must be larger than the number of perceptron fields. The default value is 1.

Hidden Layers. Specify the number of hidden layers. Use a comma between multiple hidden layers. The default value is 1.

Output Layer. Specify the number of output layers. The default value is 1.

Random Seed. Click Generate if you want to generate the seed used by the random number generator.

Model Building

Max iterations. Specify the maximum number of iterations to perform. The default value is 10.

Expert Only

BlockSize. Select the Expert Mode option in the Model Building section if you want to specify the block size for stacking input data in matrices. This can speed up the computation. The default block size is 128.

The following table shows the relationship between the settings in the SPSS® Modeler MultiLayerPerceptron-AS node dialogs and the Spark KDE library parameters.

Table 1. Node properties mapped to Spark parameters
SPSS Modeler setting	Script name (property name)	Spark parameter
Predictors	`features`
Target	`label`
Perceptrons Layer	`layers[0]`	`layers[0]`
Hidden Layers	`layers[1…<latest-1>]`	`layers[1…<latest-1>]`
Output Layer	`layers[<latest>]`	`layers[<latest>]`
Random Seed	`seed`	`seed`
Max iterations	`maxiter`	`maxiter`

¹ "Class MultilayerPerceptronClassifier." Apache Spark. JavaDoc. Web. 5 Oct 2018.