Features

Building custom networks

• Building SPNs from single operations or blocks (layers)of operations
• Support for categorical variables and continuous variables
• Realizing joint and conditional distributions
• Realizing MPNs and SPNs
• Adding explicit latent variables
• Building convolutional SPNs with a layer-oriented interface

Automatic SPN validity checking and scope calculation

SPN structure generation

• Generating dense random SPNs of varying complexity

Inference

• Marginal and MPE inference
• Inferring values of explicit latent variables (often used for classification)
• Sampling

Learning

• Batch and online learning
• Hard Expectation-Maximization learning
• Gradient descent learning using TensorFlow optimization

Saving and loading of trained SPNs

SPN network visualization