API

The API of PHOTONAI Graph is based on the PHOTONAI package. It provides additional methods that can be used for graph machine learning, and a few utility functions for handling the conversions of graphs as well as writing and saving them. The different methods are implemented so that they can be used as elements of a photon pipeline or on their own, as all elements are also sklearn compatible as they implement fit/transform/predict functions. There is a certain workflow so that the graph machine learning steps work together. Each implemented method is either a transformer or a predictor.

Transformer Classes

Transformer objects are performing transformations of the input data. Therefore each transformer has to provide either a fit() and a transform() function or a fit_transform() function.

classDiagram-v2
    class Transformer{
    +fit()
    +transform()
    +fit_transform()
    }
    Transformer <|-- Graph Constructor
    Transformer <|-- Graph Measures
    Transformer <|-- Graph Kernel
    Transformer <|-- Graph Embedding

Predictor Classes

Predictor objects are the actual learning algorithm of the respective pipeline. Therefore each predictor has to provide at least a fit() and a predict() function.

PHOTONAI Graph provides predefined models for classification and regression.

Classification

classDiagram-v2
    class Predictor{
    +fit()
    +predict()
    }
    Predictor <|-- DglModel  
    DglModel <|-- GCNClassifierModel
    DglModel <|-- SGConvClassifierModel
    DglModel <|-- GATClassifierModel

Regression

classDiagram-v2
    class Predictor{
    +fit()
    +predict()
    }
    Predictor <|-- DglModel 
    DglModel <|-- GCNRegressorModel
    DglModel <|-- SGConvRegressorModel
    DglModel <|-- GATRegressorModel