GOOD.networks.models.BaseGNN

Base classes for Graph Neural Networks

Classes

BasicEncoder(config, **kwargs)	Base GNN feature encoder.
GNNBasic(config, *args, **kwargs)	Base class for graph neural networks

class GOOD.networks.models.BaseGNN.BasicEncoder(config: Union[CommonArgs, Munch], **kwargs)

Bases: Module

Base GNN feature encoder.

Parameters

config (Union[CommonArgs, Munch]) – munchified dictionary of args (config.model.dim_hidden, config.model.model_layer, config.model.model_level, config.model.global_pool, config.model.dropout_rate)

config = munchify({model: {dim_hidden: int(300),
                   model_layer: int(5),
                   model_level: str('node'),
                   global_pool: str('mean'),
                   dropout_rate: float(0.5),}
                   })

class GOOD.networks.models.BaseGNN.GNNBasic(config: Union[CommonArgs, Munch], *args, **kwargs)

Bases: Module

Base class for graph neural networks

Parameters

• *args (list) – argument list for the use of arguments_read()

• **kwargs (dict) – key word arguments for the use of arguments_read()

arguments_read(*args, **kwargs)

It is an argument reading function for diverse model input formats. Support formats are: model(x, edge_index) model(x, edge_index, batch) model(data=data).

Notes

edge_weight is optional for node prediction tasks.

Parameters

• *args – [x, edge_index, [batch]]

• **kwargs – data, [edge_weight]

Returns

Unpacked node features, sparse adjacency matrices, batch indicators, and optional edge weights.