readout.graph_readout

class pydgn.model.readout.graph_readout.LinearGraphReadout(*args: Any, **kwargs: Any)

Class that implements a simple readout mapping for graph prediction

forward(node_embeddings: torch.tensor, batch: torch.Tensor, **kwargs) → Tuple[torch.Tensor, Optional[torch.Tensor], Optional[List[object]]]

Implements a linear graph readout

Parameters

node_embeddings (torch.Tensor) – the node embeddings of size Nxd
batch (torch.Tensor) – a tensor specifying to which graphs nodes belong to in the batch
kwargs (dict) – additional parameters (unused)

Returns

a tuple (output, node_embeddings)

readout.link_readout

class pydgn.model.readout.link_readout.DotProductLinkReadout(*args: Any, **kwargs: Any)

Class that implements a simple readout mapping for link prediction via dot product

forward(node_embeddings: torch.tensor, batch: torch.Tensor, **kwargs) → Tuple[torch.Tensor, Optional[torch.Tensor], Optional[List[object]]]

Implements a dot product scorer for link prediction

Parameters

node_embeddings (torch.Tensor) – the node embeddings of size Nxd
batch (torch.Tensor) – a tensor specifying to which graphs nodes belong to in the batch
kwargs (dict) – additional parameters (unused)

Returns

a tuple (None, node_embeddings, [link scores, dense adjacency matrix])

class pydgn.model.readout.node_readout.LinearNodeReadout(*args: Any, **kwargs: Any)

Class that implements a simple readout mapping for node prediction

forward(node_embeddings: torch.tensor, batch: torch.Tensor, **kwargs) → Tuple[torch.Tensor, Optional[torch.Tensor], Optional[List[object]]]

Implements a linear readout of node embeddings

Parameters

node_embeddings (torch.Tensor) – the node embeddings of size Nxd
batch (torch.Tensor) – a tensor specifying to which graphs nodes belong to in the batch
kwargs (dict) – additional parameters (unused)

Returns

a tuple (output, node_embeddings)