models.nn.pyg_infograph

Module Contents

Classes

SUPEncoder	Encoder used in supervised model with Set2set in paper `”Order Matters: Sequence to sequence for sets”
Encoder	Encoder stacked with GIN layers
FF	Residual MLP layers.
InfoGraph	Implimentation of Infograph in paper `”InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation

class models.nn.pyg_infograph.SUPEncoder(num_features, dim, num_layers=1)

Bases: torch.nn.Module

Encoder used in supervised model with Set2set in paper “Order Matters: Sequence to sequence for sets” <https://arxiv.org/abs/1511.06391> and NNConv in paper “Dynamic Edge-Conditioned Filters in Convolutional Neural Networks on Graphs” <https://arxiv.org/abs/1704.02901>

forward(self, x, edge_index, batch, edge_attr)

class models.nn.pyg_infograph.Encoder(in_feats, hidden_dim, num_layers=3, num_mlp_layers=2, pooling='sum')

Bases: torch.nn.Module

Encoder stacked with GIN layers

in_featsint

Size of each input sample.

hidden_featsint

Size of output embedding.

num_layersint, optional

Number of GIN layers, default: 3.

num_mlp_layersint, optional

Number of MLP layers for each GIN layer, default: 2.

poolingstr, optional

Aggragation type, default : sum.

forward(self, x, edge_index, batch, *args)

class models.nn.pyg_infograph.FF(in_feats, out_feats)

Bases: torch.nn.Module

Residual MLP layers.

..math::

out = mathbf{MLP}(x) + mathbf{Linear}(x)

in_featsint

Size of each input sample

out_featsint

Size of each output sample

forward(self, x)

class models.nn.pyg_infograph.InfoGraph(in_feats, hidden_dim, out_feats, num_layers=3, unsup=True)

Bases: models.BaseModel

Implimentation of Infograph in paper `”InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation

Learning via Mutual Information Maximization” <https://openreview.net/forum?id=r1lfF2NYvH>__. `

in_featsint

Size of each input sample.

out_featsint

Size of each output sample.

num_layersint, optional

Number of MLP layers in encoder, default: 3.

unsupbool, optional

Use unsupervised model if True, default: True.

static add_args(parser)

Add model-specific arguments to the parser.

classmethod build_model_from_args(cls, args)

Build a new model instance.

classmethod split_dataset(cls, dataset, args)

reset_parameters(self)

forward(self, batch)

sup_forward(self, x, edge_index=None, batch=None, label=None, edge_attr=None)

unsup_forward(self, x, edge_index=None, batch=None)

sup_loss(self, prediction, label=None)

unsup_loss(self, x, edge_index=None, batch=None)

unsup_sup_loss(self, x, edge_index, batch)

static mi_loss(pos_mask, neg_mask, mi, pos_div, neg_div)