"""Torch module for SGCN."""
import torch
import torch.nn as nn
import torch.nn.functional as F
from grb.utils.normalize import GCNAdjNorm
[docs]class SGCN(nn.Module):
r"""
Description
-----------
Simplifying Graph Convolutional Networks (`SGCN <https://arxiv.org/abs/1902.07153>`__)
Parameters
----------
in_features : int
Dimension of input features.
out_features : int
Dimension of output features.
hidden_features : int or list of int
Dimension of hidden features. List if multi-layer.
n_layers : int
Number of layers.
layer_norm : bool, optional
Whether to use layer normalization. Default: ``False``.
activation : func of torch.nn.functional, optional
Activation function. Default: ``torch.tanh``.
k : int, optional
Hyper-parameter, refer to original paper. Default: ``4``.
feat_norm : str, optional
Type of features normalization, choose from ["arctan", "tanh", None]. Default: ``None``.
adj_norm_func : func of utils.normalize, optional
Function that normalizes adjacency matrix. Default: ``GCNAdjNorm``.
dropout : float, optional
Rate of dropout. Default: ``0.0``.
"""
def __init__(self,
in_features,
out_features,
hidden_features,
n_layers,
activation=torch.tanh,
feat_norm=None,
adj_norm_func=GCNAdjNorm,
layer_norm=False,
batch_norm=False,
k=4,
dropout=0.0):
super(SGCN, self).__init__()
self.in_features = in_features
self.out_features = out_features
self.feat_norm = feat_norm
self.adj_norm_func = adj_norm_func
if type(hidden_features) is int:
hidden_features = [hidden_features] * (n_layers - 1)
elif type(hidden_features) is list or type(hidden_features) is tuple:
assert len(hidden_features) == (n_layers - 1), "Incompatible sizes between hidden_features and n_layers."
if batch_norm:
self.batch_norm = nn.BatchNorm1d(in_features)
else:
self.batch_norm = None
self.in_conv = nn.Linear(in_features, hidden_features[0])
self.out_conv = nn.Linear(hidden_features[-1], out_features)
self.activation = activation
self.layers = nn.ModuleList()
for i in range(n_layers - 2):
if layer_norm:
self.layers.append(nn.LayerNorm(hidden_features[i]))
self.layers.append(SGConv(in_features=hidden_features[i],
out_features=hidden_features[i + 1],
k=k))
if dropout > 0.0:
self.dropout = nn.Dropout(dropout)
else:
self.dropout = None
@property
def model_type(self):
"""Indicate type of implementation."""
return "torch"
@property
def model_name(self):
return "sgcn"
[docs] def forward(self, x, adj):
r"""
Parameters
----------
x : torch.Tensor
Tensor of input features.
adj : torch.SparseTensor
Sparse tensor of adjacency matrix.
Returns
-------
x : torch.Tensor
Output of model (logits without activation).
"""
if self.batch_norm is not None:
x = self.batch_norm(x)
x = self.in_conv(x)
x = F.relu(x)
if self.dropout is not None:
x = self.dropout(x)
for layer in self.layers:
if isinstance(layer, nn.LayerNorm):
x = layer(x)
else:
x = layer(x, adj)
if self.activation is not None:
x = self.activation(x)
if self.dropout is not None:
x = self.dropout(x)
x = self.out_conv(x)
return x
[docs]class SGConv(nn.Module):
r"""
Description
-----------
SGCN convolutional layer.
Parameters
----------
in_features : int
Dimension of input features.
out_features : int
Dimension of output features.
k : int, optional
Hyper-parameter, refer to original paper. Default: ``4``.
Returns
-------
x : torch.Tensor
Output of layer.
"""
def __init__(self, in_features, out_features, k):
super(SGConv, self).__init__()
self.in_features = in_features
self.out_features = out_features
self.linear = nn.Linear(in_features, out_features)
self.k = k
[docs] def forward(self, x, adj):
r"""
Parameters
----------
x : torch.Tensor
Tensor of input features.
adj : torch.SparseTensor
Sparse tensor of adjacency matrix.
Returns
-------
x : torch.Tensor
Output of layer.
"""
for i in range(self.k):
x = torch.spmm(adj, x)
x = self.linear(x)
return x