base_model.py

import math
import torch
from torch.nn import Parameter
import torch.nn.functional as F
from torch_geometric.nn.conv import MessagePassing, GCNConv
from torch_geometric.utils import remove_self_loops, add_self_loops, add_remaining_self_loops, softmax
from inits import glorot, zeros, reset
from torch_scatter import scatter_add
from torch.nn import Linear

class GCN(MessagePassing):
    def __init__(self, in_channels, out_channels, weight, bias,
    improved=False, cached=False, normalize=True, **kwargs):
        super(GCN, self).__init__(aggr='add', **kwargs)

        self.in_channels = in_channels
        self.out_channels = out_channels
        self.improved = improved
        self.cached = cached
        self.normalize = normalize

        self.weight = weight
        self.bias = bias
        self.reset_parameters()

    def reset_parameters(self):
        #glorot(self.weight)
        #zeros(self.bias)
        self.cached_result = None
        self.cached_num_edges = None


    @staticmethod
    def norm(edge_index, num_nodes, edge_weight=None, improved=False,
             dtype=None):        
        if edge_weight is None:
            edge_weight = torch.ones((edge_index.size(1), ), dtype=dtype,
                                     device=edge_index.device)

        fill_value = 1 if not improved else 2
        edge_index, edge_weight = add_remaining_self_loops(
            edge_index, edge_weight, fill_value, num_nodes)

        row, col = edge_index
        deg = scatter_add(edge_weight, row, dim=0, dim_size=num_nodes)
        deg_inv_sqrt = deg.pow(-0.5)
        deg_inv_sqrt[deg_inv_sqrt == float('inf')] = 0

        return edge_index, deg_inv_sqrt[row] * edge_weight * deg_inv_sqrt[col]

    def forward(self, x, edge_index, edge_weight=None):
        """"""
        x = torch.matmul(x, self.weight)

        if self.cached and self.cached_result is not None:
            if edge_index.size(1) != self.cached_num_edges:
                raise RuntimeError(
                    'Cached {} number of edges, but found {}. Please '
                    'disable the caching behavior of this layer by removing '
                    'the `cached=True` argument in its constructor.'.format(
                        self.cached_num_edges, edge_index.size(1)))

        if not self.cached or self.cached_result is None:
            self.cached_num_edges = edge_index.size(1)
            if self.normalize:
                edge_index, norm = self.norm(edge_index, x.size(
                    self.node_dim), edge_weight, self.improved, x.dtype)
            else:
                norm = edge_weight
            self.cached_result = edge_index, norm

        edge_index, norm = self.cached_result

        return self.propagate(edge_index, x=x, norm=norm)


    def message(self, x_j, norm):
        return norm.view(-1, 1) * x_j if norm is not None else x_j

    def update(self, aggr_out):
        if self.bias is not None:
            aggr_out = aggr_out + self.bias
        return aggr_out

    def __repr__(self):
        return '{}({}, {})'.format(self.__class__.__name__, self.in_channels,
                                   self.out_channels)

class GAT(MessagePassing):

    def __init__(self, in_channels, out_channels, weight, att, bias, heads=1, dropout=0, concat=True,
                 negative_slope=0.2, **kwargs):
        super(GAT, self).__init__(aggr='add', **kwargs)
        self.in_channels = in_channels
        self.out_channels = int(out_channels / heads)
        self.heads = heads
        self.concat = True
        self.negative_slope = negative_slope
        self.dropout = dropout

        self.weight = weight
        self.att = att
        self.bias = bias

    def forward(self, x, edge_index, size=None):
        if size is None and torch.is_tensor(x):
            edge_index, _ = remove_self_loops(edge_index)
            edge_index, _ = add_self_loops(edge_index,
                                           num_nodes=x.size(self.node_dim))

        if torch.is_tensor(x):
            x = torch.matmul(x, self.weight)
        else:
            x = (None if x[0] is None else torch.matmul(x[0], self.weight),
                 None if x[1] is None else torch.matmul(x[1], self.weight))

        return self.propagate(edge_index, size=size, x=x)


    def message(self, edge_index_i, x_i, x_j, size_i):
        # Compute attention coefficients.
        x_j = x_j.view(-1, self.heads, self.out_channels)
        if x_i is None:
            alpha = (x_j * self.att[:, :, self.out_channels:]).sum(dim=-1)
        else:
            x_i = x_i.view(-1, self.heads, self.out_channels)
            alpha = (torch.cat([x_i, x_j], dim=-1) * self.att).sum(dim=-1)

        alpha = F.leaky_relu(alpha, self.negative_slope)
        alpha = softmax(alpha, edge_index_i, size_i)

        # Sample attention coefficients stochastically.
        alpha = F.dropout(alpha, p=self.dropout, training=self.training)

        return x_j * alpha.view(-1, self.heads, 1)

    def update(self, aggr_out):
        if self.concat is True:
            aggr_out = aggr_out.view(-1, self.heads * self.out_channels)
        else:
            aggr_out = aggr_out.mean(dim=1)

        if self.bias is not None:
            aggr_out = aggr_out + self.bias
        return aggr_out

    def __repr__(self):
        return '{}({}, {}, heads={})'.format(self.__class__.__name__,
                                             self.in_channels,
                                             self.out_channels, self.heads)

class GIN(MessagePassing):
    def __init__(self, weight, bias, weight2, bias2, **kwargs):
        super(GIN, self).__init__(aggr='add', **kwargs)
        self.weight = weight
        self.bias = bias
        self.weight2 = weight2
        self.bias2 = bias2

    def reset_parameters(self):
        self.eps.data.fill_(self.initial_eps)

    def forward(self, x, edge_index):
        """"""
        x = x.unsqueeze(-1) if x.dim() == 1 else x
        edge_index, _ = remove_self_loops(edge_index)
        out = F.linear(x + self.propagate(edge_index, x=x), self.weight, self.bias)
        out = F.relu(out)
        out = F.linear(out, self.weight2, self.bias2)
        return out
        
    def message(self, x_j):
        return x_j

class SGC(MessagePassing):
    def __init__(self, weight, bias, K=2, cached=False, **kwargs):
        super(SGC, self).__init__(aggr='add', **kwargs)

        self.K = K
        self.cached = cached
        self.weight = weight
        self.bias = bias
        self.reset_parameters()

    def reset_parameters(self):
        self.cached_result = None
        self.cached_num_edges = None

    def forward(self, x, edge_index, edge_weight=None):
        """"""
        if self.cached and self.cached_result is not None:
            if edge_index.size(1) != self.cached_num_edges:
                raise RuntimeError(
                    'Cached {} number of edges, but found {}. Please '
                    'disable the caching behavior of this layer by removing '
                    'the `cached=True` argument in its constructor.'.format(
                        self.cached_num_edges, edge_index.size(1)))

        if not self.cached:
            x = F.linear(x, self.weight, self.bias)

        if not self.cached or self.cached_result is None:
            self.cached_num_edges = edge_index.size(1)
            edge_index, norm = GCNConv.norm(edge_index, x.size(self.node_dim),
                                            edge_weight, dtype=x.dtype)
            for k in range(self.K):
                x = self.propagate(edge_index, x=x, norm=norm)
            self.cached_result = x
        return x

    def message(self, x_j, norm):
        return norm.view(-1, 1) * x_j