models.py

# -*- coding:utf-8 -*-

import torch.nn.functional as F
from torch import nn
from torch_geometric.nn import GATConv, GCNConv, SAGEConv


class GCN_LP(nn.Module):
    def __init__(self, in_channels, hidden_channels, out_channels):
        super(GCN_LP, self).__init__()
        self.conv1 = GCNConv(in_channels, hidden_channels)
        self.conv2 = GCNConv(hidden_channels, out_channels)

    def encode(self, data):
        x, edge_index = data.x, data.edge_index
        x = F.relu(self.conv1(x, edge_index))
        x = F.dropout(x, p=0.5, training=self.training)
        x = self.conv2(x, edge_index)

        return x

    def decode(self, z, edge_label_index):
        src = z[edge_label_index[0]]
        dst = z[edge_label_index[1]]
        r = (src * dst).sum(dim=-1)
        return r

    def forward(self, data, edge_label_index):
        z = self.encode(data)
        return self.decode(z, edge_label_index)


class SAGE_LP(nn.Module):
    def __init__(self, in_channels, hidden_channels, out_channels):
        super(SAGE_LP, self).__init__()
        self.conv1 = SAGEConv(in_channels, hidden_channels)
        self.conv2 = SAGEConv(hidden_channels, out_channels)

    def encode(self, data):
        x, edge_index = data.x, data.edge_index
        x = F.relu(self.conv1(x, edge_index))
        x = F.dropout(x, p=0.5, training=self.training)
        x = self.conv2(x, edge_index)

        return x

    def decode(self, z, edge_label_index):
        src = z[edge_label_index[0]]
        dst = z[edge_label_index[1]]
        r = (src * dst).sum(dim=-1)
        return r

    def forward(self, data, edge_label_index):
        z = self.encode(data)
        return self.decode(z, edge_label_index)


class GAT_LP(nn.Module):
    def __init__(self, in_channels, hidden_channels, out_channels):
        super(GAT_LP, self).__init__()
        self.conv1 = GATConv(in_channels, hidden_channels, heads=8, concat=False)
        self.conv2 = GATConv(hidden_channels, out_channels, heads=8, concat=False)

    def encode(self, data):
        x, edge_index = data.x, data.edge_index
        x = F.relu(self.conv1(x, edge_index))
        x = F.dropout(x, p=0.5, training=self.training)
        x = self.conv2(x, edge_index)

        return x

    def decode(self, z, edge_label_index):
        src = z[edge_label_index[0]]
        dst = z[edge_label_index[1]]
        r = (src * dst).sum(dim=-1)
        return r

    def forward(self, data, edge_label_index):
        z = self.encode(data)
        return self.decode(z, edge_label_index)