Source code for torch_geometric.nn.conv.supergat_conv

from typing import Optional
from torch_geometric.typing import OptTensor

import math

import torch
from torch import Tensor
from torch.nn import Parameter
import torch.nn.functional as F

from torch_geometric.nn.conv import MessagePassing
from torch_geometric.nn.dense.linear import Linear
from torch_geometric.utils import (remove_self_loops, add_self_loops, softmax,
                                   is_undirected, negative_sampling,
                                   batched_negative_sampling, to_undirected,
                                   dropout_adj)

from ..inits import glorot, zeros


[docs]class SuperGATConv(MessagePassing): r"""The self-supervised graph attentional operator from the `"How to Find Your Friendly Neighborhood: Graph Attention Design with Self-Supervision" <https://openreview.net/forum?id=Wi5KUNlqWty>`_ paper .. math:: \mathbf{x}^{\prime}_i = \alpha_{i,i}\mathbf{\Theta}\mathbf{x}_{i} + \sum_{j \in \mathcal{N}(i)} \alpha_{i,j}\mathbf{\Theta}\mathbf{x}_{j}, where the two types of attention :math:`\alpha_{i,j}^{\mathrm{MX\ or\ SD}}` are computed as: .. math:: \alpha_{i,j}^{\mathrm{MX\ or\ SD}} &= \frac{ \exp\left(\mathrm{LeakyReLU}\left( e_{i,j}^{\mathrm{MX\ or\ SD}} \right)\right)} {\sum_{k \in \mathcal{N}(i) \cup \{ i \}} \exp\left(\mathrm{LeakyReLU}\left( e_{i,k}^{\mathrm{MX\ or\ SD}} \right)\right)} e_{i,j}^{\mathrm{MX}} &= \mathbf{a}^{\top} [\mathbf{\Theta}\mathbf{x}_i \, \Vert \, \mathbf{\Theta}\mathbf{x}_j] \cdot \sigma \left( \left( \mathbf{\Theta}\mathbf{x}_i \right)^{\top} \mathbf{\Theta}\mathbf{x}_j \right) e_{i,j}^{\mathrm{SD}} &= \frac{ \left( \mathbf{\Theta}\mathbf{x}_i \right)^{\top} \mathbf{\Theta}\mathbf{x}_j }{ \sqrt{d} } The self-supervised task is a link prediction using the attention values as input to predict the likelihood :math:`\phi_{i,j}^{\mathrm{MX\ or\ SD}}` that an edge exists between nodes: .. math:: \phi_{i,j}^{\mathrm{MX}} &= \sigma \left( \left( \mathbf{\Theta}\mathbf{x}_i \right)^{\top} \mathbf{\Theta}\mathbf{x}_j \right) \phi_{i,j}^{\mathrm{SD}} &= \sigma \left( \frac{ \left( \mathbf{\Theta}\mathbf{x}_i \right)^{\top} \mathbf{\Theta}\mathbf{x}_j }{ \sqrt{d} } \right) .. note:: For an example of using SuperGAT, see `examples/super_gat.py <https://github.com/pyg-team/pytorch_geometric/blob/master/examples/ super_gat.py>`_. Args: in_channels (int): Size of each input sample, or :obj:`-1` to derive the size from the first input(s) to the forward method. out_channels (int): Size of each output sample. heads (int, optional): Number of multi-head-attentions. (default: :obj:`1`) concat (bool, optional): If set to :obj:`False`, the multi-head attentions are averaged instead of concatenated. (default: :obj:`True`) negative_slope (float, optional): LeakyReLU angle of the negative slope. (default: :obj:`0.2`) dropout (float, optional): Dropout probability of the normalized attention coefficients which exposes each node to a stochastically sampled neighborhood during training. (default: :obj:`0`) add_self_loops (bool, optional): If set to :obj:`False`, will not add self-loops to the input graph. (default: :obj:`True`) bias (bool, optional): If set to :obj:`False`, the layer will not learn an additive bias. (default: :obj:`True`) attention_type (string, optional): Type of attention to use. (:obj:`'MX'`, :obj:`'SD'`). (default: :obj:`'MX'`) neg_sample_ratio (float, optional): The ratio of the number of sampled negative edges to the number of positive edges. (default: :obj:`0.5`) edge_sample_ratio (float, optional): The ratio of samples to use for training among the number of training edges. (default: :obj:`1.0`) is_undirected (bool, optional): Whether the input graph is undirected. If not given, will be automatically computed with the input graph when negative sampling is performed. (default: :obj:`False`) **kwargs (optional): Additional arguments of :class:`torch_geometric.nn.conv.MessagePassing`. """ att_x: OptTensor att_y: OptTensor def __init__(self, in_channels: int, out_channels: int, heads: int = 1, concat: bool = True, negative_slope: float = 0.2, dropout: float = 0.0, add_self_loops: bool = True, bias: bool = True, attention_type: str = 'MX', neg_sample_ratio: float = 0.5, edge_sample_ratio: float = 1.0, is_undirected: bool = False, **kwargs): kwargs.setdefault('aggr', 'add') super(SuperGATConv, self).__init__(node_dim=0, **kwargs) self.in_channels = in_channels self.out_channels = out_channels self.heads = heads self.concat = concat self.negative_slope = negative_slope self.dropout = dropout self.add_self_loops = add_self_loops self.attention_type = attention_type self.neg_sample_ratio = neg_sample_ratio self.edge_sample_ratio = edge_sample_ratio self.is_undirected = is_undirected assert attention_type in ['MX', 'SD'] assert 0.0 < neg_sample_ratio and 0.0 < edge_sample_ratio <= 1.0 self.lin = Linear(in_channels, heads * out_channels, bias=False, weight_initializer='glorot') if self.attention_type == 'MX': self.att_l = Parameter(torch.Tensor(1, heads, out_channels)) self.att_r = Parameter(torch.Tensor(1, heads, out_channels)) else: # self.attention_type == 'SD' self.register_parameter('att_l', None) self.register_parameter('att_r', None) self.att_x = self.att_y = None # x/y for self-supervision if bias and concat: self.bias = Parameter(torch.Tensor(heads * out_channels)) elif bias and not concat: self.bias = Parameter(torch.Tensor(out_channels)) else: self.register_parameter('bias', None) self.reset_parameters()
[docs] def reset_parameters(self): self.lin.reset_parameters() glorot(self.att_l) glorot(self.att_r) zeros(self.bias)
[docs] def forward(self, x: Tensor, edge_index: Tensor, neg_edge_index: OptTensor = None, batch: OptTensor = None) -> Tensor: r""" Args: neg_edge_index (Tensor, optional): The negative edges to train against. If not given, uses negative sampling to calculate negative edges. (default: :obj:`None`) """ N, H, C = x.size(0), self.heads, self.out_channels if self.add_self_loops: edge_index, _ = remove_self_loops(edge_index) edge_index, _ = add_self_loops(edge_index, num_nodes=N) x = self.lin(x).view(-1, H, C) # propagate_type: (x: Tensor) out = self.propagate(edge_index, x=x, size=None) if self.training: pos_edge_index = self.positive_sampling(edge_index) pos_att = self.get_attention( edge_index_i=pos_edge_index[1], x_i=x[pos_edge_index[1]], x_j=x[pos_edge_index[0]], num_nodes=x.size(0), return_logits=True, ) if neg_edge_index is None: neg_edge_index = self.negative_sampling(edge_index, N, batch) neg_att = self.get_attention( edge_index_i=neg_edge_index[1], x_i=x[neg_edge_index[1]], x_j=x[neg_edge_index[0]], num_nodes=x.size(0), return_logits=True, ) self.att_x = torch.cat([pos_att, neg_att], dim=0) self.att_y = self.att_x.new_zeros(self.att_x.size(0)) self.att_y[:pos_edge_index.size(1)] = 1. if self.concat is True: out = out.view(-1, self.heads * self.out_channels) else: out = out.mean(dim=1) if self.bias is not None: out += self.bias return out
def message(self, edge_index_i: Tensor, x_i: Tensor, x_j: Tensor, size_i: Optional[int]) -> Tensor: alpha = self.get_attention(edge_index_i, x_i, x_j, num_nodes=size_i) alpha = F.dropout(alpha, p=self.dropout, training=self.training) return x_j * alpha.view(-1, self.heads, 1)
[docs] def negative_sampling(self, edge_index: Tensor, num_nodes: int, batch: OptTensor = None) -> Tensor: num_neg_samples = int(self.neg_sample_ratio * self.edge_sample_ratio * edge_index.size(1)) if not self.is_undirected and not is_undirected( edge_index, num_nodes=num_nodes): edge_index = to_undirected(edge_index, num_nodes=num_nodes) if batch is None: neg_edge_index = negative_sampling(edge_index, num_nodes, num_neg_samples=num_neg_samples) else: neg_edge_index = batched_negative_sampling( edge_index, batch, num_neg_samples=num_neg_samples) return neg_edge_index
[docs] def positive_sampling(self, edge_index: Tensor) -> Tensor: pos_edge_index, _ = dropout_adj(edge_index, p=1. - self.edge_sample_ratio, training=self.training) return pos_edge_index
[docs] def get_attention(self, edge_index_i: Tensor, x_i: Tensor, x_j: Tensor, num_nodes: Optional[int], return_logits: bool = False) -> Tensor: if self.attention_type == 'MX': logits = (x_i * x_j).sum(dim=-1) if return_logits: return logits alpha = (x_j * self.att_l).sum(-1) + (x_i * self.att_r).sum(-1) alpha = alpha * logits.sigmoid() else: # self.attention_type == 'SD' alpha = (x_i * x_j).sum(dim=-1) / math.sqrt(self.out_channels) if return_logits: return alpha alpha = F.leaky_relu(alpha, self.negative_slope) alpha = softmax(alpha, edge_index_i, num_nodes=num_nodes) return alpha
[docs] def get_attention_loss(self) -> Tensor: r"""Compute the self-supervised graph attention loss.""" if not self.training: return torch.tensor([0], device=self.lin.weight.device) return F.binary_cross_entropy_with_logits( self.att_x.mean(dim=-1), self.att_y, )
def __repr__(self): return '{}({}, {}, heads={}, type={})'.format(self.__class__.__name__, self.in_channels, self.out_channels, self.heads, self.attention_type)