pygmtools.utils.build_aff_mat_from_pyg

pygmtools.utils.build_aff_mat_from_pyg(G1, G2, node_aff_fn=None, edge_aff_fn=None, backend=None)[source]

Convert torch_geometric.data.Data object to affinity matrix

Parameters

  • G1 – Graph object, whose type must be torch_geometric.data.Data

  • G2 – Graph object, whose type must be torch_geometric.data.Data

  • node_aff_fn – (default: inner_prod_aff_fn) the node affinity function with the characteristic node_aff_fn(2D Tensor, 2D Tensor) -> 2D Tensor, which accepts two node feature tensors and outputs the node-wise affinity tensor. See inner_prod_aff_fn() as an example.

  • edge_aff_fn – (default: inner_prod_aff_fn) the edge affinity function with the characteristic edge_aff_fn(2D Tensor, 2D Tensor) -> 2D Tensor, which accepts two edge feature tensors and outputs the edge-wise affinity tensor. See inner_prod_aff_fn() as an example.

  • backend – (default: pygmtools.BACKEND variable) the backend for computation.

Returns

the affinity matrix corresponding to the torch_geometric.data.Data object G1 and G2

Example
>>> import torch
>>> from torch_geometric.data import Data
>>> import pygmtools as pygm
>>> pygm.set_backend('pytorch')

# Generate Graph object
>>> x1 = torch.rand((4, 2), dtype=torch.float)
>>> e1 = torch.tensor([[0, 0, 1, 1, 2, 2, 3], [1, 2, 0, 2, 0, 3, 1]], dtype=torch.long)
>>> G1 = Data(x=x1, edge_index=e1)
>>> x2 = torch.rand((5, 2), dtype=torch.float)
>>> e2 = torch.tensor([[0, 0, 1, 1, 2, 2, 3, 4, 4], [1, 3, 2, 3, 1, 3, 4, 2, 3]], dtype=torch.long)
>>> G2 = Data(x=x2, edge_index=e2)

# Obtain Affinity Matrix
>>> K = pygm.utils.build_aff_mat_from_pyg(G1, G2)
>>> K.shape
(20, 20)

# The affinity matrices K can be further processed by GM solvers