CWN#
CWN class.
- class topomodelx.nn.cell.cwn.CWN(in_channels_0, in_channels_1, in_channels_2, hid_channels, n_layers, **kwargs)[source]#
Implementation of a specific version of CW network [1].
- Parameters:
- in_channels_0int
Dimension of input features on nodes (0-cells).
- in_channels_1int
Dimension of input features on edges (1-cells).
- in_channels_2int
Dimension of input features on faces (2-cells).
- hid_channelsint
Dimension of hidden features.
- n_layersint
Number of CWN layers.
- **kwargsoptional
Additional arguments CWNLayer.
References
[1]Bodnar, et al. Weisfeiler and Lehman go cellular: CW networks. NeurIPS 2021. https://arxiv.org/abs/2106.12575
- forward(x_0, x_1, x_2, adjacency_0, incidence_2, incidence_1_t)[source]#
Forward computation through projection, convolutions, linear layers and average pooling.
- Parameters:
- x_0torch.Tensor, shape = (n_nodes, in_channels_0)
Input features on the nodes (0-cells).
- x_1torch.Tensor, shape = (n_edges, in_channels_1)
Input features on the edges (1-cells).
- x_2torch.Tensor, shape = (n_faces, in_channels_2)
Input features on the faces (2-cells).
- adjacency_0torch.Tensor, shape = (n_edges, n_edges)
Upper-adjacency matrix of rank 1.
- incidence_2torch.Tensor, shape = (n_edges, n_faces)
Boundary matrix of rank 2.
- incidence_1_ttorch.Tensor, shape = (n_edges, n_nodes)
Coboundary matrix of rank 1.
- Returns:
- x_0torch.Tensor, shape = (n_nodes, in_channels_0)
Final hidden states of the nodes (0-cells).
- x_1torch.Tensor, shape = (n_edges, in_channels_1)
Final hidden states the edges (1-cells).
- x_2torch.Tensor, shape = (n_edges, in_channels_2)
Final hidden states of the faces (2-cells).