Simplicial convolutional neural network implementation for complex classification.
- class topomodelx.nn.simplicial.scnn.SCNN(in_channels, hidden_channels, conv_order_down, conv_order_up, aggr_norm=False, update_func=None, n_layers=2)[source]#
Simplicial convolutional neural network implementation for complex classification.
Note: At the last layer, we obtain the output on simplcies, e.g., edges. To perform the complex classification task for this challenge, we consider pass the final output to a linear layer and compute the average.
- Parameters:
- in_channelsint
Dimension of input features.
- hidden_channelsint
Dimension of features of hidden layers.
- out_channelsint
Dimension of output features.
- conv_order_downint
Order of lower convolution.
- conv_order_upint
Order of upper convolution.
- aggrbool
Whether to aggregate features on the nodes into 1 feature for the whole complex. Default: False.
- n_layersint
Number of layers.
- forward(x, laplacian_down, laplacian_up)[source]#
Forward computation.
- Parameters:
- xtorch.Tensor, shape = (n_simplices, channels)
Tensor of features node/edge/face.
- laplacian_downtorch.Tensor, shape = (n_simplices, n_simplices)
Down Laplacian. For node features, laplacian_down = None.
- laplacian_up: torch.Tensor, shape = (n_edges, n_nodes)
Up Laplacian.
- Returns:
- torch.Tensor, shape = (n_simplices, hidden_channels)
Final hidden representation of one-dimensional cells.