skada.deep.losses.gda_loss
- skada.deep.losses.gda_loss(s, t, metric='euc', laplac='laplac1')[source]
Compute the GDA loss between two graphs.
Inspired by https://github.com/CrownX/SPA
- Parameters:
- storch.Tensor
Source features.
- ttorch.Tensor
Target features.
- metricstr, default="euc"
The metric to use for the adjacency matrix.
- laplacstr, default="laplac1"
The Laplacian matrix to use.