POT
POT copied to clipboard
Linear OT mapping across different spaces
🚀 Feature
Would it be possible to perform linear domain adaptation (like is done by ot.da.LinearTransport) but across different spaces ?
Motivation
The LinearTransport is very useful to align distributions without losing information from the original geometry. It would be great to have the same feature but for distributions in different spaces / in the Gromov-Wasserstein context