nerdslab / latentot Goto Github PK

Dear Authors,
Thanks for sharing the code. I have a small confusion in the code. I think the transport function in ot.py should be mapping points through barycentric projection. I am then confused why does the code-logic differ from the one in the Python OT library.

To be more specific, the former directly performs a matrix vector product of coupling matrix & the vector of target samples while the latter first computes marginal from the joint coupling and then does matrix vector product of matrix with the marginal distribution and the vector target samples.

Please help me with this confusion. Thanks for your time.

Piyushi