search

eddardd / WBTransport

This repo contains the implementation of the Wasserstein Barycenter Transport proposed in "Wasserstein Barycenter Transport for Acoustic Adaptation" at ICASSP21 and "Wasserstein Barycenter for Multi-Source Domain Adaptation" in CVPR21
https://openaccess.thecvf.com/content/CVPR2021/html/Montesuma_Wasserstein_Barycenter_for_Multi-Source_Domain_Adaptation_CVPR_2021_paper.html
GNU General Public License v3.0
31 stars 5 forks source link

question about eq10 #3

Closed Fay-why closed 1 year ago

Fay-why commented 1 year ago

Hi, dear author, thanks for your great job.

I have a big question about the following equation, since this step aims to calculate the source domain aggregation barycenter, why the coupling γ space is (μ_s, μ_t)? I think it should be (μ_s, μ_b) ?

image

image

eddardd commented 1 year ago

Hi Helen,

Thank you for your question. Yes, this is actually a typo. It is a coupling between μ_s and μ_b.

Fay-why commented 1 year ago

ok~ Thanks for your reply!

And I also want to know if this kind of domain adaptation [source domain, target domain] should have same label space? Because I have seen the following equation in your paper.

image

eddardd commented 1 year ago

This is the covariate shift assumption. This is a necessary condition for doing unsupervised domain adaptation (i.e. without target labels). Under this assumption the distributional shift problem can be solved by applying a transformation to the features, while keeping labels unchanged.

Fay-why commented 1 year ago

Thanks! I will close this issue.