-
Notifications
You must be signed in to change notification settings - Fork 7
[ENH] Stabilized and epsilon scaled sinkhorn solvers #84
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
…just logging format
…ust evaluation frequency
…checks and update logging messages
…ction and adjust train_params unpacking
… plan computation
bthirion
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just curious. This solver is not in POT ?
There's a cost in maintaining our own solvers...
…implify function calls
…enhance readability
…improved precision
Dense solvers are similar to POT, but the sparse implementations are unique to FUGW. |
bthirion
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I can't spot any issue.
At the end of the day, how many solvers do we support ? Do we have a clear use case for all of them ?
…r improved accuracy
bthirion
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe the code could be organized a bit better ?
|
I removed the dense solvers which are already implemented in POT, which means less code to maintain. |
I removed the nested functions. |
This PR implements stabilized and epsilon scaled sinkhorn solvers (Schmitzer, B. 2016), supporting sparse cost matrices.