Skip to content
This repository was archived by the owner on Dec 6, 2023. It is now read-only.

Add proximal opeartor for the 2D total variation penalty#101

Open
fabianp wants to merge 8 commits into
scikit-learn-contrib:masterfrom
fabianp:TV2DPenalty
Open

Add proximal opeartor for the 2D total variation penalty#101
fabianp wants to merge 8 commits into
scikit-learn-contrib:masterfrom
fabianp:TV2DPenalty

Conversation

@fabianp

@fabianp fabianp commented Nov 5, 2016

Copy link
Copy Markdown
Member

Also includes an example of using this proximal operator for denoising. The output of this example is

image

Other changes:

  • Added tolerance parameter for FISTA

@mblondel

mblondel commented Nov 6, 2016

Copy link
Copy Markdown
Member

Nice!

I have recently worked on a more efficient implementation of FISTA:
https://gist.github.com/mblondel/5105786d740693a6996bcb8e482c7083

The main advantage is that it is possible to compute the projection and regularization term in the same function call, which allows to share some expensive computations in some cases (e.g. nuclear norm). I also support a tolerance constant.

@mblondel

mblondel commented Nov 7, 2016

Copy link
Copy Markdown
Member

@fabianp I am not very familiar with the TV proximity operator. Perhaps you could ask someone else who is to give a quick review?

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants