Feasibility-based Fixed Point Networks
Updated: Jul 15
TD;DR - Projection algorithms can now be fused with data-driven regularization to effectively solve inverse problems with many constraints.
This blog is based upon a recent paper (preprint available here).
(On mobile devices the embedded document below may not display appropriately, in which case you may view it as a .pdf file here. On desktops, if you opened the .pdf viewer in full screen, then you may need to refresh the page to reload the display.)
As always, I am happy to discuss in the comments below any comments, suggestions, criticisms, and/or questions pertaining to this post.