Talk:MM algorithm

Latest comment: 2 years ago by 2A02:8108:50BF:C694:1CE0:EDD5:D67C:A239 in topic "MM itself is not an algorithm"

Comments

edit

The following comments were moved from the text and need to be addressed

  • Is convexity really a requirement ? The function drawn below is not. With regard to maximization, I'd rather consider concave functions actually.

Billlion (talk) 08:10, 2 July 2014 (UTC)Reply

Headings

edit

I think "How it works" is a bit too informal for a heading. Thoughts?130.225.93.185 (talk) 08:01, 2 July 2014 (UTC) That was me Billlion (talk) 08:02, 2 July 2014 (UTC)Reply

"MM itself is not an algorithm"

edit

Well, I agree that MM is a description of how to construct an optimization algorithm (given a function to be optimized), but that description can be seen as an algorithm itself, too. A "meta-algorithm", if you will. Obviously that "meta-algorithm" could not be implemented on computer hardware since it involves a highly non-deterministic step in symbolic math (finding a suitable function  ). This is a terminological problem. Perhaps a better wording would be that MM is a class of (optimization) algorithms following a common mathematical design. --2A02:8108:50BF:C694:1CE0:EDD5:D67C:A239 (talk) 09:08, 7 January 2022 (UTC)Reply