Comments on: An ML/AI approach to P != NP https://brenocon.com/blog/2010/08/an-mlai-approach-to-p-np/ cognition, language, social systems; statistics, visualization, computation Tue, 25 Nov 2025 13:11:20 +0000 hourly 1 By: chris https://brenocon.com/blog/2010/08/an-mlai-approach-to-p-np/#comment-34049 chris Sat, 28 Aug 2010 14:54:37 +0000 http://anyall.org/blog/?p=851#comment-34049 I found the attempt close in spirit to the Mulmuley programme of attacking P/NP via Algebraic Geometry (approach summarized last year by Lance Fortnow in CACM).

Graphical models find natural generalization in certain toric algebraic varieties (eg. Strumfels et al. work on “algebraic statistics”), and sheaves are natural setting for model theory.

Frankly I can’t see how ML per se fits in here, but the general direction seems more plausible now when two independent perspectives in phrased different languages end up resembling each other.

]]> By: jimmy https://brenocon.com/blog/2010/08/an-mlai-approach-to-p-np/#comment-32396 jimmy Tue, 10 Aug 2010 19:11:54 +0000 http://anyall.org/blog/?p=851#comment-32396 http://rjlipton.wordpress.com/2010/08/09/issues-in-the-proof-that-p%E2%89%A0np/
http://michaelnielsen.org/polymath1/index.php?title=Deolalikar%27s_P!%3DNP_paper

]]> By: brendano https://brenocon.com/blog/2010/08/an-mlai-approach-to-p-np/#comment-32315 brendano Mon, 09 Aug 2010 23:18:04 +0000 http://anyall.org/blog/?p=851#comment-32315 Right, statistical physics is the root of everything. I personally learned the material starting from machine learning and artificial intelligence courses. The borrowing happened 10-20 years ago, as far as I can tell. And that depends on what you count as a significant borrowing… I guess the Hinton and Sejnowski early 1980′s stuff on Boltzmann machines was already adopting this sort of thing, and that then became part of the neural networks literature, which was cutting edge machine learning at the time.

]]> By: Yisong https://brenocon.com/blog/2010/08/an-mlai-approach-to-p-np/#comment-32308 Yisong Mon, 09 Aug 2010 21:40:59 +0000 http://anyall.org/blog/?p=851#comment-32308 I would actually phrase it as a statistical physics-centric approach — at least the part that uses probabilistic graphical models. ML/AI borrowed much of that (very useful) machinery from the physics community.

]]> By: brendano https://brenocon.com/blog/2010/08/an-mlai-approach-to-p-np/#comment-32305 brendano Mon, 09 Aug 2010 20:33:34 +0000 http://anyall.org/blog/?p=851#comment-32305 Thanks, updated post. Fast-paced news….

]]> By: jimmy https://brenocon.com/blog/2010/08/an-mlai-approach-to-p-np/#comment-32298 jimmy Mon, 09 Aug 2010 18:38:20 +0000 http://anyall.org/blog/?p=851#comment-32298 updated version.
http://www.hpl.hp.com/personal/Vinay_Deolalikar/Papers/pnp_updated.pdf

]]>