|
This document uses a surprising amount of statistics, and may actually be comprehensible to some of us. I won't pretend to understand any of it. But it looks interesting. http://www.win.tue.nl/~gwoegi/P-versus-NP/Deolalikar.pdf link from personal website of the researcher: http://www.hpl.hp.com/personal/Vinay_Deolalikar/Papers/pnp_preliminary.pdf
showing 5 of 8
show all
|
|
If you look at Vinay's industrial research, you'll recognize that they're mostly related with Data Mining and ML. Also most of the problems in AI, ML and DM are NP-complete, so the proof of this theorem might effect our field too. Actually, it would be really interesting to see that P=NP. And you can follow the developments about the proof from the Dick Lipton's blog: https://rjlipton.wordpress.com/2010/08/15/the-p%e2%89%a0np-proof-is-one-week-old/ |
|
http://www.ugcs.caltech.edu/~stansife/pnp.html for anyone who wants the short ride... |
btw, Scott Aaronson is betting $200,000 the proof doesn't hold up
it would be interesting to watch, from what i could gather it relies on clustering and graphical models as part of the proof.
Announcing that you proved something without peer reviews doesn't feel right at all. Fleischmann and Pons cold fusion a good example. Of course now paper will be reviewed, get much more attention and in the end mistakes would probably be found.
@ilya: you're being unfair. The paper wasn't announced globally by the author; it was sent to a small number of peers for review, who then promptly leaked the paper to the larger public. As of right now some people are reading it and others are preparing to read it, and many of these people would qualify as peers.
For results of this caliber it is better to conduct the proofing in the open than to trust absolutely the two or three poor reviewers from a journal who would have to check far more than they're used to (specially since this paper borrows results from lots of different areas) with far greater consequences than accepting/rejecting a wrong/right paper of less significance. Proofing Perelman's proof of Poincare's conjecture happenned mostly around arXiv at first, IIRC.
You are probably right, I might be unfair. I feel that paper first should have been sent to journal, reviewed, published, and then it would be hammered by mathematicians around the world.
If the journal revieweing mechanism were perfect, maybe. The way it is, no single three reviewers picked by a journal editor could probably be able to say that this proof really solves the problem, that it has some flaws that can be fixed or that it is a dead end because of some obscure theorem in complexity. It's better to release it to the larger community so that experts in all sub-areas it uses can verify its arguments and experts on complexity can evaluate the overall structure.
@ilya If you are a serious researcher, whose email top researchers would read, you don't send your paper to JACM and wait for them to forward it to reviewers.
No matter how odd it might sound, it is a common practice. E.g. Consider the paper that gave structure of DNA, it wasn't peer reviewed the journal editor decided that it was too important and published it directly. I think the journal was Science or Nature.
More recently consider the Primes in P paper, the authors also didnt wait for a journal to give a review they sent it to researchers in the field directly.
Thus in this case what he did was correct and appropriate.
Aug 11 update from Terry Tao (issues with proof are probably too major to be fixed) http://www.ugcs.caltech.edu/~stansife/pnp.html