Hacker News Reader

Comment by master_yoda_1 | Hacker News Reader

I request the moderator of hn to please flag this post. This fraud is claiming to solve np hard problem in title and some jokers at hn make it to the front page of hn

View on HN

I didn't read the title that way. You can have a (heuristic) solver that finds solutions to some instances of an NP-hard problem in reasonable time. That's not the same as claiming to have an algorithm that, in the strict theoretical CS sense, would solve that problem (i.e. all instances of it) in polynomial time. The latter claim would be highly suspicious by default; the former need not be, as it doesn't imply anything extraordinary.

Quote from OP in another comment:

> I think a better way to phrase it is that we are writing a solver for a reasonable subset of inputs to an NP-hard problem.

Many expert spend life on approximation of np hard problem. For the name of holy God if you don’t understand something then please don’t trivialize it.

Was the headline changed from "problem" to "puzzle"?

Is this even a NP-hard problem?

I would really enjoy a proof that it is not!

Could you be more specific about the fraud? Optimizing your code to use appropriate data structures, and using state reduction to remove symmetries and duplicate states seems like a perfectly cromulent way of reducing a problem to its true core, which may be small enough to bruteforce.

Claiming to solve np hard problem is a fraud. Just put optimizing code in title.

> Claiming to solve np hard problem is a fraud.

Claiming to solve an np hard problem in polynomial time on all inputs would be either a fraud or a breakthrough. This is not such a claim. The algorithm is organized to perform well on many but not all inputs--its worst case is exponential time and it doesn't pretend to be otherwise. If you play chess against a chess engine like Stockfish, the exact same thing is going on, and in fact the algorithms involved are closely related to the one in the article.

I'd suggest that maybe "puzzle" has a slightly different connotation than "problem".

I think a "puzzle" is a concrete instance of a more formal and abstract "problem".

So I don't think claiming to solve an instance (or many!) of a problem is inappropriate.

People solve specific np-hard problems all the time, every day. They just don't solve THE np-hard problem.

Reply on news.ycombinator.com