lebeautemps asked for a simple explanation of the whole P=NP thing that's been circulating the internet recently. I figured I'd give it a go - I'd appreciate it if the more knowledgable on my friends list could correct anything particularly stupid I'd said...
A "P" problem is one that is easily solved through a series of steps, without trying every single combination - such as "sort a bunch of names into alphabetical order". You don't have to try every single possible list of names in order to sort a list of names, you can just use a series of comparisons to shuffle them up and down until they're sorted.
An non-P problem is one that has no known efficient set of steps for producing an answer, and thus requires you to try possibilities until you find one that's right. A common example of one that is thought to be hard is The Travelling Salesman problem - "given a bunch of cities and a bunch of roads connecting them, what is the shortest route that will take the salesman to each city exactly once?" There's no known solution to this problem other than "start trying solutions and keep going until you've found one." (although there are ways of excluding obviously wrong answers quickly).
NP is the superset of all problems that are easy to check, P is the smaller subset of all problems that are easy to solve, and proving that they are the same thing would mean that all problems that are easy to check are also easy to solve.
One of the reasons this is important is that pretty much all of the security methods we use online rely on things like "integer factorization" not being P - the fact that we can multiply two large numbers together to get an answer very quickly, but breaking that large number back into the two component parts requires every possibility to be checked (which takes years/decades for very large numbers).
The recent fuss is because of a paper that was published claiming that P!=NP - i.e. that there are definitely problems that are easy to check, but not easy to solve. This would mean, for example, that there is no simple way to convert the big number back into its two components, and thus that all of our security is safe (from that particular direction, anyway).
no subject
Date: 2010-08-13 08:42 am (UTC)no subject
Date: 2010-08-13 08:46 am (UTC)no subject
Date: 2010-08-13 09:19 am (UTC)no subject
Date: 2010-08-13 09:18 am (UTC)no subject
Date: 2010-08-13 09:24 am (UTC)No, an NP problem is any problem where you can efficiently check the answer. All problems in P are in NP, for example. What's known about a problem doesn't affect what complexity class it's in.
no subject
Date: 2010-08-13 09:27 am (UTC)no subject
Date: 2010-08-13 09:33 am (UTC)Before I get to that - it's also not true to say that problems in NP which are not known to be in P are those where you have to try every combination to find a solution. There might be much more efficient search strategies than that, which cut out huge chunks of the search space, but don't manage to actually be polynomial time. In particular, they don't have to be exponential time - lots of algorithms are super-polynomial but sub-exponential.
NP-complete means roughly "as hard as any problem in NP". A polynomial-time algorithm for any NP-complete problem could be used to solve *any* problem in NP in polynomial time, using a device called a "reduction".
no subject
Date: 2010-08-13 09:36 am (UTC)However, there's no reason that all problems in NP but not in P are NP-complete. There could perfectly well be less general NP problems that are still too hard to be in P.
no subject
Date: 2010-08-13 11:57 am (UTC)Indeed, I seem to recall integer factorisation is in that class? I don't actually really understand this stuff, unfortunately, but iirc I remember Aaronson's blog saying that a quantum computer would help with integer factorisation, but would not help with NP-complete problems in general. But I'm not sure I understand correctly.
If I do, a minor nitpick might be to correct the last paragraph of the post, which says that P!=NP would mean that there is no efficient integer factorisation, to say that there probably isn't, or that there is no better solution to the travelling salesman problem[1].
[1] We soon need to replace "travelling salesman" with an equivalent network-routing problem to stay current with what most people are familiar with :)
no subject
Date: 2010-08-13 09:31 am (UTC)Doing that is a set of steps for producing an answer! You probably meant to say "no efficient set of steps" here, or some such.
Also, technically, NP is a superset of P – there's no implication that "an NP problem" doesn't have a polynomial-time solution. Colloquial usage is to only describe problems as NP if they aren't (known to be) in P, on the pragmatic basis that if they were easier than that you'd have said so, but that's not what it really means.
I think the major thing I would add to this explanation is a paragraph explaining why the question is open, along the lines of:
no subject
Date: 2010-08-13 09:50 am (UTC)no subject
Date: 2010-08-13 10:01 am (UTC)Perhaps move the bit about our limited knowledge down to the TSP sentence? So that a non-P problem is simply described as "one which has no efficient set of steps", and then "A common example of a problem which is thought to be hard in this way is The Travelling Salesman problem...".
no subject
Date: 2010-08-13 10:05 am (UTC)no subject
Date: 2010-08-13 09:38 am (UTC)And in cryptography, it's not worst-case hardness that matters - it's average case hardness.
If P turns out to equal NP, cryptography will have big problems - we'll be on the search for problems where the polynomial describing their hardness has a big exponent - but if we prove P != NP we'll still be a long way from having a cryptosystem we can prove good things about without making any assumptions about hardness.
no subject
Date: 2010-08-13 09:50 am (UTC)(And thanks - I really appreciate this)
no subject
Date: 2010-08-13 09:58 am (UTC)A P problem is one that can be solved in poly time.
An NP problem is one where you can check the solution in poly time.
If P != NP, there are problems that don't have efficient solutions, even though if we saw the solution we could efficiently check it.
no subject
Date: 2010-08-13 10:03 am (UTC)I'm not going to mention poly time - the whole point was not to turn off people who go blank at the mention of things like that. I need to use examples that are in simple english, so that people can grasp at them. Without examples most people don't grasp things.
no subject
Date: 2010-08-13 10:54 am (UTC)"Dude. One way to measure P vs. NP's importance is this. If NP problems were feasible, then mathematical creativity could be automated. The ability to check a proof would entail the ability to find one. Every Apple II, every Commodore, would have the reasoning power of Archimedes or Gauss. So by just programming your computer and letting it run, presumably you could immediately solve not only P vs. NP, but also the other six Clay problems. (Or five, now that Poincaré is down.) "