Algorithm 版 (精华区)
发信人: ssos (存在与虚无·守拙), 信区: Algorithm
标 题: The Importance of the P versus NP Question
发信站: 哈工大紫丁香 (2003年07月27日11:15:13 星期天), 站内信件
The Importance of the P versus NP Question
STEPHEN COOK
University of Toronto, Toronto, Ont., Canada
The P versus NP problem is to determine whether every language
accepted by some nondeterministic Turing machine in polynomial time is
also accepted by some deterministic Turing machine in polynomial
time. Unquestionably this problem has caught the interest of the
mathematical community. For example, it is the first of seven
milliondollar ``Millennium Prize Problems'' listed by the Clay
Mathematics Institute [www.claymath.org]. The Riemann Hypothesis and
Poincar\"e Conjecture, both mathematical classics, are farther down
the list. On the other hand, Fields Medalist Steve Smale lists P
versus NP as problem number three, after Riemann and Poincar\"e, in
``Mathematical Problems for the Next Century'' [Smale 1998].
But P versus NP is also a problem of central interest in computer
science. It was posed thirty years ago [Cook 1971; Levin 1973] as a
problem concerned with the fundamental limits of feasible
computation. Although this question is front and center in complexity
theory, NPcompleteness proofs have become pervasive in many other
areas of computer science, including artificial intelligence,
databases, programming languages, and computer networks (see Garey and
Johnson [1979] for 300 early examples).
If the question is resolved, what would be the consequences? Consider
first a proof of P=NP. It is possible that the proof is
nonconstructive, in the sense that it does not yield an algorithm for
any NPcomplete problem. Or it might give an impractical algorithm,
for example, running in time n^100 . In either of these cases, the
proof would probably have few practical consequences other than to
disappoint complexity theorists. However, experience has shown that
when natural problems are proved to be in P, a feasible algorithm can
be found. There are potential counterexamples to this assertion; most
famously, the deep results of Robertson and Seymour [1993--1995], who
prove that every minor closed family of graphs can be recognized in
time O(n^3), but their algorithm has such huge constants it is not
practical. But practical algorithms are known for some specific
minorclosed families (such as planar graphs), and possibly could be
found for other examples if sufficient effort is expended.
If P=NP is proved by exhibiting a truly feasible algorithm for an
NPcomplete problem such as SATISFIABILITY (deciding whether a
collection of propositional clauses has a satisfying assignment), the
practical consequences would be stunning. First, most of the hundreds
of problems shown to be NPcomplete can be efficiently reduced to
SATISFIABILITY, so many of the optimization problems important to
industry could be solved. Second, mathematics would be transformed,
because computers could find a formal proof of any theorem which has a
proof of reasonable length. This is because formal proofs (say in
Zermelo--Fraenkel set theory) are easily recognized by efficient
algorithms, and hence bounded proof existence is in NP. Although the
formal proofs may not be intelligible to humans, the problem of
finding intelligible proofs would be reduced to that of finding a good
recognition algorithm for formal proofs. Similar remarks apply to the
fundamental problems of artificial intelligence: planning, natural
language understanding, vision, and even creative endeavors such as
composing music and writing novels. In each case, success would depend
on finding good algorithms for recognizing good results, and this
fundamental problem itself would be aided by the SAT solver by
allowing easy testing of recognition theories.
One negative consequence of a feasible proof that P=NP is that
complexity-based cryptography would become impossible. The security of
the Internet, including most financial transactions, depends on
assumptions that computational problems such as large integer
factoring or breaking DES (the Data Encryption Standard) cannot be
solved feasibly. All of these problems are efficiently reducible to
SATISFIABILITY. (On the other hand, quantum cryptography would survive
a proof of P=NP and might solve the Internet security problem.)
Now consider the consequences of a proof that P≠NP. Such a proof
might just answer the most basic of a long list of important related
questions that could keep complexity theorists busy far in the
future. How large is the time lower bound for SATISFIABILITY: is it
barely superpolynomial or is it truly exponential, or is it in
between? Does it apply just for the worst case inputs, or are there
convincing average case lower bounds [Levin 1986; Gurevich 1991]? What
about lower bounds for NP approximation problems [Vazirani 2001]? Are
there lower bounds for problems such as integer factorization that are
reducible to NP problems but may not be NPhard? In general, proving
the security of cryptographic protocols such as RSA or DES is much
harder than proving P≠NP.
Most complexity theorists, including the author, believe that P ≠ NP
(see Gasarch [2002] for a recent poll). I would summarize the argument
in favor of P ≠ NP by saying that we are really good at inventing
efficient algorithms, but really bad at proving algorithms don't
exist. There are powerful techniques that are part of the standard
undergraduate computer science curriculum for devising efficient algo
rithms for diverse problems. Millions of smart people, including
engineers and programmers, have tried hard for many years to find a
provably efficient algorithm for one or more of the 1000 or so
NPcomplete problems, but without success.
Contrast this with the efforts of the small set of mathematicians who
seriously work on proving P ≠ NP. There are reasons why the main
techniques tried for proving complexity lower bounds may not work for
showing P ≠ NP: a proof based on diagonalization cannot relativize
[Baker et al. 1975], and a proof based on Boolean circuit lower bounds
cannot be ``natural'' [Razborov and Rudich 1997]. Further, there are
natural complexity class separations that we know exist but we cannot
prove. Consider the sequence of complexity class inclusions
LOGSPACE \subseteq P \subseteq NP \subseteq PSPACE.
A simple diagonal argument shows that the first is a proper subset of
the last, so it follows that one of the three adjacent inclusions must
be proper. But no proof is known that any particular one is proper.
Assuming that P ≠ NP, when and if will a proof be found? Apparently
by the year 2100, if one believes the majority opinion from the poll
[Gasarch 2002]. It is difficult to say whether much progress has been
made to date, since there is no convincing program toward finding a
proof. There are recent beautiful results in complexity theory
involving probabilistically checkable proofs [Arora et al. 1998] and
derandomization [Impagliazzo et al. 1999] which create deep incites
into the nature of computation, and it is nice to think that these
ideas will someday contribute to a proof of P ≠ NP.
REFERENCES
ARORA, S., LUND, C., MOTWANI, R., SUDAN, M., AND SZEGEDY,
M. 1998. Proof verification and the hardness of approximation
problems. J. ACM 45, 3 (May), 501--555.
BAKER, T., GILL, J., AND SOLOVAY, R. 1975. Relativizations of the P =?
NP question. SICOMP: SIAM J. Comput.
COOK, S. 1971. The complexity of theoremproving procedures. In
Conference Record of 3rd Annual ACM Symposium on Theory of
Computing. ACM New York, pp. 151--158.
GASARCH, W. 2002. Guest column: The P=?NP poll. SIGACT NEWS 33, 2
(June), 34--47.
GAREY,M .R.,AND JOHNSON, D. S. 1979. Computers and Intractibility, a
Guide to the Theory of NP Completeness. Freeman, San Francisco,
Calif.
GUREVICH, Y. 1991. Average case
completeness. J. Comput. Syst. Sci. 14, 3 (June), 346--398.
IMPAGLIAZZO, R., SHALTIEL, R., AND WIGDERSON, A. 1999. Nearoptimal
conversion of hardness into pseudorandomness. In Proceedings of the
40th Symposium on Foundations of Computer Science. IEEE Computer
Society Press, Los Alamitos, Calif., pp. 181--190.
LEVIN, L. 1973. Universal search problems (in Russian). Problemy
Peredachi Informatsii 9, 3, 265--266. (English translation in
Trakhtenbrot, B. A.: A survey of Russian approaches to Perebor
(bruteforce search) algorithms. Ann. Hist. Comput. 6 (1984),
384--400.)
LEVIN, L. 1986. Average case complete problems. SIAM J. Comput. 15,
285--286.
RAZBOROV,A.A.,AND RUDICH, S. 1997. Natural
proofs. J. Comput. Syst. Sci. 55, 1 (Aug.), 24--35.
ROBERTSON, N., AND SEYMOUR, P. D. 1983--1995. Graph minors
i--xiii. J. Combinat. Theory B.
SMALE,S. 1998. Mathematical problems for the next century. MATHINT:The
Mathematical Intelligencer 20.
VAZIRANI, V. 2001. Approximation Algorithms. SpringerVerlag, Berlin,
Germany.
--
<<社会契约论>>是一本好书,应当多读几遍
风味的肘子味道不错,我还想再吃它
※ 来源:·哈工大紫丁香 bbs.hit.edu.cn·[FROM: 202.118.230.220]
Powered by KBS BBS 2.0 (http://dev.kcn.cn)
页面执行时间:205.167毫秒