Cook's Theorem On NP-completeness SATISFIABILITY Is NP-complete
Cook's Theorem On NP-completeness SATISFIABILITY Is NP-complete
TIMING CONSIDERATIONS
Horizontal axis: n=number of preferences about party guests;
Vertical axis: time, measured in, say, seconds, required to decide
if there is a satisfying guest list.
Compare the polynomial n5 -time solution to the exponential-
time 2n one, as n increases beyond 20. Exponential-time is always
infinitely slower than polynomial-time as problem size becomes
infinitely large.
The only known solutions to SATISFIABILITY essentially involve checking every possible guest list and this takes exponential time: 2n steps. The problem
is generally believed to lie outside the class P of decision problems with a quick (Polynomial) solution. This remains true even if everybody has exactly three
preferences (so-called 3-Satisfiability). But the problem belongs to NP and this property, of admitting polynomially checkable candidates, turns out to be
fundamental (unlike the distinction between decision problems and optimisation problems — find an actual guest list! — which is a theoretical convenience).
The theorem appears to refer only to a very specific problem but relates to one of the most famous open questions in math-
ematics, referred to as P=NP? Cook’s 1971 theorem says that if we can find a polynomial algorithm for SATISFIABILITY
then the classes P and NP are identical, and we can solve all NP problems in polynomial time, no matter how different from
SATISFIABILITY they might appear to be. Leonid Levin published essentially the same result in 1973 and this theorem is
also known as the Cook-Levin Theorem.
Web link: rjlipton.wordpress.com/about/: if P=NP? is settled you will read about it here first ... unless you are the one winning the
$1 million: en.wikipedia.org/wiki/Millennium Prize Problems!
Further reading: Computers and Intractability by M.R. Garey and D.S. Johnson, W.H.Freeman & Co Ltd, 1979.
Created by Robin Whitty for www.theoremoftheday.org