home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.ai.neural-nets:3174 comp.ai:3077 comp.theory:1752
- Newsgroups: comp.ai.neural-nets,comp.ai,comp.theory
- Path: sparky!uunet!elroy.jpl.nasa.gov!usc!cs.utexas.edu!torn!watserv2.uwaterloo.ca!watmath!jfbuss
- From: jfbuss@math.uwaterloo.ca (Jonathan Buss)
- Subject: Re: measuring complexity (was Re: Kolmogorov Complexity)
- Message-ID: <BsxH2n.Ks4@math.uwaterloo.ca>
- Organization: University of Waterloo
- References: <1992Aug10.163725.26986@julian.uwo.ca> <1992Aug10.115352.1@acad3.alaska.edu> <1992Aug12.150857.219@access.digex.com>
- Date: Thu, 13 Aug 1992 15:15:58 GMT
- Lines: 25
-
- In article <1992Aug12.150857.219@access.digex.com> dzik@access.digex.com (Joseph Dzikiewicz) writes:
-
- >This lack of infinite memory is why digital computers are of equal
- >computing power to a finite state machine and considerably less powerful
- >than a Turing machine. Admittedly, a computer with megabytes of memory
- >is a finite state machine with a lot of states, but it's still not
- >Turing-equivalent.
- >
-
- This observation begs the question, why are digital computers modelled
- as infinite-state machines? Can one obtain more realistic results by
- using finite-state machines as the model?
-
- To particularize, do results like the following have finite-state
- analogues?
-
- (1) Mergesort has time complexity O(n log n).
-
- (2) The Satisfiability problem is NP-complete.
-
- These seem to be useful results, but their statement implicitly uses
- an infinite-state model.
-
-
- Jonathan Buss
-