home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!usc!sdd.hp.com!ux1.cso.uiuc.edu!cs.uiuc.edu!kadie
- From: kadie@cs.uiuc.edu (Carl M. Kadie)
- Subject: Re: Data for Travelling Salesman Prob Wanted...
- Message-ID: <C19uH5.JIv@cs.uiuc.edu>
- Organization: University of Illinois, Dept. of Comp. Sci., Urbana, IL
- References: <1993Jan21.114216.1196@uoft02.utoledo.edu> <C17ty4.42p@cs.uiuc.edu> <uh311ae.727724965@sunmanager> <C19rLG.FFx@cs.uiuc.edu>
- Date: Fri, 22 Jan 1993 20:05:29 GMT
- Lines: 31
-
- uh311ae@sunmanager.LRZ-Muenchen.DE (Henrik Klagges) writes:
-
- [...]
- check out the paper by fritzke in the /pub/neuroprose archive at the aftp
- server archive.cis.ohio-state.edu. He has a very slick 'cell-based' alg.
- on TSP that has excellent scaling behaviour.
- [...]
-
- From looking at the paper, it looks as though the Fritzke's program
- doesn't guarantee the quality of its solution. Based on the _Discover_
- article, I think the Johnson's method can. Also, Johnson's program
- seems to be 2 or 3 orders of magnitude faster and to produce better
- solutions. For comparison:
-
- Johnson Fritzke
- # of cites: 1000000 2392
- CPU time: 3 cpu hours 1.13 cpu hours
- Quality: within 2% of optimum about 9% of optimum
- Guaranteed
- error bounds: yes (I think) no
-
- - Carl
-
- Sources: _Discover_ magazine, Jan. 1992, p 91
- anonymous ftp
- archive.cis.ohio-state.edu/pub/neuroprose/fritzke.linear-tsp.ps.Z
-
-
- --
- Carl Kadie -- I do not represent any organization; this is just me.
- = kadie@cs.uiuc.edu =
-