home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!saimiri.primate.wisc.edu!ames!news.hawaii.edu!tarski!ross
- From: ross@tarski.tmc.edu (David Ross)
- Subject: Re: nonstandard analysis
- Message-ID: <1992Dec14.212618.17566@news.Hawaii.Edu>
- Sender: root@news.Hawaii.Edu (News Service)
- Nntp-Posting-Host: tarski.math.hawaii.edu
- Organization: University of Hawaii Mathematics Department
- References: <1992Dec6.025006.16915@athena.mit.edu> <24210@galaxy.ucr.edu> <1992Dec11.234035.1668@news.Hawaii.Edu> <Bz5F04.5p3@mentor.cc.purdue.edu>
- Date: Mon, 14 Dec 1992 21:26:18 GMT
- Lines: 41
-
- In article <Bz5F04.5p3@mentor.cc.purdue.edu> hrubin@mentor.cc.purdue.edu (Herman Rubin) writes:
- >In article <1992Dec11.234035.1668@news.Hawaii.Edu> ross@tarski.tmc.edu (David Ross) writes:
- >
- >>There are many results in analysis, especially probability theory, for which
- >>the only known proofs use nonstandard analysis.
- >
- >This is at best illusory. Nonstandard analysis is useful in finding theorems
- >and simplifying arguments, but with caveats involved in interpretation.
- >
- >However, since any nonstandard proof of a standard theorem can be mechanically
- >translated into a standard proof
-
- This is only true in a very weak sense. Nonstandard arguments which just
- use Los Transfer can indeed be translated mechanically into standard
- arguments; since the first 15 years of n.s.a. arguments relied solely
- on transfer, the argument made sense through, say, 1975 (and still
- applies to arguments which use IST).
-
- However, 'nonstandard hull' arguments, which have become common in
- measure theory and functional analysis, are not solely transfer-based, so
- are not amenable to easy translation, except perhaps by a 'local
- recreation' of the nonstandard model (i.e., taking a sufficiently
- saturated ultraproduct of a Banach space or of a measure space). I
- know of several results - for example, Keisler's strong existence
- theorems for stochastic differential equations, Arkeryd's work on the
- Boltzmann equation, some of my own results on stochastic couplings -
- for which, while there are possibly standard proofs without the
- aforementioned local re-creation, such proofs are not obtainable by
- mechanical translation and are not easy to see. (Even arguments where
- the nonstandard hull is used primarily to simplify a weak limit argument
- - for example, in the superprocess constructions of Doug Perkins - it
- conversion is not automatic.)
-
-
- - David
-
-
- --
- David Ross, Dept. of Math., Univ. of Hawaii at Manoa, Honolulu HI 96822
- Internet: ross@math.hawaii.edu -or- ross@tarski.math.hawaii.edu
- Phone: 808-956-9949
-