home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!uvaarpa!darwin.sura.net!sgiblab!spool.mu.edu!howland.reston.ans.net!paladin.american.edu!auvm!BEN.DCIEM.DND.CA!MMT
- Return-Path: <@VMD.CSO.UIUC.EDU:mmt@ben.dciem.dnd.ca>
- Message-ID: <9301271844.AA26025@chroma.dciem.dnd.ca>
- Newsgroups: bit.listserv.csg-l
- Date: Wed, 27 Jan 1993 13:44:19 EST
- Sender: "Control Systems Group Network (CSGnet)" <CSG-L@UIUCVMD.BITNET>
- From: mmt@BEN.DCIEM.DND.CA
- Subject: Re: Occam
- Lines: 41
-
- [Martin Taylor 930127 13:30]
- (Bruce Nevin 930127 08:54:45)
-
- >Martin --
- >
- >I've read through the Occam piece, but have not got to the bottom
- >of my disquiet. I started to write something here, but decided
- >to wait until I have been able to give it more thought.
- >Everything ought to be re-framed, I believe, in terms of word
- >dependencies in the sublanguage grammar for the science in which
- >the hypotheses are stated.
-
- That can't be relevant. Whatever dependencies there are can only
- contribute to the redundancy. The point is that whatever constructs
- and concepts YOU already hold can be used for free in describing new
- stuff for YOU. The hypothesis length of concern is the shortest way
- YOU can describe the new stuff (which includes all the observations
- that you didn't previously describe). So no discussion, of whatever
- kind it might be, about HOW the hypotheses are described, could
- possibly be relevant to the argument.
-
- Is your problem perhaps related to a misconception that the minimum
- description length is an objective quantity? It isn't. It applies
- to YOU and to no-one else. It depends on what YOU already know (or
- believe you know).
-
- "The sublanguage grammar for the science" is of no value in stating
- the hypothesis unless you have already incorporated that sublanguage
- into your conceptual models. It becomes part of the prior assumptions.
- If you already have it, you can use it for free. If you don't, then a
- description of the sublanguage (and perhaps of the science) must be
- included as part of the length of the hypothesis statement.
-
- Before you complete what you were going to write, read my posting of
- 930121 18:30 "more on Occam." There, the idea is extended, to show why
- classical psychologists have trouble accepting PCT (of course, not in
- those words--much more generally than that). That's the posting in which
- I pointed out that there should be no conflict between Rissanen and
- Wallace&Freemen about MDL in the readings list you posted.
-
- Martin
-