On Tue, 08 May 2001, you wrote: > > Paul C, > I may be missing the point of Occam's razor but this website does not > seem to understand it. As Schumpeter notes, the debate was one > between nominalists and realists, though the realists in maintaining > that only ideas and general concepts had real existence were at odds > with the contemporary meaning of realism. The nominalists on the > other hand argued that there were only singular individuals so that > words designating a group of individuals or concept were only useful > conventions but did not designate a reality and were thus to be > mistrusted. Occam thus maintained that "one should not needlessly > increase the number of abstract entities," that principle of economy > being known as Occam's razor. > > Just the simple injunction to minimize constants in a regression > formula does not seem to get at the heart of Occam's razor. The Minimum Description Length principle is a modern formalisation of the underlying intuition in Occams razor. MDL has application to much more than minimising the constants in a regression formula, though that is one instance of it. The principle comes into its own in automated reasoning systems which have to construct some model to explain input data - for example in computer vision systems. Here the choice between theorems has to be automatic. In such cases one needs a means of measuring the information content of the theorem in general, including constants and other arithmetic operations that one employs to arrive at the answer. Thus if the formula one was using involved very complicated expressions this would also count against the theorem. > > In fact the relationship of Occam's razor to contemporary statistical > work is at another level. For insofar as modern statistical knowledge > attempts to substitute "synthetic substitutes for multiple things", > e.g., THE price index, for increases in price; THE unemployment rate, > for unemployed persons, it attempts to constitute and support > realities of a superior level, while still being grounded in > nominalist and individualist conventions. That the status of reality > is granted to two levels indicates the ground that has been covered > since Occam--as argued by Alain Desrosieres, The Politics of Large > Numbers: A History of Statistical reasoning. p. 70 I am sorry but I do not see the relevance of this paragraph. Perhaps I do not fully understand what you are getting at. > > Yours, Rakesh > > > > > ' -- Paul Cockshott, University of Glasgow, Glasgow, Scotland 0141 330 3125 mobile:07946 476966 paul@cockshott.com http://www.dcs.gla.ac.uk/people/personal/wpc/ http://www.dcs.gla.ac.uk/~wpc/reports/index.html
This archive was generated by hypermail 2b30 : Sat Jun 02 2001 - 00:00:06 EDT