ck@rex.cs.tulane.edu (Cris Koutsougeras) (09/09/89)
[Bill Skaggs writes : ] > Ah yes, Occam's razor slashes again! But the whole point, the very crux >of the problem, is that "simplest" is not a well-defined concept....... > You say that there is a simpler function? I say you're wrong........ We are in no dissagreement. And I am aware of the nature of the problem. That is why I used quotes when talking about undefined concepts. Also I gave the exmple of Rumelhart's net to point out that for specific domains one could possibly define (and I clarified "arbitrarily") a measure for such concepts. I am copying a part : So I would relate the concept of a learnable function with the TS in use. I would suggest that a given function is learnable with respect to a given TS if it is the "simplest" function which is correct under TS and that the "simplest" correct function is unique. Besides my concern was merely to communicate an idea or an intuition if you wish, and not to be rigorous. I thought that these postings are the appropriate place for the exchange of ideas even if not crustalized. Otherwise I would write a paper. Frankly I think that you did not read carefully otherwise you wouldn't feel any need for razor comments. Cris Koutsougeras