Re: Serious *Mistake* by Schmidthuber

From: Bill Jefferys <bill.domain.name.hidden>
Date: Sat, 7 Sep 2002 15:08:51 -0500

At 6:15 PM -0700 on 9/6/02, Osher Doctorow wrote:
>From: Osher Doctorow osher.domain.name.hidden, Fri. Sept. 6, 2002 6:17PM
>
>Bill Jefferys says:
>
>< Nonsense. It's done all the time for events of low probability.
>
>If *doing something all the time* is your reply to nonsense, then can I
>assume that not doing something is your reply to *sense*? Ah well, the
>subtleties of logic!
>
>Do you really want to argue about division by 0 and near 0 denominator?
>Why don't you think about if for a few days or weeks. I would hate to see
>you lose so easily.

Who said anything about dividing by zero? You said:

> >Thus, not only can
> >conditional probability not model events of probability 0, but it cannot
> >even model events of probability close to 0 (Rare Events).

Proof by example:

Example 1: The probability that if I toss a fair coin it will land on
its edge is pretty close to zero, say epsilon.

        P(edge|fair coin tossed fairly)<<1

Are you saying that I cannot model this using conditional
probability? I just did.

Example 2: P(A|~A)=0, for any A. Are you saying that I cannot model
this using conditional probability? I just did. [Note: ~ means "not"]

Example 3: I can calculate the conditional probability that someone
else at the poker table has a straight flush, given that I have a
particular straight flush. This is very low probability. Are you
saying that I can't model this with conditional probability? This is
just a problem in combinatorics, and can be done.

Note that never have to divide by zero, either.

I am guessing that you must have something else in mind, but I sure
don't know what it is.

Bill
Received on Sat Sep 07 2002 - 13:15:17 PDT