marc.geddes.domain.name.hidden wrote:
>
>
> On Sep 10, 5:06 pm, Brent Meeker <meeke....domain.name.hidden> wrote:
>
>> Yes there is. In fact descriptions with fewer free parameters are automatically
>> favored by Bayesian inference.
>>
>> http://quasar.as.utexas.edu/papers/ockham.pdf
>>
>> Brent Meeker
>>
>
> Nice try. That's an interesting paper, but it's merely one guys
> attempt to try to define the problem in terms of Bayesianism. It does
> not provide solutions to (a) and (b), which remain unresolved.
I didn't say it solved all problems. I just pointed out that Bayesian inference
does inherently favor simplicity.
>
> These types of attempts to try to reduce Occam's razor to Bayes soon
> run into a big big problem, which I have already mentioned:
>
> There is more than one meaure of complexity. For example,
> *information* is not the same thing as *knowledge*. Shannon
> information is simply a measure of the degree of randomness in a
> string, whereas *knowledge* is more a measure of the amount of work
> that went into producing a string (ie it is *meaningful* information).
Knowledge is usually defined as true belief that is casually connected to the
facts that make it true. That has nothing to do with work (free energy?
computational steps?). You can certainly do a lot of work and end up with a
false belief.
>
> Effective use of Occam's razor also requires us to judge the
> simplicity/complexity of *meaningful information* (ie knowledge), not
> just Shannon information. Bayesianism Induction cannot possibly do
> this, since it cannot handle the *semantics* (meaning) of the
> information, only the Shannon information.
Bayesian inference only assigns probabilities to propositions in such a way as
to maintain a certain kind of consistency. It already assumes that these
propositions have meanings - otherwise it would be impossible to say what it
meant for one to have a certain probability. It's just an extension of logic to
allow values between "true" and "false".
>This it is because it only
> deals with the *functional* aspects of information... ie patterns as
> they appear to external observers, rather than what the patterns
> signify ( the *semantic* aspects of information).
But patterns only signify (have a semantic meaning) in a context that includes
action and goals. How information influences those actions provides a
functional definition of it's content.
Brent Meeker
> >
>
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list.domain.name.hidden
To unsubscribe from this group, send email to everything-list+unsubscribe.domain.name.hidden
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---
Received on Wed Sep 10 2008 - 02:13:44 PDT