opencog-dev team mailing list archive

["Ben Goertzel" <ben@xxxxxxxxxxxx>]

getCount refers to the amount of evidence n associated with a proposition

It may seem odd, but there's no paradox about having a non-integer n

it's really no odder than the notion of, say, 1.73 apples...

A heuristic formula sometimes used is

weightOfEvidence = n/(n+k)

where k is a "personality parameter" ... in this case, clearly, if one
gets a wOfE from an inference formula, the corresponding n need not be
an integer

For indefinite probabilities we use a different formula than the
above, but the phenomenon may be the same

We could of course round n to an integer, but that would just
introduce extra error...

ben

On Mon, Jun 9, 2008 at 1:53 PM, Joel Pitt <joel.pitt@xxxxxxxxx> wrote:
> Hi Trent
>
> On Mon, Jun 9, 2008 at 4:38 PM, Trent Waddington
> <trent.waddington@xxxxxxxxx> wrote:
>> Can someone please document SimpleTruthValue and the subclasses.
>> IndefiniteTruthValue.h has a comment:
>>
>>   float confidenceLevel;  // referred as "b" in the paper
>>
>> what paper?  If I can get it then I'll happily change all these other
>> variables to have descriptive names.  Perhaps it might even shine some
>> light on why getCount() returns a float.
>
> These are related to the indefinite interval probabilities that are an
> integral part of probabilistic logic networks (PLN).
>
> A book by Ben is appearing soon, and there are several papers that he
> may be able to forward to you.
>
> I'm due to start working on PLN later this week or early next, and
> Cesar (a GSoC student) is possibly going to start properly
> implementing the Indefinite truth value mechanics soon. I'll try and
> make the TruthValue docs more clear during this time.
>
> RE: confidenceLevel. This is the chance that after k more observations
> (k is usually a system or inference parameter) the mean of these
> TruthValue strength value will lie within the Indefinite probability
> interval [L U]. Will have to leave it to someone else to answer why
> getCount returns a float however as I haven't started digging into
> that code yet...
>
> J
>
> _______________________________________________
> Mailing list: https://launchpad.net/~opencog-dev
> Post to     : opencog-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/people/+me/+editemails
> More help   : https://help.launchpad.net/ListHelp
>



-- 
Ben Goertzel, PhD
CEO, Novamente LLC and Biomind LLC
Director of Research, SIAI
ben@xxxxxxxxxxxx

"Nothing will ever be attempted if all possible objections must be
first overcome "  - Dr Samuel Johnson