On 9/6/06, M Riad <[EMAIL PROTECTED]> wrote:
>
> Interesting. ILP is new for me. I did some basic reading and it's really a different form of supervised learning. But I still don't see how this can help build general knowledge. Using your bottle example, lets assume your ILP system recognizes bottles in general, does it understand the properties of a bottle (eg can it use a bucket if no bottles are available) or is it an entity of its own. Maybe a better example would be with animals, If your ILP recognizes dogs and cats, will it also inherently recognize four-legged creatures or does it need a different logic program for that?
>
> Maybe you deal with this elsewhere in your model, but it would be very interesting if this kind of knowledge is intrinsically represented.
>
> Interesting. ILP is new for me. I did some basic reading and it's really a different form of supervised learning. But I still don't see how this can help build general knowledge. Using your bottle example, lets assume your ILP system recognizes bottles in general, does it understand the properties of a bottle (eg can it use a bucket if no bottles are available) or is it an entity of its own. Maybe a better example would be with animals, If your ILP recognizes dogs and cats, will it also inherently recognize four-legged creatures or does it need a different logic program for that?
>
> Maybe you deal with this elsewhere in your model, but it would be very interesting if this kind of knowledge is intrinsically represented.
I think in most cases, what we need for pattern definitions are just conjunctions and disjunctions of many predicates. So calling them "Prolog programs" may be an exaggeration.
In the case of animals, there may be a predicate for "mammal with 4 legs" which is in turn used to define "cats" and "dogs", and in doing so makes the latter definitions simpler. In other words, we may extract basic predicates that simplify the definition of other things. I think this is already done in some ILP algorithms.
YKY
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
