Sujet : Re: Expert systems in forth
De : albert (at) *nospam* spenarnc.xs4all.nl
Groupes : comp.lang.forthDate : 05. Jan 2025, 18:51:17
Autres entêtes
Organisation : KPN B.V.
Message-ID : <nnd$28b6bf50$1efdeaad@10290eb862f73416>
References : 1 2 3 4
User-Agent : trn 4.0-test77 (Sep 1, 2010)
In article <
2025Jan5.160913@mips.complang.tuwien.ac.at>,
Anton Ertl <
anton@mips.complang.tuwien.ac.at> wrote:
melahi_ahmed@yahoo.fr (ahmed) writes:
I'll see how to change the flow of the inference using the action field
of facts and executing them during the inference, like this we can
choose the next rule to use.
>
Potential improvements:
>
Also have rules that work for both truth and falsness. E.g., for
non-extinct animals, all birds have feathers and only birds have
feathers. So if you ask the "feathers" question, and you get a "yes",
you know it is a bird, and if you get a "no", you know that it is no
bird.
>
And then you do not need to ask about wings and egg-laying unless the
answer is "don't know" (supporting that would be another improvement).
>
s" platypus :- swim , not-fly , eat-meat , hoofs , hair .;" >rules
>
It seems to me that the platypus has claws, not hoofs. The most
puzzling property of the platypus, though, is that it is a mammal and
lays eggs.
I had an animals database code in c. I considered a property as true only
if the majority of the respondents considered it true. That weeds out
unanswerable questions whether a leopard has mainly sweat glands on its belly.
You were supposed to have an animal in mind and answer the questions.
At the end you are left with a correct answer or undistingishable animals.
In game theory fashion the questions are selected by chance to give
the most information, and the answers are accumulated, such that
a good question goes to the fore. (In game theory you are supposed
to try unfavourable strategies once in a while. If you have a
solid reputation as a poker player, you can shove all in with
2 8 not suited, once in a while.)
I imagine that it was a good medical database. If the questions are
"has the patient a rash of a type similar to figure 10a"
the answers are definitive, not based on stereotypical images.
(You can ask a three year old whether an elephant has a trunk,
before she have ever seen an elephant.)
Then there is the possibility to attach costs for each question. "Has
the patient globules in his liver, revealed by an MRI scan?". If there
are cost effective questions to be answered, that eliminates diagnoses,
these would be favoured first.
Now AI takes over. A simple metafysical database where you have
decide whether this is hoofs or claws, is old fashioned.
It reminds me of Plato where the idea of hoofs exist independent
of the human minds. Where hoofs are a shadow of the ideal hoofs
outside of the cave.
- anton
Groetje Albert
-- Temu exploits Christians: (Disclaimer, only 10 apostles)Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style WallArt For Home, Office And Garden Decor - Perfect For Windows, Bars,And Gifts For Friends Family And Colleagues.