Matt's AGI <http://mattmahoney.net/agi2.html> uses an approximates
conditional Kolmogorov Complexity:

K(Y|X)

with conditional compression :

C(Y|X) = C(XY) - C(X)

C compresses and provides the resulting length.
Juxtaposition concatenates, ie "XY" is concatenate(X,Y).
He then uses this to route messages to the most intelligent agent for that
message via a measure semantic distance between strings X and Y:
D(X,Y) =C(Y|X) + C(X|Y)

If one has the less ambitious goal of scoring relative causation of
substrings of the same corpus, one needs a different measure.  Let's say
you have a corpus string Z and two of its substrings X and Y.

PROPOSITION:

We can say X is more causative in Z than is Y if:

CausalScore(Z|X) < CausalScore(Z|Y)
WHERE
CausalScore(A|B) = C(A-B)/Size(A-B)
"-" deletes the right hand string out of the left hand string.

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T9a8649791989a23a-M0b75909812e0a928987f48ea
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to