Hey,

I have a question for someone familiar with the fuzzy artmap algorithm
implemented in python in the "brain" area.

It seems in the classification that they're using fuzzy "and" to calculate
how close the input vectors are to the weight vectors that they build up in
training.  There are two places where this happens -- one where they're
getting the "activation levels" and one where they are calculating the match
value.

The activation levels are only used to order the categories so that
(presumably) the most likely categories get checked first.  Then, the match
is calculated to see if it exceeds the vigilance level to signal a match.

The thing that's concerning is that fuzzy "and" isn't symmetric across these
vectors.  a ^ b != b ^ a.  And, one of the calculations is input ^ weight
and the other is weight ^ input.

Does anyone know what value is derived by the different orientations of the
equation?  (I'm asking because of a low level optimization I could make if
the comparisons were the same... I'm assuming they are not for some good
reason, but I'm not intuiting what that reason might be...)

 Doug
_______________________________________________
Pyro-users mailing list
[EMAIL PROTECTED]
http://emergent.brynmawr.edu/mailman/listinfo/pyro-users

Reply via email to