Roger Critchlow wrote:
It's not obvious how to program the same capabilities without the
teaching the agents how to apply the same common sense and expert
senses which the programmer uses to frame a fixed solution space.
Framing a fixed solution space is tricky! Take a data set with lots of
possible signals and chop it in two pieces. If one set gives you a
variable with states A, B and the other set has the same variables with
only C and D then a model that predicts on distinctions between A and B
won't be useful. Using the other half of the data to inform what can
be considered a stable signal is in some sense cheating as much as
fitting a model to all of the data and then testing using it too.
Marcus
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org
