Learning Mechanisms
Learning Mechanisms of Theo
Macros are formed by examining the explanations of previously successful
inferences. The TMAC (Theo-MACros) system achieves this for Theo by composing a sequence
of successful inference steps, computing a generalization of that sequence,
then storing this macro. This accomplishes a type of explanation-based
learning. These explanations record dependencies among beliefs, so that when a belief is changed, Theo can remove dependent beliefs. This provides a simple forgetting mechanism for truth-maintenance, as well as a form of abstraction.
To return, press HOME.
To go to the next document, press NEXT.