Generalization in Soar
Soar's learning mechanism,
dependency analysis to
determine which working memory elements at the start of an
impasse resulted in the
final resolution of the subgoal.
Since this dependency analysis uses only those working memory
elements that were used directly in impasse resolution, chunking
generalizes implicitly. The situation
that led to a chunk does not need to be repeated exactly the same for a
chunk to fire later; only those elements which lead directly to the chunk
are necessary. Thus, the chunk can fire not only for the situation that
created it but in any situation in which its conditions match.
Sometimes these conditions can be
too specific or too general. However,
this implicit generalization usually works well, leading to
transfer of learning.
To go to the Soar property list, press UP
To return, press HOME.
To go to the next document, press NEXT.