Learning Hierarchical Skills from Problem Solutions Using Means-Ends Analysis


Humans acquire skills in different ways, one of which involves learning from worked-out solutions to problems. In this paper, we present an extension to the Icarus cognitive architecture that lets it acquire complex hierarchical skills in this manner. Unlike previous work on this topic, our approach relies on an existing architectural mechanism, means-ends analysis, to explain steps in the problem solution and to learn new structures. We illustrate this method in the domains of multi-column subtraction and football, after which we discuss related work and consider directions for future research in this area.

Back to Friday Posters