Relative relevance of subsets of agent's knowledge
Sławomir Nowaczyk, Jacek Malec · 2007
We study agents situated in partially observable environments, who do not have the resources to create conformant plans. Instead, they create con-ditional plans which are partial, and learn from experience to choose the best of them for execution. Our agent employs an incomplete symbolic de-duction system based on Active Logic and Situation Calculus for reasoning about actions and their consequences. An Inductive Logic Programming al-gorithm generalises observations and deduced knowledge in order to distin-guish “bad ” plans early, before agent’s computational resources are wasted on considering them. In this paper we present experiments which show that in order for learn-ing to be successful, an agent’s knowledge needs to be filtered. We argue that this filtering nicely matches the intuitive notion of “knowledge relevance”. We also present a heuristic scheme, combining several natural rules, which can be used to automatically determine which formulae should be used for learning. 1