Iterated Phantom Induction: A Little Knowledge Can Go a Long Way

Weadvance a knowledge-based learning method that augments conventional generalization to permit concept acquisition in failure domains. These are domains in whichlearning must proceed exclusively with failure examplesthat are relatively uninformative for conventional methods. A domain theory is used to explain and then systematically perturb the observed failures so that they can be treated as if they were positive training examples. The concept induced from these "phantom"examplesis exercised in the world, yielding additional observations, and the process repeats. Surprisingly, an accurate concept can often be learned even if the phantomexamplesare themselves failures and the domaintheory is only imprecise and approximate. Weinvestigate the behavior of the methodin a stylized air-hockey domain which demandsa nonlinear decision concept. Learning is shownempirically to be robust in the face of degraded domain knowledge. Aninterpretation is advanced which indicates that the in...

• Type: Conference
• Year: 1998
• Where: AAAI
• Authors: Mark Brodie, Gerald DeJong
• Keywords: AAAI 1998, Concept Acquisition, Intelligent Agents, Knowledge-based Learning Method, Nonlinear Decision Concept