A Dynamic Bayesian Network Framework for Learning from Observation

Learning from Observation (a.k.a. learning from demonstration) studies how computers can learn to perform complex tasks by observing and thereafter imitating the performance of an expert. Most work on learning from observation assumes that the behavior to be learned can be expressed as a state-to-action mapping. However most behaviors of interest in real applications of learning from observation require remembering past states. We propose a Dynamic Bayesian Network approach to learning from observation that addresses such problem by assuming the existence of non-observable states.