U.S.-U.K. Cooperative Research: Constructing Diagnostic Explanations Using Schema-Structured Bayesian Networks

点击翻译按钮获取中文摘要

9900485LugerThis three-year award supports US-UK collaborative research in knowledge and cognitive systems between George Luger of the University of New Mexico and Brendan McGonigle at the University of Edinburgh's Laboratory for Cognitive Neuroscience and Intelligent Systems. Their research involves the development of probabilistic models of diagnostic reasoning and testing of new algorithms developed by the US investigator. Both groups are concerned with performance improvement of robots through error identification and failure recovery. The investigators propose that iterative performance will improve by embedding abductive reasoning in the robot's environment. The US investigator brings to this collaboration expertise in schema-based abduction, a form of causal reasoning, which employs Bayesian Networks as the underlying representation. This is complemented by the Edinburgh group's expertise in intelligent systems and takes advantage of learning experiments utilizing their NOMAD robot. Collaboration with the Edinburgh group provides an opportunity to study the cognitive plausibility of the US investigator's proposed model of diagnostic reasoning.