Probabilistic Methods in Dynamic Environments

我们建议研究人工智能中的贝叶斯表示和推理方法，特别关注动态环境中的决策。 动态环境为智能系统研究提供了机遇和挑战：在这种环境中表现良好所需的灵活性很可能需要人工智能方法，但人工智能研究传统上集中在更静态的环境中。 本论文在贝叶斯网推理研究的基础上，重点研究了贝叶斯网的真实的实时推理和动态表示构造。 研究结果将有助于我们理解动态环境中决策理论技术的问题和潜力，并为复杂的实时监测，评估和控制应用提供决策理论基础。

We propose research on Bayesian representation and inference methods in AI, focused especially on decision-making in dynamic environments. Dynamic environments provide both an opportunity and a challenge to intelligent systems research: the flexibility required to perform well in such environments will most likely demand AI methods, but AI research traditionally has focused on more static environments. The proposed research builds on our current research on inference in Bayes nets, and focuses on real- time inference and dynamic representation construction. The results will contribute to our understanding of the problems of and potential for decision-theoretic techniques in dynamic environments, and will provide a decision-theoretic foundation for complex real-time monitoring, assessment, and control applications.