Interactive Learning in Noisy and Changing Environments

点击翻译按钮获取中文摘要

9625557 Kuh This project will study learning models in noisy and changing environments. The environments are generally unknown and are learned through observation examples presented at discrete times. Previous research has modeled and analyzed how supervised learning algorithms (algorithms where inputs receive labels) behave when confronted with slowly changing environments. This research will continue to analyze the behavior of supervised learning algorithms, but will focus on more complex reinforcement learning algorithms. The primary objectives are to obtain a better understanding of reinforcement learning methods to problems in signal processing and communications. Two applications areas that are well suited to reinforcement learning approaches are sequential detection problems and networking problems. ***