KDI: Automated Learning in Network Traffic Control

点击翻译按钮获取中文摘要

9873469Mishra, Bhubaneswar, Parikh, RohitNew York UniversityKDI: Automated Learning in Network Traffic ControlThis is a design and implementation study of automated intelligent agents suitable for controlling and optimizing resource allocation in large-scale networks. This area is characterized by sparsely-connected, computational elements that interact solely through the use of joint resources in massively-distributed environments. The mathematical underpinnings of this research program are found in economics. In contrast to classical theory, however, the underlying computational model of strategy selection here involves inductive learning and discovery processes. The here goal is to provide a rigorous treatment of the reasoning behavior of automated agents by using belief-revision logics to model belief-based learning. The long term significance of the proposed research is the dissemination of automated agents suitable for control and optimization in large-scale, geographically distributed networks, without the benefit of common knowledge of the state of the system.