An instance-based state representation for network repair

Abstract

We describe a formal framework for diagnosis and re-pair problems that shares elements of the well known partially observable MDP and cost-sensitive classifica-tion models. Our cost-sensitive fault remediation model is amenable to implementation as a reinforcement-learning system, and we describe an instance-based state representation that is compatible with learning and planning in this framework. We demonstrate a system that uses these ideas to learn to efficiently restore net-work connectivity after a failure.