Learning-Augmented Algorithms

Learning-Augmented Online Decision Making
Many system problems such as robot tracking, data center resource provisioning, battery management for EV charging station, cooling system control, etc can be modeled as online decision processes. For various online decision processes, we design novel learning-augmented algorithms to minimize the expected cost (or maximize the expected reward ) while guaranteeing anytime competitiveness. The anytime competitiveness requires that for any sequence, the cost is upper bounded by the cost of an expert policy prior scaling by a preset parameter.