2019/05/12 by Aditya Modi, Modi, Aditya, Debadeepta Dey +11 · 1 citation
Computer Science · #Software Engineering Research #Mobile Crowdsensing and Crowdsourcing #Advanced Software Engineering Methodologies
paper · pdf · doi:10.48550/arxiv.1905.05179
Assemblies of modular subsystems are being pressed into service to perform\nsensing, reasoning, and decision making in high-stakes, time-critical tasks in\nsuch areas as transportation, healthcare, and industrial automation. We address\nthe opportunity to maximize the utility of an overall computing system by\nemploying reinforcement learning to guide the configuration of the set of\ninteracting modules that comprise the system. The challenge of doing\nsystem-wide optimization is a combinatorial problem. Local attempts to boost\nthe performance of a specific module by modifying its configuration often leads\nto losses in overall utility of the system's performance as the distribution of\ninputs to downstream modules changes drastically. We present metareasoning\ntechniques which consider a rich representation of the input, monitor the state\nof the entire pipeline, and adjust the configuration of modules on-the-fly so\nas to maximize the utility of a system's operation. We show significant\nimprovement in both real-world and synthetic pipelines across a variety of\nreinforcement learning techniques.\n