Micro-Armed Bandit: Lightweight & Reusable Reinforcement Learning for Microarchitecture Decision-Making
Gerasimos Gerogiannis, Josep Torrellas
Abstract
Online Reinforcement Learning (RL) has been adopted as an effective mechanism in various decision-making problems in microarchitecture. Its high adaptability and the ability to learn at runtime are attractive characteristics in microarchitecture settings. However, although hardware RL agents are effective, they suffer from two main problems. First, they have high complexity and storage overhead. This complexity stems from decomposing the environment into a large number of states and then, for each of these states, bookkeeping many action values. Second, many RL agents are engineered for a specific application and are not reusable.
Topics & Concepts
Reinforcement learningMicroarchitectureComputer scienceOverhead (engineering)AdaptabilityComputer architectureEmbedded systemArtificial intelligenceOperating systemEcologyBiologyReinforcement Learning in RoboticsEvolutionary Algorithms and ApplicationsVLSI and FPGA Design Techniques