USENIX Association Proceedings of the 2 nd JavaTM Virtual Machine

This article investigates how machine learning methods might enhance current garbage collection techniques in that they contribute to more adaptive solutions. Machine learning is concerned with programs that improve with experience. Machine learning techniques have been successfully applied to a number of real world problems, such as data mining, game playing, medical diagnosis, speech recognition and automated control. Reinforcement learning provides an approach in which an agent interacts with the environment and learns by trial and error rather than from direct training examples. In other words, the learning task is specified by rewards and penalties that indirectly tell the agent what it is supposed to do instead of telling it how to accomplish the task. In this article we outline a framework for applying reinforcement learning to optimize the performance of conventional garbage collectors. In this project we have researched an adaptive decision process that makes decisions regarding which garbage collector technique should be invoked and how it should be applied. The decision is based on information about the memory allocation behavior of currently running applications. The system learns through trial and error to take the optimal actions in an initially unknown environment.