自动化内存垃圾收集机制越来越成为现代编程语言的必备特性.但是目前的垃圾收集算法仍然存在收集不及时,消耗CPU资源过大的问题.通过对基于RBF的神经网络进行训练,智能的决定垃圾收集机制的调用时机.该方法分为预先训练和动态训练,以提高预测的准确率.具体描述了其实现并通过实际的测试和分析证明该方法取得了较好的效果.
Automatic memory garbage collection mechanism is now the necessary feature in modern programming language. The downside of GC is that collection is not called in time and consuming much more CPU resource. The paper presents a RBF neural network based scheduling method to determine the time of GC running. The method is divided into predefined training and dynamic training to improve the precise. The paper describes implementation of the method in detail and analyzes the performance tests which prove that it a practical method.