并行绘制系统中基于多响应随机森林的负载估计  

Multivariate Random Forests Based Load Estimation in Parallel Rendering System

作　　者：段思羽 贾文娟 DUAN Si-yu;JIA Wen-juan(College of Computer Science,Sichuan University,Chengdu 610065)

机构地区：[1]四川大学计算机学院,成都610000

出　　处：《现代计算机》2018年第24期22-25,共4页Modern Computer

摘　　要：基于屏幕空间调度的并行绘制系统中,提高绘制效率的关键在于达到多个绘制结点间的负载平衡。若是可以准确预知当前帧的负载分布就能够得出任务的分配方式,以实现负载平衡;使用多响应的随机森林(MRF),并针对绘制数据不平衡的特点进行改进,预测负载的分布情况,在预测准确性上有良好的效果。In screen based and dispatched parallel rendering system, the key to improving rendering efficiency is to achieve load balancing among multiple rendering nodes. If the load distribution of the current frame can be accurately predicted, the task allocation can be obtained to achieve load balancing. Uses a Multivariate Random Forest(MRF) and improves the characteristics of the data imbalance to predict the distribution of the load. It has an excellent effect on prediction accuracy.

关 键 词：回归森林 多响应 不平衡数据 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP181[自动化与计算机技术—计算机科学与技术]