摘要
为了在高吞吐量应用程序中快速进行测量来监视物理过程,论文将通信吞吐量传输到中心服务器之前的测量数据预过滤过程转化为代码路径分支混淆技术,并将其视为二进制分类问题。利用决策树与随机随机森林构建了每种体系结构并提出了不同的实现方案,使用改进的冯·诺依曼体系结构的CPU理论模型和现场可编程门阵列(FPGA)来构建计算体系结构,并在FPGA上实现随机森林的应用。所提方法在Zedboard开发板上进行了实验,试验结果表明所提方法能有效地提高吞吐量和资源量,且降低了主板设备的功耗。
In order to monitor the physical process quickly in high throughput applications,this paper transforms the pre-fil⁃tering process of measured data before the transmission of communication throughput to the central server into code path branch ob⁃fuscation technology,and regards it as a binary classification problem.Each architecture is constructed by using decision tree and random forest,and different implementation schemes are proposed.The computational architecture is constructed by using the CPU theoretical model of improved von Neumann architecture and field programmable gate array(FPGA),and the application of random forest is implemented on FPGA.The proposed method is tested on Zedboard development board.The experimental results show that the proposed method can effectively improve the throughput and resources,and reduce the power consumption of the motherboard equipment.
作者
刘翌
潘小辉
胡浔惠
LIU Yi;PAN Xiaohui;HU Xunhui(State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 210000;Nari Group Limited,Nanjing 210000)
出处
《计算机与数字工程》
2021年第2期353-359,共7页
Computer & Digital Engineering
