| 本文已被：浏览 4823次 下载 6033次 |
码上扫一扫！ |
|
|
| 一种主动容错的序列流并行分析算法 |
|
赵峰1,2, 李庆华1,2, 金莉1
|
|
1.华中科技大学,计算机科学与技术学院,湖北,武汉,430074;2.国家高性能计算中心(武汉),湖北,武汉,430074
|
|
| 摘要: |
| 提出一种主动容错的序列流并行分析算法--FTPSA算法(proactive fault-tolerant parallel sequence stream analysis algorithm),以解决噪声环境下大规模序列流的自适应分析问题.算法利用学习网络描述流序列,并存于0-1矩阵中;将低比例和高比例不良数据分层考虑,分别采用基于容错和基于结构优化的学习方法;同时,经过全局筛选,有效地减少了中间结果集合,降低了内存和通信消耗.真实数据集上的实验结果表明,FTPSA算法准确率高,占用的存储空间小,并有良好的容错性和扩展性. |
| 关键词: 序列流 主动容错 知识学习 并行算法 |
| DOI： |
| 分类号: |
| 基金项目:Supported by the National Natural Science Foundation of China under Grant Nos.60503048, 60273075 (国家自然科学基金) |
|
| A Parallel Analysis Algorithm for Sequence Stream Based on Proactive Fault Tolerance |
|
ZHAO Feng,LI Qing-Hua,JIN Li
|
| Abstract: |
| A parallel sequence stream analysis algorithm named FTPSA (proactive fault-tolerant parallel sequence stream analysis algorithm) is proposed in order to deal with sequence stream’s adaptive analysis in noisy environment, which is based on proactive fault-tolerant knowledge learning. The algorithm utilizes learning network to describe sequence stream and stores those in 0-1 matrix, delaminates the low-proportion and the high-proportion noisy data and utilizes fault-tolerant and structure-optimize learning methods, utilizes global filtration to depress memory cost and communication cost. The experimental results on real stream show that FTPSA algorithm is more fault-tolerant, scaleable, accurate, and less memory. |
| Key words: sequence stream proactive fault tolerance knowledge learning parallel algorithm |
