基于MapReduce的高阶矩阵乘法分布式并行算法研究

发布时间：2018-04-11 01:23

本文选题：MapReduce + 高阶矩阵　；参考：《小型微型计算机系统》2015年12期

【摘要】：高阶矩阵的存储和处理在信息、经济、生物等学科领域都有十分重要的应用,但是由于单节点计算机CPU、内存等资源的限制,导致了对高阶矩阵的处理存在一定的困难.在研究云计算平台Hadoop及其核心组件MapReduce的基础上,研究实现了处理高阶矩阵乘法的通用并行算法(内积法),在此基础上,对内积法进行了改进,提出一种基于缓存的分布式并行算法(缓存法),通过实验仿真表明,缓存法相比内积法执行效率更高,不仅适合处理高阶稀疏矩阵,而且可以处理高阶稠密矩阵,并且在并行效果上接近理论线性加速比.
[Abstract]:The storage and processing of high order matrices are very important in the fields of information, economy, biology and so on. However, because of the limitation of resources such as single node computer CPU and memory, it is difficult to deal with higher order matrices.Based on the research of cloud computing platform Hadoop and its core component MapReduce, a general parallel algorithm (inner product method) for dealing with high order matrix multiplication is developed. On this basis, the inner product method is improved.A cache based distributed parallel algorithm (cache method) is proposed. The experimental results show that the cache method is more efficient than the inner product method, which is not only suitable for dealing with high order sparse matrix, but also can deal with high order dense matrix.And the parallel effect is close to the theoretical linear speedup.
【作者单位】：中国地质大学武汉计算机学院;
【基金】：国家自然科学基金青年项目(61305087,61402425)资助;国家自然科学基金面上项目(61272470)资助中国博士后科学基金项目(2014M562086)资助
【分类号】：TP338.8

【参考文献】