Research on advancement of computer-intensive methods for statistical estimations along with Graphical Processing Units and development of application software

统计计算新算法的发展需要对计算环境中的趋势进行仔细和持续的关注。近年来，人们对多核CPU（如多核处理器）和GPGPU（通用计算图形处理单元）的使用越来越感兴趣。本研究针对在大型并行计算机上实现涉及大量简单迭代计算的统计算法进行了设计和开发。实际上，我们研究的统计方法是在一个特殊的多层计算体系结构上运行的，每层计算体系结构都是分布式并行计算机系统。因此，大规模并行计算有望缓解由维度诅咒引起的难题。

A Development of new algorithms for statistical computations needs careful and continuous attention to trends in the computational environments. Recently much interest is laid on making use of the CPU with multiple cores such as the multi-core processor and General Purpose computation on Graphical Processing Units (GPGPU). In this study the statistical algorithms that involve a huge number of a simple iterative calculation are designed to develop in aiming at its implementation on the massive parallel computer. Actually we study the statistical methods which runs on a special computational architecture having the multiple layers, each of which is the distributed parallel computer system. As a result, the massive parallel computation is expected to mitigate a difficult problem arising from the curse of dimension.