Hardware Acceleration of Persistent Homology Computation
Hardware Acceleration of Persistent Homology Computation
复制标题
持久同调计算的硬件加速
DOI：
10.1007/978-3-030-33642-4_9
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Chen, C
中科院分区：
文献类型：
--
作者：
Wang, F;Deng, C;Yuan, B;Chen, C
持久同源性作为一种强有力的拓扑数据分析工具，能够鲁棒地捕捉数据的拓扑结构。它的相关信息总结在一个持久性图，记录拓扑结构，以及他们的显着性。近年来，人们对各种结构域的持续同源性越来越感兴趣。在生物医学图像分析中，持续同源性已被应用于脑图像、神经元图像、心脏图像和癌症病理图像。同时，由于在一个称为边界矩阵的大矩阵上进行列运算，持久同调的计算可能是耗时的。本文旨在通过边界矩阵列运算的硬件实现来加速持久同调计算。通过设计专用硬件来处理快速矩阵约简，所提出的硬件加速器可能实现高达20 k-30 k倍的加速。
As a powerful tool for topological data analysis, persistent homology captures topological structures of data in a robust manner. Its pertinent information is summarized in a persistence diagram, which records topological structures, as well as their saliency. Recent years have witnessed an increased interest of persistent homology in various domains. In biomedical image analysis, persistent homology has been applied to brain images, neuron images, cardiac images and cancer pathology images. Meanwhile, the computation of persistent homology could be time-consuming due to column operations over a large matrix, called the boundary matrix. This paper seeks to accelerate persistent homology computation with a hardware implementation of the column operations of the boundary matrix. By designing a dedicated hardware to process fast matrix reduction, the proposed hardware accelerator could potentially achieve up to 20k–30k times speed-up.