Hardware acceleration for similarity computations of feature vectors

Abstract

In Web mining applications, an enormous amount of data needs to be processed quickly and efficiently. Thus, hardware support is crucial to enhance the performance of these operations. In this work, chip-level hardware support for similarity measure calculation is introduced and then extended to similarity matrix computation. Both software and hardware versions of various similarity computations are implemented with a hierarchical platform-based design approach to facilitate component reuse at different levels of abstraction. Software modules are executed on a reconfigurable soft processor core on the same field-programmable gate array (FPGA) as the hardware implementations for the purpose of performance comparison. Preliminary results using parallel hardware are also presented. In addition, performance gain from hardware as opposed to a general-purpose processor is examined. Experimental results show that the performance of similarity computation for a set of feature vectors could be significantly enhanced by using on-chip hardware.