System-on-a-chip (SoC)-based Hardware Acceleration for Extreme Learning Machine

Extreme learning machine (ELM) is a popular class of supervised models in machine learning that is used in a wide range of applications, such as image object classification, video content analysis (VCA) and human action recognition. However, ELM classification is a computationally demanding task, and the existing hardware implementations are not efficient for embedded systems. This work addresses the implementation of extreme learning machine (ELM) in a system on a chip field-programmable gate-array (SoC FPGA)-based customized architecture to efficiently utilize hardware accelerator. The optimization process consists of parallelism extraction, algorithm tuning and deep pipelining.