Xiaorang Guo: HW/SW Co-design of K-means Clustering Algorithm on Zynq MPSoC (Project work)

29.09.2021, 11:00 am

Invitation to the presentation of Mr. Xiaorang Guo

Topic: HW/SW Co-design of K-means Clustering Algorithm on Zynq MPSoC

Project: Project work

Supervisors: Ahmed Kamaledin, Muhammad Ali

Abstract: K-means clustering algorithm is a famous unsupervised machine learning algorithm for classifications. Because of its simple control flow and fine-grain parallelism, this clustering algorithm is very suitable for optimization on the FPGA. This project work presents an optimized K-means algorithm by using the HLS tool to reduce the overall latency, and before the acceleration, the time-consuming part is analyzed and identified. The optimized algorithm is packaged into an HLS IP and is integrated with the SW part using the Xilinx tool, where the integration is finished using the Jupyter Notebook integrated in the Linux system of the Pynq-Z1 board. In the end, the execution time, resource utilization, and power will be evaluated and compared between the pure SW implementation and the HW/SW co-design system.