Technical methods for accelerating CNNs for IoT using FPGAs and machine learning

This paper aims to explore Field Programmable Gate Arrays (FPGAs) and their potential in implementing Machine Learning, as well as reviewing the area of acceleration of deep learning networks, in particular, convolution neural networks (CNNs) on FPGAs. The paper begins by exploring machine learning algorithms, in particular, CNNs as they have wide applications in the area of image and video recognition and the sudden shift in direction when it comes to the acceleration of these networks by utilizing hardware accelerators especially FPGAs. Then the paper reviewed methods for optimizing and compressing CNN networks as well as enhancing the design of FPGA hardware to provide efficient designs for the IoT. Finally, recommendations were discussed both in FPGA architecture and in the optimization of neural networks implemented, as well as the deployment of FPGAs in IoT.