FPGA Based Implementations of RNN and CNN: A Brief Analysis

Abstract

Deep neural network (DNNs) is an extensive field which used for application those have complex nature such as processing of voice and image. It has two main varieties namely Convolutional neural network (CNNs) and recurrent neural network (RNNs) are got recent success in industrial applications. CNN used for applications like image classification, RNNs used for time-variant problems. Even though both belong to the DNNs family they implementations show substantial differences. I get more potential results on Field Programmable Gate Array (FPGA) besides CPU and GPU's implementations. Evolution shows remarkable advantages of FPGA implementations over GPUs and CPUs. In this Research article FPGA implementation of CNNs and RNNs are compared and analyse its optimizations. Benefits and drawbacks of FPGA implementations of deep neural networks are also highlighted.