Convergence of Deep Learning and Edge Computing using Model Optimization

Edge systems are undergoing a groundbreaking computing evolution to support artificial intelligence, deep learning, and complex computational algorithms. Using cloud servers to perform deep learning model inference poses challenges such as response delays, increased communication costs, and data privacy concerns. Therefore, significant efforts have been made to push the processing of deep learning models to edge systems, which has led to the creation of edge intelligence as the intersection of learning and edge computing. Learning models, especially deep convolutional neural networks, have made significant achievements in machine vision, which provide high accuracy and predictability by spending computing power and memory. If these models are optimized and deployed on edge systems, there will be a revolution in the applications of edge systems in real time. In this paper, by using optimization techniques such as quantization, weight pruning, and weight clustering, the possibility of deploying a typical convolutional neural network model on edge systems that have limited computing resources and memory is investigated. The results show that by using a collaborative algorithm, despite the slight decrease in the accuracy of the model, it is possible to achieve a small-sized model that can even be deployed on microcontrollers.