Edge AI – A Promising Technology

Edge Artificial Intelligence (Edge AI) has become the buzzword for every industry organization. Edge intelligence utilizes edge computing to access and analyze the data from locally harvested areas and use artificial intelligence (AI) that enables the machine to make accurate decisions and predictions of such data. Two significant reasons for the efficacy of the deployment of AI models at the edge are innovations of sophisticated computing frameworks like Deep Neural Networks (DNNs) and advances in computing infrastructure. The challenges of deploying DNNs on edge are their huge memory requirement and computational complexity. DNN Compression techniques minimize the number of parameters and bits required without much accuracy loss. It reduces the memory and bandwidth requirements of the DNNs to best fit edge devices. In this chapter, we discuss different DNN model compression techniques and devices best suited for acceleration, considering multiple factors like the type of compressions and device characteristics.