Brief overview of Edge AI accelerators for energy-constrained edge

Authors:
Publisher website
Google Scholar
Abstract
In the last several years, there has been a big boom in the field of AI, and particularly in a subset called deep learning. The models used in deep learning are often relatively computationally expensive and problematic to execute in real-time in environments with computational and energetic constraints. In order to do so, specialized hardware, so-called AI Accelerators are often utilized. The term AI Accelerators encompass a wide variety of different devices using different technologies. This work surveys a market and introduces an overview of such accelerators for a network's edge constrained by electrical energy.