Embracing Stochasticity to Enable Neuromorphic Computing at the Edge

The advancements of the Internet of Things (IoT) and Artificial Intelligence (AI) have resulted in the proliferation of intelligent computing devices. However, AI-based solutions are often compute-intensive, severely limiting the performance of edge devices operating in energy-constrained environments. As CMOS scaling trends continue to saturate, several emerging technologies have shown the promise of efficient computation. In this paper, we present an overview of devices that exhibit inherent stochasticity and how they form the building blocks of a stochastic computing framework that can execute energy and memory-efficient stochastic algorithms for intelligent computing.