Probabilistic computing based on magnetic tunnel junction
Abstract
Von Neumann computers based on the Turing machine theory have been developed to an extremely high level. However, facing today’s massive data and computing needs, energy consumption, transmission bandwidth, and other issues make further development challenging. Therefore, a more efficient non-traditional computing architecture is urgently needed to overcome these problems. In recent years, new architectures, such as quantum computing, neuromorphic computing, and probabilistic computing, have emerged. Among them, probabilistic computing, with its great potential in energy saving and high-speed computing, has become a strong candidate for non-traditional computing. As a representative of spintronic devices, the magnetic tunnel junction (MTJ) is an important component of probabilistic computer. This paper reviews the working principles of MTJs, systematically summarizes common MTJ types and tuning methods, and introduces the applications of MTJ-based probabilistic computing. It also explores the potential of alternative materials in probabilistic computing. Finally, the challenges and future directions of probabilistic computing are discussed and anticipated.
