Why Spintronics-Based Neuromorphic Computing?

Neuromorphic computing refers to a wide range of approaches regarding implementing artificial intelligence (AI)/machine learning (ML) algorithms on unconventional hardware (conventional hardware being modern digital computers like CPUs and GPUs). Many of these unconventional hardware architectures are inspired by the internal structure and working of the brain, thereby acquiring the name “neuromorphic.” They also often intertwine the memory and computing elements unlike conventional hardware, which follows the von Neumann architecture and have the memory and computing units separated from each other. So these unconventional architectures are often called in-memory or non-von Neumann architectures as well. Many neuromorphic architectures also use spike-based computation and communication schemes, much like the brain. This chapter provides an introduction to neuromorphic/non-von Neumann computing and discusses which aspects and features regarding the emerging technology of nanomagnetism and spintronics make it interesting for neuromorphic/non-von Neumann implementations. This chapter also provides information on which subsequent chapter of the book covers which aspect in more detail.