Neuromorphic Computing Using Memristor Crossbar Networks: A Focus on Bio-Inspired Approaches

Neuromorphic computing systems, which employ electronic circuits based on digital and analog components to mimic the neurobiological structures in nervous systems, have attracted broad interest as a promising approach for future computing applications [1]-[3]. The interest is driven both from the bottom-up, where continued performance gains following Moore's law have become increasingly harder to obtain [4], [5], and from the top-down, where prolific applications of data-centric tasks such as artificial intelligence [6]-[8], bigdata analysis [9]-[11], and large-scale numerical simulations [12], [13] demand computer architectures that can efficiently address the von Neumann bottleneck [14]. Neuromorphic computing allows massive amounts of data to be processed in parallel, potentially with minimal data movement, thus offering a promising solution for current and future computing needs.