In-memory computing with emerging memory devices: Status and outlook

Abstract

In-memory computing (IMC) has emerged as a new computing paradigm able to alleviate or suppress the memory bottleneck, which is the major concern for energy efficiency and latency in modern digital computing. While the IMC concept is simple and promising, the details of its implementation cover a broad range of problems and solutions, including various memory technologies, circuit topologies, and programming/processing algorithms. This Perspective aims at providing an orientation map across the wide topic of IMC. First, the memory technologies will be presented, including both conventional complementary metal-oxide-semiconductor-based and emerging resistive/memristive devices. Then, circuit architectures will be considered, describing their aim and application. Circuits include both popular crosspoint arrays and other more advanced structures, such as closed-loop memory arrays and ternary content-addressable memory. The same circuit might serve completely different applications, e.g., a crosspoint array can be used for accelerating matrix-vector multiplication for forward propagation in a neural network and outer product for backpropagation training. The different algorithms and memory properties to enable such diversification of circuit functions will be discussed. Finally, the main challenges and opportunities for IMC will be presented.

Memristive tonotopic mapping with volatile resistive switching memory devices

“…Major research efforts have gone in this direction, where the memristive element is used as a static first-order memory element to store the synaptic weight of an artificial neural network 39 . In this framework, memristors can act as accelerators of hardware neural networks within the context of in-memory computing 40 , 41 . Other research directions aim at the exploration of biologically plausible paradigms using memristors as synaptic dynamic elements.…”

A review of in-memory computing for machine learning: architectures, options

“…Dynamic IMC embraces the advantage of static IMC and enables the additional strength in switching memory devices to reproduce additional functions, such as neuron activation. The general case of applying IMC is to combine them into the same platform and provide energy-efficient computing systems that enable learning and adaptation, with significant challenges on device structures and reliability (Mannocci et al , 2023).…”

An organized view of reservoir computing: a perspective on theory and technology development

“…Therefore, various approaches, in which training is simplified or restricted to only part of the system, have been considered for years. [1][2][3] Furthermore, development of in-memory computing approaches, 4,5) the free von Neumann bottleneck, 6,7) requires completely new computing paradigms, different from commonly used Turingian algorithms. [8][9][10][11] There are two main issues limiting computing efficiency: the von Neumann bottleneck and the informational "black hole" problem.…”