Memristive Devices and Networks for Brain‐Inspired Computing

Abstract

As the era of big data approaches, conventional digital computers face increasing difficulties in performance and power efficiency due to their von Neumann architecture. As a result, there is recently a tremendous upsurge of investigations on brain-inspired neuromorphic hardware with high parallelism and improved efficiency. Memristors are considered as promising building blocks for the realization of artificial synapses and neurons and can therefore be utilized to construct hardware neural networks. Here, a review is provided on existing approaches for the implementation of artificial synapses and neurons based on memristive devices; and the respective advantages and disadvantages of these approaches are evaluated. This is followed by a discussion of hardware accelerators and neuromorphic computing systems that exploit the parallel, in-memory and analog characteristics of memristive crossbar arrays as well as the intrinsic dynamics of memristors. Finally, the outstanding challenges are addressed that have not yet been resolved in the present studies, and future advances are discussed that might be needed for building intelligent and energy efficient neuromorphic systems.

Stimuli‐Enabled Artificial Synapses for Neuromorphic Perception: Progress and Perspectives

“…Phase-change material (PCM)-based devices are promising for use in next-generation nonvolatile memory device and neuromorphic synaptic device applications because they offer superfast switching speeds and low fabrication costs. [85][86][87][88][89] PCMs can show large differences in their electrical resistivity values when they are switched between their amorphous (low-conductance) and crystalline (high-conductance) phases. GeSbTe (GST) compounds are the most widely used PCMs because of their high cyclability.…”

Neuromorphic Engineering for Hardware Computational Acceleration and Biomimetic Perception Motion Integration

“…Although LIF neurons have successfully simulated the basic LIF function, they sacrificed and ignored the details of ion dynamics in biological neuron cell membranes. [ 151 ] As an alternative, HH neurons are used to describe certain complex characteristics of biological neurons, such as ion channel dynamics, all‐or‐none law and hyperpolarization phenomena, etc., [ 35 ] which seem to be more biologically authentic and actively adopted in neurocomputing science. [ 161 ] However, HH neurons involving sophisticated neuron dynamics usually bring more complex circuit structures, which is different from LIF neurons that can even be implemented in a single memristor.…”

“…The MLP contains perception, hidden connection, and reaction layers, which can solve the linear inseparable problem as an alternative to the SLP. [ 35 ] Both the connection layer and the reaction layer have information processing functions. The first layer implements binary separate classification, and the second layer implements AND operation, which solves the XOR problem.…”

Recent Progress in Synaptic Devices Based on 2D Materials

“…In nonfilamentary synaptic devices working by other mechanisms such as charge trapping or detrapping, higher switching uniformity could be obtained. However, the switching behaviors cannot produce a larger dynamic range, limiting the weight linearity and symmetry . Phase change materials could demonstrate inherent unipolar switching behaviors because of thermally driven switching mechanisms.…”