Two-Terminal Neuromorphic Devices for Spiking Neural Networks: Neurons, Synapses, and Array Integration

Area of Science:

• Neuromorphic Engineering
• Materials Science
• Computational Neuroscience

Background:

• Conventional computing struggles with massive, complex datasets.
• Neuromorphic computing and artificial neural networks (ANNs) offer potential solutions.
• Spiking neural networks (SNNs) are a promising third-generation ANN for real-time spatiotemporal processing.

Purpose of the Study:

• To review progress in memristor-based spiking neural network (SNN) neuromorphic hardware.
• To highlight the role of memristor-based neurons, synapses, and array integration.
• To provide perspectives for developing next-generation brain-like computing systems.

Main Methods:

• Focus on memristor devices for artificial neurons and synapses.
• Investigate the integration of these components into arrays.
• Incorporate relevant biological insights into hardware design.

Main Results:

• Memristor-based devices are crucial building blocks for SNN hardware.
• Overcoming material and integration challenges is key for biomimetic behavior.
• Advancements are being made in creating high-density crossbar arrays for multiply-accumulate (MAC) operations.

Conclusions:

• Memristor-based SNNs represent a significant step towards brain-like computing.
• Material properties and device integration are critical for effective SNN hardware.
• Further research is needed to fully realize the potential of these systems.