Spiking Neural Networks for Inference and Learning: A Memristor-based Design Perspective

TL;DR

This paper reviews recent advances in implementing spiking neural networks with memristor-based hardware, emphasizing how interdisciplinary plasticity rules can enhance neuromorphic computing's density, power efficiency, and learning capabilities.

Contribution

It introduces novel multifactor plasticity rules tailored to memristor dynamics, surpassing traditional STDP in performance and enabling more effective neuromorphic hardware implementations.

Findings

Memristor-based neuromorphic systems can outperform traditional architectures in efficiency.

Interdisciplinary plasticity rules leverage memristor stochasticity for better learning.

Recent developments show promising hardware implementations of spiking neural networks.

Abstract

On metrics of density and power efficiency, neuromorphic technologies have the potential to surpass mainstream computing technologies in tasks where real-time functionality, adaptability, and autonomy are essential. While algorithmic advances in neuromorphic computing are proceeding successfully, the potential of memristors to improve neuromorphic computing have not yet born fruit, primarily because they are often used as a drop-in replacement to conventional memory. However, interdisciplinary approaches anchored in machine learning theory suggest that multifactor plasticity rules matching neural and synaptic dynamics to the device capabilities can take better advantage of memristor dynamics and its stochasticity. Furthermore, such plasticity rules generally show much higher performance than that of classical Spike Time Dependent Plasticity (STDP) rules. This chapter reviews the recent development in learning with spiking neural network models and their possible implementation with memristor-based hardware.