Avalanches and edge-of-chaos learning in neuromorphic nanowire networks

Joel Hochstetter,Alon Loeffler,Adrian Diaz-Alvarez,Tomonobu Nakayama,Ruomin Zhu,Zdenka Kuncic

doi:10.1038/s41467-021-24260-z

Abstract

The brain’s efficient information processing is enabled by the interplay between its neuro-synaptic elements and complex network structure. This work reports on the neuromorphic dynamics of nanowire networks (NWNs), a unique brain-inspired system with synapse-like memristive junctions embedded within a recurrent neural network-like structure. Simulation and experiment elucidate how collective memristive switching gives rise to long-range transport pathways, drastically altering the network’s global state via a discontinuous phase transition. The spatio-temporal properties of switching dynamics are found to be consistent with avalanches displaying power-law size and life-time distributions, with exponents obeying the crackling noise relationship, thus satisfying criteria for criticality, as observed in cortical neuronal cultures. Furthermore, NWNs adaptively respond to time varying stimuli, exhibiting diverse dynamics tunable from order to chaos. Dynamical states at the edge-of-chaos are found to optimise information processing for increasingly complex learning tasks. Overall, these results reveal a rich repertoire of emergent, collective neural-like dynamics in NWNs, thus demonstrating the potential for a neuromorphic advantage in information processing.

Highlights

The brain’s efficient information processing is enabled by the interplay between its neurosynaptic elements and complex network structure
The increasing prevalence of streaming data requires a shift in neuro-inspired information processing paradigms, beyond static Artificial Neural Network (ANN) models used in Artificial Intelligence (AI)
It has been widely postulated that optimal information processing in non-linear dynamical systems may be achieved close to a phase transition, in a state known as criticality[29]

Summary

Introduction

The brain’s efficient information processing is enabled by the interplay between its neurosynaptic elements and complex network structure. The brain’s unique capacity for adaptive, real-time learning is enabled by the complex interplay between its neuro-synaptic non-linear elements and a recurrent network topology[1], with information processing manifested through emergent collective dynamics[2]. The interplay between memristive switching and a recurrent network topology promotes emergence of collective, adaptive dynamics, such as formation of conducting transport pathways that can be dynamically tuned[18,19], giving networks a biologically plausible structural plasticity[20]. These neuromorphic properties equip NWNs with unique learning potential, with applications ranging from shortest-path optimisation[21] to associative memory[22]. Edge-of-chaos dynamics have been observed in cortical networks[35] and appear to optimise computational performance in recurrent neural networks[36], echo state networks[37] and random boolean networks[38]

Methods

Results

Conclusion