Perovskite neural trees

Haitian Zhang ,Martin V Holt,Tae Joon Park,Evgeny Nazaretski,Badri Narayanan,M Krishnamurthy ,Oleg Shpyrko ,Subramanian K R S Sankaranarayanan ,Shakti N Wadekar ,Alex Frañó ,Chih-Yung Huang ,Yong S Chu,Mathew J Cherukara,Mingyuan Ge,Zhan Zhang,Robert Andrawis,Peter O Sprau,Sukriti Manna,Qi Wang,Gopalakrishnan Srinivasan,Kaushik Roy,Ivan A Zaluzhnyy,Zhen Zhang,Nelson Hua,Yifei Sun,Hua Zhou,Xiaojing Huang,Hanfei Yan,Shriram Ramanathan

doi:10.1038/s41467-020-16105-y

Abstract

Trees are used by animals, humans and machines to classify information and make decisions. Natural tree structures displayed by synapses of the brain involves potentiation and depression capable of branching and is essential for survival and learning. Demonstration of such features in synthetic matter is challenging due to the need to host a complex energy landscape capable of learning, memory and electrical interrogation. We report experimental realization of tree-like conductance states at room temperature in strongly correlated perovskite nickelates by modulating proton distribution under high speed electric pulses. This demonstration represents physical realization of ultrametric trees, a concept from number theory applied to the study of spin glasses in physics that inspired early neural network theory dating almost forty years ago. We apply the tree-like memory features in spiking neural networks to demonstrate high fidelity object recognition, and in future can open new directions for neuromorphic computing and artificial intelligence.

Highlights

Trees are used by animals, humans and machines to classify information and make decisions
We show that tree-like structures can be experimentally realized at room temperature in strongly correlated rare-earth perovskite nickelates (ReNiO3, where Re is a rare-earth ion), a class of quantum materials whose electrical properties are largely dominated by the strong interactions among electrons in them[28,29,30]
The tree-structured algorithmic weight changes as a function of time steps are used for simulation in a spiking neural network (SNN) as synapses between the input layer and the excitatory layer to demonstrate proof-of-concept application in learning using MNIST digits, a widely used database by the neuromorphic engineering community

Summary

Introduction

Trees are used by animals, humans and machines to classify information and make decisions. We report experimental realization of tree-like conductance states at room temperature in strongly correlated perovskite nickelates by modulating proton distribution under high speed electric pulses. This demonstration represents physical realization of ultrametric trees, a concept from number theory applied to the study of spin glasses in physics that inspired early neural network theory dating almost forty years ago. Evolution has enabled synapses (that are responsible for memory) to reach selflimiting weights as well as branching to avoid runaway effects and prevent catastrophic breakdown of neural circuits while still retaining the ability to learn throughout their lifespan This remarkable synaptic weight update mechanism can be summarized as a tree structure. Proton migration-driven organic systems have been reported as artificial synapses suggesting their broad relevance in emerging memory devices[33]

Methods

Results

Conclusion