An atomic Boltzmann machine capable of on-chip learning

Abstract

The Boltzmann Machine (BM) is a neural network composed of stochastically firing neurons that can learn complex probability distributions by adapting the synaptic interactions between the neurons. BMs represent a very generic class of stochastic neural networks that can be used for data clustering, generative modelling and deep learning. A key drawback of software-based stochastic neural networks is the required Monte Carlo sampling, which scales intractably with the number of neurons. Here, we realize a physical implementation of a BM directly in the stochastic spin dynamics of a gated ensemble of coupled cobalt atoms on the surface of semiconducting black phosphorus. Implementing the concept of orbital memory utilizing scanning tunnelling microscopy, we demonstrate the bottom-up construction of atomic ensembles whose stochastic current noise is defined by a reconfigurable multi-well energy landscape. Exploiting the anisotropic behaviour of black phosphorus, we build ensembles of atoms with two well-separated intrinsic time scales that represent neurons and synapses. By characterizing the conditional steady-state distribution of the neurons for given synaptic configurations, we illustrate that an ensemble can represent many distinct probability distributions. By probing the intrinsic synaptic dynamics, we reveal an autonomous reorganization of the synapses in response to external electrical stimuli. This self-adaptive architecture paves the way for on-chip learning directly in atomic-scale machine learning hardware.

References in corpus (3)

• Training and Operation of an Integrated Neuromorphic Network Based on Metal-Oxide Memristors
• Efficient supervised learning in networks with binary synapses
• Adsorption of cobalt on graphene: Electron correlation effects from a quantum chemical perspective

Cited by in corpus (12)

• Hardware-aware $insitu$ Boltzmann machine learning using stochastic magnetic tunnel junctions
• Precise atom manipulation through deep reinforcement learning
• Atomic-Scale Visualization of Chiral Charge Density Wave States and Their Reversible Transition
• Anisotropic effects in two-dimensional materials
• CMOS-compatible Ising and Potts Annealing Using Single Photon Avalanche Diodes
• Screening nearest-neighbor interactions in networks of exciton-polariton condensates through spin orbit coupling
• Physical Neural Networks with Self-Learning Capabilities
• High-speed coherent photonic random-access memory in long-lasting sound waves
• Gating orbital memory with an atomic donor
• A self-learning magnetic Hopfield neural network with intrinsic gradient descent adaption
• Gradient descent in materia through homodyne gradient extraction
• Artificial stochastic neural network on the base of double quantum wells