Search

Systems that emulate biological neural networks offer an efficient way of running AI algorithms, but they can’t be trained using the conventional approach. The symmetry of these ‘physical’ networks provides a neat solution.

Neural networks often forget old knowledge or become too rigid when learning new data. Here, authors introduce Metaplasticity from Synaptic Uncertainty, a Bayesian learning rule that scales learning by uncertainty and forgets in a controlled way, enabling robust continual learning and reliable detection of unknown inputs.

A memory technology that combines the functions of memristors and ferroelectric capacitors in a single stack can be used for on-chip training and inference of artificial neural networks.

On International Women in Engineering Day, members of our editorial board highlight individuals who have inspired them during their research careers.

Memristors have been intensely investigated for memory applications and neural network accelerators. In this Comment, we discuss the requirements for how memristor technologies should evolve for Bayesian in-memory computing.

By combining several probabilistic AI algorithms, a recent study demonstrates experimentally that the inherent noise and variation in memristor nanodevices can be exploited as features for energy-efficient on-chip learning.

Bayesian electronics harness the randomness of noisy sensor data to quantify uncertainty and make predictions at low computational cost. This Perspective shows how they can be realized to improve reliability and reduce energy in wearable devices, smart industrial sensors and autonomous robots

The Editorial Board and Editorial Team are delighted to present a selection of short Research Highlights describing some of our favourite Communications Engineering publications of 2023.

The deployment of AI on edge computing devices raises significant challenges in terms of high energy consumption and limited functionality, which could be efficiently mitigated by brain-inspired neuromorphic engineering. Here, the authors introduce voltage-dependent synaptic plasticity for efficient, unsupervised learning in memristive synapses, demonstrating state-of-the-art performance and robustness on pattern recognition tasks

The authors present an AI engine with 32,768 memristors powered by a miniature solar cell. This circuit exploits near-memory computing, naturally adjusting its accuracy depending on the illumination level, and paves the way for self-powered AI.

Deep neural networks usually rapidly forget the previously learned tasks while training new ones. Laborieux et al. propose a method for training binarized neural networks inspired by neuronal metaplasticity that allows to avoid catastrophic forgetting and is relevant for neuromorphic applications.

A Bayesian machine can be implemented in a system with distributed memristors, allowing it to locally perform computation with minimal energy movement.

Spin-torque nano-oscillators have sparked interest for their potential in neuromorphic computing, however concrete demonstration are limited. Here, Romera et al show how spin-torque nano-oscillators can mutually synchronise and recognize temporal patterns, much like neurons, illustrating their potential for neuromorphic computing.

Deep learning has an increasing impact to assist research. Here, authors show that a dynamical neural network, trained on a minimal amount of data, can predict the behaviour of spintronic devices with high accuracy and an extremely efficient simulation time.

Clement Turck and colleagues present an alternative computing platform leveraging the property of logarithm to transfer multiplication operation into addition. They demonstrate the energy efficiency and superior performance of the prototype on gesture recognition and sleep stage recognition benchmark tasks.

Bayesian networks gain importance in safety-critical applications. Authors conducted experiments with a memristor-based Bayesian network trained with variational inference with technological loss, achieving accurate heartbeats classification and prediction certainty.

The non-ideal characteristics of resistive memory devices can be used to develop low-power and resilient probabilistic neuromorphic computing hardware, suitable for highly constrained edge-based applications.

Mona Ezzadeen and co-authors demonstrate a compute-in memory cell with a low consumed power per operation. In silicon implementation with 23 inputs is successfully used to solve benchmarking tasks of digit recognition.

A network of four spin-torque nano-oscillators can be trained in real time to recognize spoken vowels, in a simple and scalable approach that could be exploited for large-scale neural networks.

Spintronic nano-neurons and synapses can be connected by radiofrequency signals into neural networks that are capable of classifying real-world radiofrequency inputs without digitization at high speed and with low energy costs—an important step for artificial intelligence at the edge.

Population coding, where populations of artificial neurons process information collectively can facilitate robust data processing, but require high circuit overheads. Here, the authors realize this approach with reduced circuit area and power consumption, by utilizing superparamagnetic tunnel junction based neurons.

Neuromorphic computing takes inspiration from the brain to create energy-efficient hardware for information processing, capable of highly sophisticated tasks. Including more physics in the algorithms and nanoscale materials used for computing could have a major impact in this field.

Spoken-digit recognition using a nanoscale spintronic oscillator that mimics the behaviour of neurons demonstrates the potential of such oscillators for realizing large-scale neural networks in future hardware.