The reservoir computing (RC) framework states that any nonlinear, input-driven dynamical system (the reservoir ) exhibiting properties such as a fading memory and input separability can be trained to perform computational tasks. This broad inclusion of systems has led to many new physical substrates for RC. Properties essential for reservoirs to compute are tuned through reconfiguration of the substrate, such as change in virtual topology or physical morphology. As a result, each substrate possesses a unique ‘quality’—obtained through reconfiguration—to realize different reservoirs for different tasks. Here we describe an experimental framework to characterize the quality of potentially any substrate for RC. Our framework reveals that a definition of quality is not only useful to compare substrates, but can help map the non-trivial relationship between properties and task performance. In the wider context, the framework offers a greater understanding as to what makes a dynamical system compute, helping improve the design of future substrates for RC.

Neural networks have revolutionized the area of artificial intelligence and introduced transformative applications to almost every scientific field and industry. However, this success comes at a great price; the energy requirements for training advanced models are unsustainable. One promising way to address this pressing issue is by developing low-energy neuromorphic hardware that directly supports the algorithm's requirements. The intrinsic non-volatility, non-linearity, and memory of spintronic devices make them appealing candidates for neuromorphic devices. Here, we focus on the reservoir computing paradigm, a recurrent network with a simple training algorithm suitable for computation with spintronic devices since they can provide the properties of non-linearity and memory. We review technologies and methods for developing neuromorphic spintronic devices and conclude with critical open issues to address before such devices become widely used.

Deep learning methods are recognised as state-of-the-art for many applications of machine learning. Recently, deep learning methods have emerged as a solution to the task of automatic music generation (AMG) using symbolic tokens in a target style, but their superiority over non-deep learning methods has not been demonstrated. Here, we conduct a listening study to comparatively evaluate several music generation systems along six musical dimensions: stylistic success, aesthetic pleasure, repetition or self-reference, melody, harmony, and rhythm. A range of models, both deep learning algorithms and other methods, are used to generate 30-s excerpts in the style of Classical string quartets and classical piano improvisations. Fifty participants with relatively high musical knowledge rate unlabelled samples of computer-generated and human-composed excerpts for the six musical dimensions. We use non-parametric Bayesian hypothesis testing to interpret the results, allowing the possibility of finding meaningfulnon-differences between systems’ performance. We find that the strongest deep learning method, a reimplemented version of Music Transformer, has equivalent performance to a non-deep learning method, MAIA Markov, demonstrating that to date, deep learning does not outperform other methods for AMG. We also find there still remains a significant gap between any algorithmic method and human-composed excerpts.

This tutorial covers physical reservoir computing from a computer science perspective. It first defines what it means for a physical system to compute, rather than merely evolve under the laws of physics. It describes the underlying computational model, the Echo State Network (ESN), and also some variants designed to make physical implementation easier. It explains why the ESN model is particularly suitable for direct physical implementation. It then discusses the issues around choosing a suitable material substrate, and interfacing the inputs and outputs. It describes how to characterise a physical reservoir in terms of benchmark tasks, and task-independent measures. It covers optimising configuration parameters, exploring the space of potential configurations, and simulating the physical reservoir. It ends with a look at the future of physical reservoir computing as devices get more powerful, and are integrated into larger systems.