Stochastic Reservoir Computers

Peter J. Ehlers,Hendra I. Nurdin,Daniel Soh

2024-05-21

Abstract:Reservoir computing is a form of machine learning that utilizes nonlinear dynamical systems to perform complex tasks in a cost-effective manner when compared to typical neural networks. Many recent advancements in reservoir computing, in particular quantum reservoir computing, make use of reservoirs that are inherently stochastic. However, the theoretical justification for using these systems has not yet been well established. In this paper, we investigate the universality of stochastic reservoir computers, in which we use a stochastic system for reservoir computing using the probabilities of each reservoir state as the readout instead of the states themselves. In stochastic reservoir computing, the number of distinct states of the entire reservoir computer can potentially scale exponentially with the size of the reservoir hardware, offering the advantage of compact device size. We prove that classes of stochastic echo state networks, and therefore the class of all stochastic reservoir computers, are universal approximating classes. We also investigate the performance of two practical examples of stochastic reservoir computers in classification and chaotic time series prediction. While shot noise is a limiting factor in the performance of stochastic reservoir computing, we show significantly improved performance compared to a deterministic reservoir computer with similar hardware in cases where the effects of noise are small.

Machine Learning,Neural and Evolutionary Computing,Systems and Control,Adaptation and Self-Organizing Systems

What problem does this paper attempt to address?

This paper discusses the theoretical basis and potential advantages of Stochastic Reservoir Computers. Traditional reservoir computing utilizes nonlinear dynamic systems to efficiently perform complex tasks, while recent advancements, particularly in quantum reservoir computing, introduce inherent randomness. However, a sufficient theoretical foundation for using these stochastic systems has not yet been established.

The main contribution of this paper is to demonstrate that Stochastic Reservoir Computers, including Stochastic Echo State Networks (SESNs), are a class of universal approximators. This means that these systems are capable of approximating any task we want to solve with arbitrary precision, which is a key attribute for achieving broad applications. The paper presents mathematical proofs to show the universality of Stochastic Echo State Networks and analyzes two practical application examples - classification and chaotic time series prediction - demonstrating that Stochastic Reservoir Computers can exhibit significant performance improvements compared to deterministic reservoir computers in scenarios with minimal noise.

One key advantage of Stochastic Reservoir Computers is that the number of computing nodes can exponentially increase with the hardware scale, providing a large computational capacity while maintaining a compact device. For example, in quantum computers based on qubits, input can be utilized to perform unitary rotations, and then the results are measured to obtain the probabilities of different outcomes as readout.

The paper also introduces two different designs of Stochastic Reservoir Computers: one based on qubits and the other based on optical setups. Both designs demonstrate enhanced performance compared to deterministic computers in limited statistical cases, but require a balance between noise influence and the number of runs.

In conclusion, this paper provides a solid theoretical foundation for Stochastic Reservoir Computers and explores their potential in practical applications, especially in handling complex time series and classification tasks.