Exploring quantum mechanical advantage for reservoir computing

TL;DR

This paper investigates how quantum properties like entanglement influence the memory capacity of quantum reservoir computing systems, highlighting the importance of quantum dynamics for enhanced machine learning performance.

Contribution

It establishes a link between quantum entanglement, phase space dimension, and memory performance, providing quantitative insights into quantum reservoir advantages.

Findings

High entanglement correlates with increased reservoir complexity.

Quantum phase space dimension impacts short-term memory capacity.

Dephasing reduces quantum reservoir performance.

Abstract

Quantum reservoir computing is an emerging field in machine learning with quantum systems. While classical reservoir computing has proven to be a capable concept of enabling machine learning on real, complex dynamical systems with many degrees of freedom, the advantage of its quantum analogue is yet to be fully explored. Here, we establish a link between quantum properties of a quantum reservoir, namely entanglement and its occupied phase space dimension, and its linear short-term memory performance. We find that a high degree of entanglement in the reservoir is a prerequisite for a more complex reservoir dynamics that is key to unlocking the exponential phase space and higher short-term memory capacity. We quantify these relations and discuss the effect of dephasing in the performance of physical quantum reservoirs.