Title: Establishing Risk Limits in Quantum Reservoir Computing

Abstract: This study introduces a methodology for determining generalization error bounds for various types of quantum reservoirs, leveraging Rademacher complexity. We derive specific, parameter-dependent limits for two distinct categories of quantum reservoirs and examine how these generalization bounds evolve as the qubit count increases. When applied to reservoir classes utilizing polynomial readout functions, our analysis demonstrates that risk bounds converge relative to the volume of training samples. While the explicit relationship between the bounds and the parameters of the quantum reservoir and readout mechanism allows for a degree of control over generalization error, it is important to note that the bounds exhibit exponential scaling with respect to the number of qubits, $n$. Furthermore, the derived upper limits on Rademacher complexity are applicable to additional reservoir classes, provided they satisfy specific assumptions regarding quantum dynamics and the readout function.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC