Aug 2026· Quantum Science and Technology· Vol 11· 0 citations· 64 references
Physics
TL;DR
This work proposes a non-Hermitian QRC in which non-Hermitian dynamics are employed as a tunable resource to significantly enhance the QRC performance, and demonstrates that the non-Hermitian reservoir can be tuned toward the edge of chaos by varying a single parameter that controls the non-Hermitian strength.
Abstract
Quantum reservoir computing (QRC) offers a powerful approach to exploit the rich dynamics of quantum systems for information processing. However, the computational performance of conventional Hermitian reservoirs is inherently constrained by their nonlinearity and information-spreading ability. In this work, we propose a non-Hermitian QRC in which non-Hermitian dynamics are employed as a tunable resource to significantly enhance the QRC performance. By incorporating an imaginary interaction term into the one-dimensional XY spin model, the reservoir’s information propagation extends beyond the Lieb–Robinson bound, resulting in accelerated information scrambling. Through spectral analysis and memory evaluation, we demonstrate that the non-Hermitian reservoir can be tuned toward the edge of chaos by varying a single parameter that controls the non-Hermitian strength. This tuning optimizes both memory and computational capacities, which are crucial for processing temporal sequences. For applications, we evaluate the predictive performance of both classical and quantum chaotic time series. Our results demonstrate superior performance compared with the Hermitian counterpart, with particularly notable advantages in predicting signals generated by the Sachdev–Ye–Kitaev model.
Quantum reservoir computing (QRC) is a machine learning approach which employs the internal dynamics of a physical system (the reservoir) to encode and process information. In this work, we explore the use of dual-unitary circuits in a brickwork architecture as a platform for QRC, well suited to current noisy intermediate-scale quantum devices. Dual-unitary circuits present both practical and conceptual advantages. Our results indicate that, under appropriate conditions, dual-unitarity can lead to an enhanced regime of operation: we numerically verify that it improves memory effects and nonlinear processing, and shields against finite-shot noise, mitigating exponential concentration. Moreover, dual unitarity offers an intuitive picture of how operator dynamics gives rise to memory and nonlinear processing in circuit-based reservoirs.
Quantum reservoir computing processes temporal information through driven many-body dynamics, but its performance is ultimately limited by how accurately past inputs can be extracted from finite measurements. Here we formulate this limitation as a local multiparameter estimation problem and introduce a delay-space quantum Fisher information matrix to quantify the distinguishability of information stored at different delays. This perspective identifies Fisher-orthogonal memory as a measurement-efficient design principle: different delays should perturb the reservoir state along mutually Fisher-orthogonal directions. We first analyze the single-qubit limit using the Gill--Massar bound, revealing an optimal write-store-routing trade-off. Guided by this structure, we construct solvable multi-qubit reservoirs based on Clifford routing orbits and Singer-cycle Pauli algebra. The resulting dynamics yield diagonal delay-space QFIMs with analytically programmable fading profiles. Under finite-shot local Pauli readout, these reservoirs retain sharp memory windows that are absent in a validation-selected Ising baseline. Their product-task behavior is governed by second-order responses inherited from the same Pauli-routing algebra. Our results provide an analytically controlled route toward measurement-efficient quantum reservoir computing.
This work introduces a classical-quantum state derived from the process tensor representing the dynamical part of this process for typical QRC protocols found in the literature, and extracts two diagnostics that characterise the nonlocal scrambling of information within, and loss of information from the substrate.
Motivated by the perspective of advanced time-series prediction and exploitation of quantum reservoir computing (QRC), we explored the design and implementation of a hybrid photonic-QRC (HPQRC) paradigm. This brings together the high-speed parallelism of photonic systems with the quantum reservoir’s capacity to model complex, nonlinear dynamics, and hence acts as a powerful tool for performing prediction in a resource-constrained environment with low latency. We have engineered a solution using this architecture to address issues such as computational bottlenecks, energy inefficiency, and sensitivity to noise which are common in existing reservoir computing models. Our simulation results show that HPQRC consistently outperforms both classical and quantum-only reservoir models on chaotic, financial, and biomedical benchmarks: on Mackey–Glass and Lorenz systems, HPQRC reduces normalised mean squared error by 25.9% and 21.8% respectively over quantum-only RC; on MIT-BIH ECG R-peak prediction, HPQRC achieves 89.4% accuracy (within ±10 ms tolerance) compared to 81.3% for QRC; and on S&P 500 hourly direction prediction, HPQRC attains 55.13% mean directional accuracy versus 53.89% for QRC and 52.64% for Classical RC. The model retains above-chance mean predictive performance under 15% Gaussian noise (one-sample t(9)=5.54, p<0.001 versus 50% chance level). HPQRC also achieves a 56.1% reduction in per-prediction simulation wall-clock time relative to Classical RC, establishing it as a compelling simulation-validated paradigm for hybrid quantum-photonic reservoir computing.
ABSTRACT One of the most significant achievements of the second quantum revolution is enabling tasks that are infeasible for classical computers. However, the stringent requirements for quantum resources, along with the presence of noise and losses, impose limitations on technologies for quantum information processing. Here, we overcome these limitations by demonstrating robust and scalable quantum reservoirs for multiphoton quantum computing, the first of their kind to operate robustly at room temperature under high levels of noise and decoherence. By implementing Fock projective measurements, we extract multiphoton quantum systems from classical fields, enabling the engineering of quantum reservoirs that are robust to losses and noise. Our multiphoton reservoirs support universal quantum information processing with up to forty particles, exhibiting quantum speedups over classical counterparts. Remarkably, the multiparticle interactions hosted by our reservoirs enable us to perform quantum simulations of nonlinear systems. Specifically, we simulate the quantum thermodynamics of many‐body systems in synthetic lattices. Furthermore, we exploit the complexity of our multiphoton quantum reservoir, consisting of eight hundred sixty‐one components, for the quantum prediction of mathematical functions. As such, our work unveils a path toward the development of robust quantum technologies for information processing, long regarded as a central goal of the field.
M. Hong, M. Quiroz-Juárez, Riley B. Dawkins et al.· Advancement of science· 0 citations
Non-Hermitian dynamics can sharply amplify the response of a quantum sensor near eigenstate coalescence, but whether this amplification yields a genuine metrological gain depends on how the non-unitary evolution is physically realised and how its resources are counted. We address this question experimentally by implementing the pseudo-Hermitian qubit sensing protocol of Chu
et al
. [Phys. Rev. Lett. 124, 020501 (2020)] on a two-qubit nuclear magnetic resonance processor via Naimark dilation. Because both the postselected signal and the full two-qubit probability are obtained from a single set of population measurements, the enhanced response and its postselection cost are accessed within the same experiment. We observe the characteristic steep response of the postselected branch, whereas the sensitivity extracted from the absolute two-qubit probability does not surpass the Ramsey benchmark for the same sensor encoding, consistent with the theoretical resource bound derived by Ding
et al
. [Phys. Rev. Lett. 131, 160801 (2023)] for the corresponding dilation scheme. The experiment complements the recent photonic implementation of the direct non-unitary route and provides an experimental benchmark of the Hermitian-dilation route with explicit resource accounting.
Y. Zhai, Chao Wei, Jun-Da Song et al.· Chinese Physics B· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
