Optimal training of finitely sampled quantum reservoir computers for forecasting of chaotic dynamics

• DOI: 10.1007/s42484-025-00261-9
• Publication Date: 2025-02-27T07:15:29Z
• Authors: Osama Ahmed; Felix Tennie; Luca Magri
• Keywords: Reservoir computing

Abstract In the current Noisy Intermediate Scale Quantum (NISQ) era, presence of noise deteriorates performance quantum computing algorithms. reservoir (QRC) is a type machine learning algorithm, which, however, can benefit from different types tuned noise. this paper, we analyze how finite sampling affects chaotic time series prediction gate-based QRC and recurrence-free (RF-QRC) models. First, examine RF-QRC show that, even without recurrent loop, it contains temporal information about previous states using leaky integrated neurons. This makes extreme machines (QELM). Second, that degrades capabilities both while affecting more due to propagation Third, optimize training finite-sampled framework two methods: (a) singular value decomposition (SVD) applied data matrix containing noisy activation (b) data-filtering techniques remove high frequencies states. We denoising improves signal-to-noise ratios with smaller loss. Finally, demonstrate signals in are highly parallelizable on multiple processing units (QPUs) as compared architecture connections. The analyses numerically showcased prototypical dynamical systems relevance turbulence. work opens opportunities for samples forecasting near-term hardware.