Reproduce “Quantum Reservoir Computing with Physics-Informed Correction for Reduced-Order PDE Forecasting” on OpenResearch

About this paper

Krishna Bhatia, Harsh, Shalini Devendrababu

This study from Fractal Analytics and IIT Jodhpur introduces a two-stage hybrid architecture, combining a Quantum Reservoir Computer (QRC) with a Physics-Informed Neural Network (PINN)-based corrector, for reduced-order forecasting of nonlinear partial differential equations. The QRC+PIC model improves predictive accuracy and physical consistency for chaotic systems like the Kuramoto–Sivashinsky equation by locally refining the QRC's initial predictions.

Key insights

• The hybrid proposal-correction architecture effectively leverages the QRC's efficient nonlinear mapping capabilities for initial predictions and the PINN's ability to enforce physical consistency locally.
• Local physics-informed correction mitigates the fragility of global PINN optimization and enhances the physical adherence of purely data-driven QRC forecasts.
• Rigorous benchmarking, including simple classical autoregressive models, is crucial as linear baselines can remain highly competitive in reduced-order chaotic and non-chaotic regimes.

Summary by alphaXiv.