2026-10-26 · arXiv Quantum Physics
arXiv:2610.32156v1 Announce Type: new
Abstract: Predicting and controlling driven quantum systems requires access to dynamics across physical parameters, initial states, and evolution times. A reusable neural model must capture this dependence while preserving unitary structure and maintaining accuracy over long-time coherent evolution. Here we introduce Quantum Evolution Networks (QENs), which map physical parameters and time to complete propagators for periodically driven three-spin registers. Shared networks learn unitary one-period propagators and compatible within-period paths, enabling predictions through known periodic composition without query-specific numerical integration. For transverse driving, the fifth-percentile fidelity, minimized over input states and sampled times, is 0.9947 over 24 periods on 256 held-out configurations. A leading-order analysis of spectrally resolved endpoint residuals predicts multicycle process error more accurately than quadratic or spectrum-averaged estimates. Interventions at matched single-period infidelity produce orders-of-magnitude differences in multicycle infidelity, demonstrating the importance of residual orientation. Compatible component replacements identify how endpoint and within-period approximations limit reconstruction accuracy. These results establish a reusable neural representation of the studied propagator families and connect its learned residual structure to its observed long-time reliability.
原文链接：https://arxiv.org/abs/2610.32156
查看更多资讯请见 AI 行业资讯。