SymQNet: Amortized Acquisition for Low-Latency Adaptive Hamiltonian Learning

2026-06-10

AI Summary

This work addresses the excessive decision latency in adaptive Hamiltonian learning, which arises from the need to re-optimize Bayesian design criteria after every posterior update, rendering it impractical for high-frequency quantum calibration. To overcome this limitation, the authors propose SymQNet, the first framework to incorporate amortized reinforcement learning into this task. By learning a posterior-conditional acquisition policy offline, SymQNet enables online decision-making through millisecond-scale forward inference, thereby preserving Bayesian adaptivity while achieving ultra-low latency. Experimental results demonstrate that, in a 5-qubit system, SymQNet reduces decision latency by factors of 47.1 and 72.6 compared to bounded Fisher information search and two-step BALD, respectively. Moreover, full simulation on a 12-qubit system completes in just 1.02 seconds, substantially outperforming existing approaches.

Abstract

Adaptive Hamiltonian learning is central to calibrating and characterizing quantum devices. In an adaptive controller, choosing the next experiment is itself a computation. Bayesian design rules are recomputed after every posterior update, and that step can take seconds. Across hundreds of shots, those seconds become a significant wall-clock cost for adaptivity. We introduce SymQNet, an amortized reinforcement-learning approach for low-latency adaptive Hamiltonian learning. SymQNet learns a posterior-conditioned acquisition policy offline, then uses a fast policy forward pass online while retaining Bayesian posterior feedback. On transverse-field Ising benchmarks, SymQNet substantially reduces acquisition latency relative to bounded Fisher-information search and bounded two-step Bayesian active learning by disagreement (BALD). At five qubits, it reduces acquisition-only decision latency by $47.1\times$ and $72.6\times$ relative to these online baselines; at twelve qubits, full simulated steps take $1.02$ s for SymQNet versus $13.27$ s for bounded two-step BALD. Overall, we show that learned acquisition can make adaptive Hamiltonian learning practical for repeated low-latency workloads.

Problem

• adaptive Hamiltonian learning
• acquisition latency
• Bayesian experimental design
• quantum device calibration
• low-latency control

Innovation

• amortized reinforcement learning
• adaptive Hamiltonian learning
• low-latency acquisition
• Bayesian experimental design
• SymQNet