-
经典计算正逼近物理极限, 而含噪声的中等规模量子硬件面临校准与控制瓶颈, 这催化了量子人工智能这一前沿交叉领域的诞生. 量子人工智能旨在利用人工智能与量子计算的协同作用, 解决彼此的瓶颈问题. 本综述系统地分析量子人工智能的双向赋能框架. 在“AI for Quantum”方面, AI正成为克服中等规模量子硬件瓶颈的工具, 关键应用包括设备自动化校准与高保真读出、开发超越传统算法的AI量子纠错解码器及AI驱动的量子线路优化. 在“Quantum for AI”方面, 量子计算正为AI算法提供新的计算范式与模型设计思路, 从早期的HHL, QSVM算法演进到中等规模量子时代主流的变分量子算法和量子神经网络, 同时综述了该领域面临的贫瘠高原等可训练性挑战及其缓解策略, 并涵盖了量子优化及量子原生智能模型. 总结指出, 量子人工智能的双向融合有望成为推动量子计算从中等规模量子时代迈向容错计算、构建下一代混合量子-经典智能系统的重要路径之一.As classical computing approaches its physical limits, quantum computing offers exponential computational advantages, yet the current noisy intermediate-scale quantum (NISQ) era is severely constrained by high error rates, decoherence, and control challenges. This has catalyzed quantum artificial intelligence (QAI), a synergistic and bidirectional interdisciplinary field. This review analyzes the “AI for Quantum” framework, where AI addresses quantum bottlenecks, and the “Quantum for AI” framework, where quantum computing enables novel computational paradigms for artificial intelligence.The “AI for Quantum” framework elaborates the role of AI in addressing hardware bottlenecks. Specifically, this encompasses machine learning for autonomous device characterization, calibration, and high-fidelity readout, along with the development of advanced AI-based decoders and hardware-aware quantum error correction codes, as well as the optimization of quantum compilation.The “Quantum for AI” framework surveys the progression from early algorithms such as HHL and QSVM to currently prevalent variational quantum algorithms and quantum neural networks. The primary obstacles are critically examined, including the barren plateau phenomenon and the exponential concentration of quantum kernels, together with their mitigation strategies. The review also covers advances in quantum optimization, such as quantum annealing and the quantum approximate optimization algorithm, and the emergence of advanced models like quantum natural language processing.This bidirectional fusion represents a pivotal strategy for facilitating the transition from the NISQ era toward fault-tolerant computation and the development of next-generation, hybrid quantum-classical intelligent systems.
搜索
x
中国物理学会期刊
下载: