基于量子机器学习的无线宽带信号检测方案

Wireless Broadband Signal Detection Scheme Based on Quantum Machine Learning

无线通信发展到B5G/6G,随着太赫兹通信等技术的诞生,信号的调制阶数、信道信息的复杂度增加,接收信号检测的难度也是不断上升。现大多研究都将无线通信领域与机器学习相结合,但经典机器学习的方案很可能会遇到算力瓶颈。量子机器学习作为量子计算与机器学习的交叉学科,有可能成为6G的潜在技术。本文首次将量子机器学习应用于无线宽带通信中的信号检测,基于张量网络设计了一种量子-经典混合机器学习模型,混合模型由含参量子电路和神经网络构成,由量子计算机和经典计算机协同实现,信号数据在量子电路里被映射至希尔伯特空间进行训练,有着快速收敛,代价低,性能高等优势,神经网络对量子电路输出结果起到非线性修正的效果。本文仿真在Google公司的Tensorflow-Quantum平台上进行,结果表明,本文提出的量子-经典混合机器学习模型在信号检测问题中,以误码率为量化指标,与现有的纯量子算法和神经网络算法相比降低大约10个百分点。

Wireless communication develops to B5G/6G.With the birth of terahertz communication and other tech⁃nologies,the modulation order of signals and the complexity of channel information increase,and the difficulty of receiving signal detection is also rising.At present,most researches combine wireless communication field with machine learning,but classical machine learning schemes are likely to encounter computational bottlenecks.Quan⁃tum machine learning,as a cross-discipline between quantum computing and machine learning,may become a po⁃tential technology of 6G.This paper applies quantum machine learning to signal detection in wireless broadband communication for the first time,and designs a quantum-classical hybrid machine learning model based on tensor network.The hybrid model is composed of quantum circuits with parameters and neural networks.The signal data is mapped to the Hilbert space in the quantum circuit for training.It has the advantages of fast convergence,low cost and high performance.The neural network can play a nonlinear correction effect on the output result of the quantum circuit.The simulation in this paper is carried out on Google’s Tensorflow-Quantum platform.The results show that the quantum-classical hybrid machine learning model proposed in this paper takes the bit error rate as the quantitative index in signal detection problems,which is about 10 percentage points lower than the existing pure quantum algorithm and neural network algorithm.