计算机科学 ›› 2024, Vol. 51 ›› Issue (11A): 231200112-7.doi: 10.11896/jsjkx.231200112
陈超, 闫文杰, 薛桂香
CHEN Chao, YAN Wenjie, XUE Guixiang
摘要： 量子神经网络结合了量子计算与经典神经网络模型的优势,为未来人工智能领域的发展提供了一种全新的思路。尽管量子神经网络已被广泛研究,但数据编码方式以及不同训练线路对模型性能的影响尚未得到充分研究。为此,提出一种面向数据分类的量子神经网络新模型,该模型探究了不同数据编码方式和不同结构训练层对分类任务的影响。该方法首先对经典图像进行预处理,采用不同的数据编码方式将其编码到不同的参数化量子线路中进行训练,对模型的输出进行测量,使用参数移位规则更新训练参数完成数据分类。在MNIST手写体数据集上的实验结果表明,所提出的模型在数字{3,6}分类任务上的分类准确率超过了97%。与目前的主流方法相比,所提方法在分类准确率上有明显的提升。
中图分类号:
| [1]LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436-444. [2]XU X,DING Y,HU S X,et al.Scaling for edge inference ofdeep neural networks[J].Nature Electronics,2018,1(4):216-222. [3]BRAVYI S,DIAL O,GAMBETTA J M,et al.The future ofquantum computing with superconducting qubits[J].Journal of Applied Physics,2022,132(16). [4]BEER K,BONDARENKO D,FARRELLY T,et al.Trainingdeep quantum neural networks[J].Nature Communications,2020,11(1):808. [5]JIE G J,ZHUANG Z Q.Quantum Neural Networks[J].Computer Science,2001(7):1-6. [6]ZHANG L F,ZHANG X P.Network Traffic Prediction based on BP Neural networkoptimized by Quantum Genetic Algorithm[J].Computer Engineering and Science,2016,38(1):114-119. [7]HUANG H Y,BROUGHTON M,COTLER J,et al.Quantum advantage in learning from experiments[J].Science,2022,376(6598):1182-1186. [8]SIM S,JOHNSON P D,ASPURU-GUZIK A.Expressibility and entangling capability of parameterized quantum circuits for hybrid quantum-classical algorithms[J].Advanced Quantum Technologies,2019,2(12):1900070. [9]ZHENG J,GAO Q,LV Y X,et al.Model and application ofquantum Convolutional Neural Network Based on Parameterized quantum Circuit[J].Control Theory & Applications,2021,38(11):1772-1784. [10]PRESKILL J.Quantum computing in the NISQ era and beyond[J].Quantum,2018,2:79. [11]WANG H,LI Z,GU J,et al.QOC:quantum on-chip trainingwith parameter shift and gradient pruning[C]//Proceedings of the 59th ACM/IEEE Design Automation Conference.2022:655-660. [12]CHEN K,ALBEVERIO S,FEI S M.Entanglement of formationof bipartite quantum states[J].Physical Review Letters,2005,95(21):210501. [13]HUANG C H.Analysis and discussion on superposition princi-ple of quantum states and its measurement[J].University Phy-sics,2013,32(4):22-24,27. [14]CROOKS G E.Gradients of parameterized quantum gates using the parameter-shift rule and gate decomposition[J].arXiv:1905.13311,2019. [15]QIAN C.Quantum Entanglement and Quantum Computation[J].Computer Science,2006(12):230-234. [16]CORTESE J,BRAJE T.System and technique for loading classical data into a quantum computer:U.S.Patent 11,113,621[P].2021-09-07. [17]GANDHI V,PRASAD G,COYLE D,et al.Quantum neuralnetwork-based EEG filtering for a brain-computer interface[J].IEEE Transactions on Neural Networks and Learning Systems,2013,25(2):278-288. [18]CHAPELLE O,WU M.Gradient descent optimization ofsmoothed information retrieval metrics[J].Information Retrieval,2010,13:216-235. [19]KOBAYASHI T.Large Margin In Softmax Cross-Entropy Loss[C]//BMVC.2019:3. [20]SHAIK E H,RANGASWAMY N.Implementation of quantumgates based logic circuits using IBM Qiskit[C]//2020 5th International Conference on Computing,Communication and Security(ICCCS).IEEE,2020:1-6. [21]CONG I,CHOI S,LUKIN M D.Quantum convolutional neural networks[J].Nature Physics,2019,15(12):1273-1278. |
|
||
首页