摘要
贝叶斯分类算法是一种基于概率统计理论的有监督学习算法,常被用于分类问题中.本文将量子计数与经典贝叶斯分类算法相结合,提出一种新的量子贝叶斯分类算法.通过量子随机访问存储器制备所需的量子态,使用oracle进行相位翻转并构造与之所对应的操作算子,在操作算子的本征态空间上重新描述量子态,借助辅助粒子进行相位估计,投影测量后即可高效地计算出贝叶斯分类所需的数据,实现量子贝叶斯分类算法.该算法在低维特征空间中与经典算法相比有着指数级加速.
Bayesian classification algorithm is a supervised learning algorithm based on the statistics theory of probability,which is often used in classification problems.In this paper,a new quantum Bayesian classification algorithm is proposed by combining quantum counting with classical Bayesian classification algorithm.The required quantum states are prepared by a quantum random access memory,the oracle is used to phase flip and construct the corresponding operator,the quantum states are redescribed on the eigenstate space of the operator,and phase estimation is performed with the help of auxiliary particles.Then,the data required for Bayesian classification can be efficiently calculated after projection measurements and the quantum Bayesian classification algorithm can be realized.Compared with the classical algorithm,this algorithm has exponential acceleration in the low dimensional feature space.
作者
陆春悦
郭躬德
林崧
Lu Chunyue;Guo Gongde;Lin Song(College of Computer and Cyber Security,Fujian Normal University,Fuzhou 350117,China)
出处
《南京师大学报(自然科学版)》
CAS
CSCD
北大核心
2021年第4期117-121,共5页
Journal of Nanjing Normal University(Natural Science Edition)
基金
国家自然科学基金项目(61976053、61772134)
福建省高等学校新世纪优秀人才支持计划
福建省自然科学基金项目(2018J01776).
关键词
量子机器学习
贝叶斯分类
二元分类
量子计数
相位估计
quantum machine learning
Bayesian classification
binary classification
quantum counting
phase estimation