一种酉权重量子感知机  

Quantum Perceptron with the Unitary Weights

作　　者：周晓彦 嵇福高[1,2] 刘文杰 安星星 潘道蒙[4] 

机构地区：[1]江苏省气象传感网技术工程中心,南京210044 [2]南京信息工程大学电子与信息工程学院,南京210044 [3]江苏省大气环境与装备技术协同创新中心,南京210044 [4]南京信息工程大学计算机与软件学院,南京210044

出　　处：《小型微型计算机系统》2018年第3期534-538,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61373131和61501247)资助;江苏省"六大人才高峰"项目(2015-XXRJ-013)资助

摘　　要：通过将量子计算与经典感知机网络相结合,提出了一种基于酉权重的新型高效量子感知机算法.在算法过程中通过求解总的权重矩阵,并对其进行奇异值分解来保持其具有酉性.与其他人提出的量子感知机算法不同,本算法在非理想训练算例(超完备和欠完备)条件下,通过一次迭代训练学习可实现Hadamard门H、相位门S、受控非门CNOT、π/8门T这些基本量子门功能,这些基本量子门是构成任意量子门的标准集合,因此理论上该算法也能够实现任意量子门功能.最后,通过选择一个由多个基本量子门构成的组合门作为实例,以及随机选取一个训练集对算法的通用性进行了进一步验证.By combining quantum computing with classical perceptron network, a novel and efficient algorithm, called quantum percep- tron algorithm, based on the unitary weights is proposed. In order to make the weight matrix to be unitary, the total weight matrix is firstly computed, and then the singular value decomposition is utilized. Different from the previous quantum perceptron algorithms, the present algorithm can realize basic quantum gate functions ： Hadamard gate, phase gate, controlled-not （ CNOT ） gate, and ^-/8gate, which can be realized by one iteration training in the imperfect training case （ such as, over-complete, less-complete ）. These basic quantum gates are the standard set of arbitrary quantum gates, so theoretically the algorithm can also realize any quantum gate func- tion. Finally, the generality of the algorithm is further verified by selecting a composite gate composed of multiple basic quantum gates as an example and randomly selected a training set.

关 键 词：量子感知机 奇异值分解 酉性 量子门 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]