量子门线路神经网络及其改进学习算法研究

量子门线路神经网络（QGCNN）是一种直接利用量子理论设计神经网络拓扑结构或训练算法的量子神经网络模型。动量更新是在神经网络的权值更新中加入动量,在改变权值向量的同时提供一个特定的惯量,从而避免权值向量在网络训练过程中持续振荡。在基本的量子门线路神经网络的学习算法中引入动量更新原理,提出了一种具有动量更新的量子门线路网络算法（QGCMA）。研究表明,QGCMA保持了网络100%的收敛率,同时,相对于基本算法,在具有相同学习速率的情况下,提高了网络的收敛速度。     

基于QPSO的BP神经网络油田节能指标预测

针对BP神经网络易陷入局部极小问题以及收敛速度慢的问题, 引入量子粒子群优化算法和BP神经网络相结合的方法, 共享BP神经网络强大的灵活性和量子粒子群全局搜索能力强的优势, 通过改进QPSO的平均最优位置的计算方法, 实现基于BP神经网络和量子粒子群的油田节能指标预测. 以大庆某采油厂注水泵机组单耗数据为训练数据, 预测结果表明该方法能达到良好的预测效果, 具有可行性.     

基于Tanh多层函数的量子神经网络算法及其应用的研究

该文提出一种新的改进激励函数的量子神经网络模型。首先为了提高学习速率,在网络权值训练过程中引入了动量项。然后为了有效实现相邻类之间具有覆盖和不确定边界的分类问题,新网络采用区分度更大的双曲正切函数的叠加作为其隐层激励函数。最后将该算法用于字符识别,将双曲正切激励函数的量子神经网络应用于数字、字母和汉字样本的多次实验,并且与原多层激励函数量子神经网络和BP网络的实验效果进行比较,发现改进后量子神经网络不仅具有较高的识别率,而且在样本训练次数上相对原多层激励函数量子神经网络有明显减少。仿真结果证明该方法的优越性。     

一种量子衍生神经网络模型

为提高神经网络的逼近能力,通过在普通BP网络中引入量子旋转门,提出了一种新颖的量子衍生神经网络模型. 该模型隐层由量子神经元组成,每个量子神经元携带一组量子旋转门,用于更新隐层的量子权值,输入层和输出层均为普通神经元. 基于误差反传播算法设计了该模型的学习算法. 模式识别和函数逼近的实验结果验证了提出模型及算法的有效性.     

基于量子小波神经网络的传动装置模式识别

针对传统神经网络收敛精度低,以及用于故障模式识别能力差的问题,提出了将量子神经网络与小波理论相结合的量子小波神经网络模型.该模型隐层量子神经元采用小波基函数的线性叠加作为激励函数,给出了网络学习算法,并以某型传动装置监测信号的小波能量谱为训练样本,识别传动装置带有缺损的齿轮故障征兆.仿真结果表明,量子小波神经网络能够提高神经网络训练精度和故障征兆识别精度.     

关于无线传感器网络定位算法仿真

针对无线传感器网络(WSNs)节点定位的问题,提出了一种通过构建粒子群机制的量子神经网络模型优化距离矢量跳跃(DV-HOP)的定位算法(PSO-QNN),根据传统DV-HOP所得到的平均距离和实测节点距离构建量子神经网络模型,并通过粒子群算法对平均距离进行训练,从而得到较优平均值,实现了对DV-HOP算法的优化.算法缩短了传统人工神经网络的训练时间,并且加快了收敛速度.仿真结果表明:与传统DV-HOP算法相比,所提出的PSO-QNN算法能够减少约20%的定位误差,定位精度显著提高.     

量子小波神经网络及其在漏钢预报中的应用

针对传统神经网络收敛速度慢,收敛精度低,以及用于模式识别泛化能力差的问题。提出了将量子神经网络与小波理论相结合的量子小波神经网络模型。该模型隐层量子神经元采用小波基函数的线性叠加作为激励函数,称之为多层小波激励函数,这样隐层神经元既能表示更多的状态和量级,又能提高网络收敛精度和速度。给出了网络学习算法。并以之在漏钢预报波形识别中的应用验证了该模型和学习算法的有效性。     

基于量子门线路的量子神经网络模型及算法

提出一种量子神经网络模型及算法.该模型为一组量子门线路.输入信息用量子位表示,经量子旋转门进行相位旋转后作为控制位,控制隐层量子位的翻转;隐层量子位经量子旋转门进行相位旋转后作为控制位,控制输出层量子位的翻转.以输出层量子位中激发态的概率幅作为网络输出,基于梯度下降法构造了该模型的学习算法.仿真结果表明,该模型及算法在收敛能力和鲁棒性方面均优于普通BP网络.     

基于LM算法的在线自适应RBF网结构优化算法

针对LM算法不能在线训练RBF网络以及RBF网络结构设计算法中存在的问题,提出一种基于LM算法的在线自适应RBF网络结构优化算法.该算法引入滑动窗口和在线优化网络结构的思想,滑动窗口的引入既使得LM算法能够在线训练RBF网络,又使得网络对学习参数的变化具有更好的鲁棒性,并且易于收敛.在线优化网络结构使得网络在学习过程中能够根据训练样本的训练误差和隐节点的相关信息,在线自适应调整网络结构,跟踪非线性时变系统的变化,使网络维持最为紧凑的结构,以保证网络的泛化性能.最后通过仿真实验验证了所提出算法的性能.     

基于改进量子遗传算法的过程神经元网络训练

针对过程神经元网络由于模型参数较多BP算法不易收敛的问题,提出一种基于量子位Bloch坐标的量子遗传算法.将该算法融合于过程神经网络的训练.按权值参数的个数确定量子染色体上的基因数并完成种群编码,通过新的量子旋转门完成个体的更新.算法中的每条染色体携带3条基因链,因此可扩展对解空间的遍历性,加速优化进程.以两组二维三角函数的模式分类问题为例,仿真结果表明该方法不仅收敛速度快,而且寻优能力强.     

量子计算机的研究与应用

介绍了量子计算的最新研究方向,简述了量子计算和量子信息技术在保密通信、量子算法、数据库搜索等重要领域的应用。分析了量子计算机与经典计算机相比所具有的优点和目前制约量子计算机应用发展的主要因素,最后展望了其未来发展趋势。     

量子小波变换算法及线路实现

随着小波理论研究的深入,以及小波分析在信号分析和图像处理等领域的广泛应用,小波分析在量子计算领域中也越来越受到重视.应用置换矩阵、W-H变换矩阵和量子傅立叶变换矩阵来对Haar小波及D(4)小波变换矩阵进行分解,给出其算法,然后得出其完整的量子逻辑线路图,最后分析其复杂度.     

A Survey on quantum computing technology

The power of quantum computing technologies is based on the fundamentals of quantum mechanics, such as quantum superposition, quantum entanglement, or the no-cloning theorem. Since these phenomena have no classical analogue, similar results cannot be achieved within the framework of traditional computing. The experimental insights of quantum computing technologies have already been demonstrated, and several studies are in progress. Here we review the most recent results of quantum computation technology and address the open problems of the field.     

Quantum estimation,control and learning: Opportunities and challenges

The development of estimation and control theories for quantum systems is a fundamental task for practical quantum technology. This vision article presents a brief introduction to challenging problems and potential opportunities in the emerging areas of quantum estimation, control and learning. The topics cover quantum state estimation, quantum parameter identification, quantum filtering, quantum open-loop control, quantum feedback control, machine learning for estimation and control of quantum systems, and quantum machine learning.     

量子力学和量子计算机

In this paper,the relationship between computation and physics and the application of the principle of Quantum mechanics to Quantum Computing and Quantum Computers was reviewed     

Mathematical models of quantum computation

In this paper, we introduce two mathematical models of realistic quantum computation. First, we develop a theory of bulk quantum computation such as NMR (Nuclear Magnetic Resonance) quantum computation. For this purpose, we define bulk quantum Turing machine (BQTM for short) as a model of bulk quantum computation. Then, we define complexity classes EBQP, BBQP and ZBQP as counterparts of the quantum complexity classes EQP, BQP and ZQP, respectively, and show that EBQP=EQP, BBQP=BQP and ZBQP=ZQP. This implies that BQTMs are polynomially related to ordinary QTMs as long as they are used to solve decision problems. We also show that these two types of QTMs are also polynomially related when they solve a function problem which has a unique solution. Furthermore, we show that BQTMs can solve certain instances of NP-complete problems efficiently. On the other hand, in the theory of quantum computation, only feed-forward quantum circuits are investigated, because a quantum circuit represents a sequence of applications of time evolution operators. But, if a quantum computer is a physical device where the gates are interactions controlled by a current computer such as laser pulses on trapped ions, NMR and most implementation proposals, it is natural to describe quantum circuits as ones that have feedback loops if we want to visualize the total amount of the necessary hardware. For this purpose, we introduce a quantum recurrent circuit model, which is a quantum circuit with feedback loops. LetC be a quantum recurrent circuit which solves the satisfiability problem for a blackbox Boolean function includingn variables with probability at least 1/2. And lets be the size ofC (i.e. the number of the gates inC) andt be the number of iterations that is needed forC to solve the satisfiability problem. Then, we show that, for those quantum recurrent circuits, the minimum value ofmax(s, t) isO(n ²2 ^n/3). Tetsuro Nishino, D.Sc.: He is presently an Associate Professor in the Department of Information and Communication Engineering, The University of Electro-Communications. He received the B.S., M.S. and D.Sc degrees in mathematics from Waseda University, in 1982, 1984 and 1991 respectively. From 1984 to 1987, he joined Tokyo Research Laboratory, IBM Japan. From 1987 to 1992, he was a Research Associate of Tokyo Denki University, and from 1992 to 1994, he was an Associate Professor of Japan Advanced Institute of Science and Technology, Hokuriku. His main interests are circuit complexity theory, computational learning theory and quantum complexity theory.     

Self-protected quantum algorithms based on quantum state tomography

Only a few classes of quantum algorithms are known which provide a speed-up over classical algorithms. However, these and any new quantum algorithms provide important motivation for the development of quantum computers. In this article new quantum algorithms are given which are based on quantum state tomography. These include an algorithm for the calculation of several quantum mechanical expectation values and an algorithm for the determination of polynomial factors. These quantum algorithms are important in their own right. However, it is remarkable that these quantum algorithms are immune to a large class of errors. We describe these algorithms and provide conditions for immunity.      

Optimal computation with non-unitary quantum walks

Quantum versions of random walks on the line and the cycle show a quadratic improvement over classical random walks in their spreading rates and mixing times, respectively. Non-unitary quantum walks can provide a useful optimisation of these properties, producing a more uniform distribution on the line, and faster mixing times on the cycle. We investigate the interplay between quantum and random dynamics by comparing the resources required, and examining numerically how the level of quantum correlations varies during the walk. We show numerically that the optimal non-unitary quantum walk proceeds such that the quantum correlations are nearly all removed at the point of the final measurement. This requires only O(logT)$O(logT)$ random bits for a quantum walk of T$T$ steps.     

Combating Quantum Burst-errors with Quantum Interleaving Technique

Based on the interleaving technique, a kn-qubit code is constructed in this paper with more error-correcting ability than one n-qubit quantum error-correcting code without introducing the redundant qubits. By converting quantum bursts of errors into quantum random errors with the help of the quantum interleaving of the several states of the same quantum code, the proposed technique becomes an effective means to combat quantum bursts of errors. It is much simple and applicable for the quantum interleaving techniques to be used in the optical-fiber communications.     

Quantum covariance and filtering

We give a tutorial exposition of the analogue of the filtering equation for quantum systems focusing on the quantum probabilistic framework and developing the ideas from the classical theory. Quantum covariances and conditional expectations on von Neumann algebras play an essential part in the presentation.