随机混成系统稀有属性的统计模型检测方法

统计模型检测,已成为随机混成系统安全性验证的重要方法.但对安全性要求较高的系统,其不安全事件和系统失效都是稀有事件.在这种情况下,统计模型检测很难采样到满足稀有属性的样本而变得不可行.针对该问题,提出了交叉熵迭代学习的统计模型检测方法首先,使用连续时间马尔可夫链表示随机混成系统的路径概率空间,推导出路径空间上的参数化概率分布函数族;然后构造了随机混成系统路径空间上的交叉熵优化模型,提出了在路径空间上迭代学习最优重要性采样分布的算法;最后给出了基于重要性采样的稀有属性验证算法.实验结果表明:该方法能够有效地对随机混成系统的稀有属性进行验证;且在相同样本数量下,与一些启发式重要性采样方法相比,该方法的估计值能够更好地分布在均值附近,标准方差和相对误差减少超过了一个数量级.     

A note on information entropy measures for vague sets and its applications

A new nonprobabilistic entropy of a vague set is proposed by means of the intersection and union of the membership degree and nonmembership degree of the vague set. The concept called vague cross-entropy of vague sets will also be discussed and its definition is given by analogy with the cross-entropy of probability distributions. Finally, two numeric examples are presented to illustrate the applications of vague cross-entropy to pattern recognition and medical diagnosis.     

Fault diagnosis of turbine based on fuzzy cross entropy of vague sets

The fuzzy cross entropy of vague sets, so-called vague cross entropy, is introduced by analogy with the cross entropy of probability distributions. And then a new method of the fault diagnosis is proposed on the basis of the vague cross entropy and is applied to the fault diagnosis of turbine. The vague cross entropy between a testing sample and the knowledge of system faults is evaluated in the fault diagnosis of the turbine vibration. If the cross-entropy value is small, the testing sample is near to a type of fault knowledge. Then, the type of vibration fault is determined according to the minimum cross-entropy value. The fault-diagnosis example of the turbine demonstrates that the proposed method cannot only diagnose the main fault types of the turbine, it can also detect useful information for future trends and multi-fault analysis.     

基于改进最小分类误差准则算法的深度学习研究——以台风卫星云图为例

针对传统基于最小分类误差准则（MCE）建立的目标函数存在样本错分类时网络出现的梯度反向问题,引入最小分类误差准则,定义带修正项的FMCE目标函数。以较高精度的交叉熵作为基函数,将FMCE作为修正函数,提出改进交叉熵目标函数CE-FMCE,使得网络在反向传播过程中提升标签类输出的概率。CE-FMCE不仅克服了传统MCE目标函数的梯度反向问题,还弥补了交叉熵函数对非标签集梯度不作区分处理的不足。分别在自建台风云图数据集和通用数据集MNIST上对CE-FMCE和MSE、交叉熵、MCE、M3CE进行对比实验,实验结果表明CE-FMCE优于其他目标函数。     

Active probing for information in control systems with quantizedstate measurements: a minimum entropy approach

In this paper the effect of state quantization in scaler discrete-time linear control systems is studied by analyzing the system as a partially observed stochastic system. The problem of optimal state information gathering and filtering is investigated using information theoretic measures and formulating the state estimation problem as an entropy optimization problem. The active probing effect of the feedback control is thoroughly studied. Optimal feedback controls which minimize various types of entropy costs are determined, and it is shown that this problem is equivalent to an optimal control problem for a controlled Markov chain     

Deterministic and stochastic control of discrete-time bilinear systems

Optimal control of a class of time invariant single-input, discrete bilinear systems is investigated in this paper. Both deterministic and stochastic problems are considered.

In the deterministic problem, for the initial state in a certain set ∑₀, the solution is the same as the solution to the associated linear system. The optimal path may be a regular path or a singular path.

The stochastic control problem is considered with perfect state observation, and additive and multiplicative noise in the state equation. It is demonstrated that the presence of noise simplifies the analysis compared to that in the determinstic case.     

混合型概率犹豫模糊熵和交叉熵测度

针对已有概率犹豫模糊熵测度构造复杂、区分能力弱等缺点,提出一种混合型概率犹豫模糊熵测度.混合型熵测度能够综合反映概率犹豫模糊元所具有的个体不确定性和整体不确定性,具有结构简单、物理意义明确、区分能力强等优势.在概率犹豫模糊元规范化的基础上,基于混合型熵测度的构造理念所设计的混合型交叉熵测度,能够克服已有交叉熵测度的设计缺陷,综合反映两个概率犹豫模糊元之间的个体区分度和整体区分度,且具有自然的对称性.基于混合型熵测度和交叉熵测度,进一步设计概率犹豫模糊环境下的多属性决策方法,并将其应用于无人机集群作战效能评估.数值和理论分析结果表明,所提出的混合型概率犹豫模糊熵和交叉熵测度能够成对设计,互为印证,具有广泛的应用前景.     

一种新的区间直觉模糊熵及其应用

提出了区间值直觉模糊集的区间直觉模糊交叉熵,这种交叉熵充分考虑了区间值直觉模糊集的隶属度,非隶属度以及犹豫度。给出一种区间值直觉模糊集的区间直觉模糊熵的公理化体系,并且基于直觉模糊交叉熵公式给出一种区间直觉模糊熵的具体测度公式。利用区间值直觉模糊集的加权相关系数,将提出的熵公式应用于解决属性权重完全未知的区间直觉模糊多属性决策问题。     

Extracting finite-state representations from recurrent neuralnetworks trained on chaotic symbolic sequences

Concerns neural-based modeling of symbolic chaotic time series. We investigate the knowledge induction process associated with training recurrent mural nets (RNN) on single long chaotic symbolic sequences. Even though training RNN to predict the next symbol leaves the standard performance measures such as the mean square error on the network output virtually unchanged, the nets extract a lot of knowledge. We monitor the knowledge extraction process by considering the nets stochastic sources and letting them generate sequences which are then confronted with the training sequence via information theoretic entropy and cross-entropy measures. We also study the possibility of reformulating the knowledge gained by RNN in a compact easy-to-analyze form of finite-state stochastic machines. The experiments are performed on two sequences with different complexities measured by the size and state transition structure of the induced Crutchfield's epsilon-machines (1991, 1994). The extracted machines can achieve comparable or even better entropy and cross-entropy performance. They reflect the training sequence complexity in their dynamical state representations that can be reformulated using finite-state means. The findings are confirmed by a much more detailed analysis of model generated sequences. We also introduce a visual representation of allowed block structure in the studied sequences that allows for an illustrative insight into both RNN training and finite-state stochastic machine extraction processes.     

Decentralized optimization for distributed-lag models of discrete systems

The approach discussed in this paper solves a general class of optimization problems for discrete dynamic systems which include distributed lag, distributed and/or multiple pure delays, and constraints both in state and control variables. The overall system equation of this problem is described by a multidimensional nonlinear difference equation of high-order which is called the distributed-lag model. Applying Lagrange duality theory to the original problem, the dual problem is formulated, and the decomposition of the optimization process in stage is obtained. It is shown that by solving the dual problem the delay terms can be easily handled and the optimal solution to the original problem is obtained without reducing the multi-dimensional high-order system equation to a conventional larger dimensional first-order system equation. It is also shown that the dual decentralized method in this paper is easier to cope with state and control constraints than the primal method in the space of control, i.e. gradient and other techniques. The approach developed in this paper is compared with other methods using a simple example, and is applied to a combined marketing and production control problem. Some computational results are included.     

Towards an approach to stochastic adaptive control using the maximum entropy principle

The crucial problem in non-linear stochastic adaptive systems via the certainly equivalence principle is the estimation of the disturbances, which is essentially a non-linear estimation. The present paper focuses mainly on this aspect of adaptation, and the basic idea is of using the maximum entropy principle together with constraints suitably chosen. In this way one proposes a new technique for solving the Fokker-Plank-Kolmogorov equation and two new techniques for determining the conditional probability density of a random disturbance in a stochastic process. Then an adiabatic elimination is proposed, which applies when the system is slowly varying with respect to the external parameter. Finally, one shows how the dynamic equations of the state moments, combined with a linearization technique, can be utilized to analyse a broad class of non-linear stochastic systems involving random disturbances with small variances.     

Some information measures for interval-valued intuitionistic fuzzy sets

A new information entropy measure of interval-valued intuitionistic fuzzy set (IvIFS) is proposed by using membership interval and non-membership interval of IvIFS, which complies with the extended form of Deluca-Termini axioms for fuzzy entropy. Then the cross-entropy of IvIFSs is presented and the relationship between the proposed entropy measures and the existing information measures of IvIFSs is discussed. Additionally, some numerical examples are given to illustrate the applications of the proposed entropy and cross-entropy of IvIFSs to pattern recognition and decision-making.     

Unit-Linking PCNN和图像熵的彩色图像分割与边缘检测

在RGB空间中,将最大香农熵准则和最小交叉熵准则两种评价准则与大概率合并和小概率合并两种合并策略相结合,提出基于Unit-Linking PCNN的四种彩色图像分割方案,并在各分量分割结果基础上利用Unit-Linking PCNN实施边缘检测,合并得到彩色图像的边缘检测结果。分析了各评价准则和合并策略的优劣,比较了各分割方案条件下的图像分割和边缘检测效果。与HSV空间中得到的相关结果进行分析比较,该文分割和边缘检测结果体现了图像的更多的细节,说明了在RGB空间中进行彩色图像分割和边缘检测的合理性。与相关文献结果相比,该方法的模型参数对图像分割结果的影响较不敏感。计算机仿真结果表明,该方法具有较好的彩色图像分割和边缘检测效果,具有较强适用性。     

基于改进微粒群优化的学习路径优化控制方法

对学生学习的路径控制在智能化教学系统中是一个重要的问题。该文以知识空间理论为基础建立了学习状态空间,通过改进的微粒群算法对该学习状态空间的学习路径进行最优化控制,并利用死亡惩罚函数法把约束最优化学习路径问题转化成了无约束的最优化学习路径控制问题,引入交换子和交换序的概念对微粒群算法进行改进。在结果分析中,通过动态参数法,即动态变化交换子保留概率的方法提高微粒群的收敛效果,达到了最优化学习路径控制的目的。     

求解PCB钻孔机刀具路径规划的交叉熵方法

工作刀具路径的优化程度是PCB钻孔机的重要性能指标,对其进行很好的优化有助于提高PCB设备的加工质量和加工速度.首先对刀具路径进行建模,然后对交叉熵算法进行描述并应用交叉熵方法对刀具路径进行求解.实验结果表明,选择交叉熵方法对环境进行建模简单、有效,在求解刀具路径规划方面具有一定的优势.     

混合交叉熵算法求解复杂零等待流水线调度问题

针对制造行业中广泛存在的一类复杂零等待流水线调度问题, 即带序相关设置时间和释放时间的零等待 流水线调度问题(NFSSP SDSTs RTs), 建立问题的排序模型并提出一种混合交叉熵算法(HCEA)进行求解, 优化目 标为最小化总提前和延迟时间. 首先, 设计了一种基于问题性质的快速评价方法, 有效降低评价解的计算复杂度. 其 次, 采用交叉熵算法学习并积累优质解的结构特征, 建立概率模型对优质解的工件块分布进行有效地估计. 通过合 理的采样和更新方法, 实现对解空间中优质区域的全局搜索. 然后, 为提高算法搜索效率, 设计带两种搜索策略的快 速局部搜索方法, 对全局搜索发现的优质区域进行细致且深入的搜索. 最后, 仿真实验与算法对比验证了HCEA可 有效求解NFSSP SDSTs RTs.     

基于路径聚合扩张卷积的图像语义分割方法

基于编码器-解码器的深度全卷积神经网络在图像语义分割中取得了重大的进展,但是深度网络中网络低层定位信息传播到网络高层路径过长,导致解码阶段难以利用低层定位信息来恢复物体边界结构,针对这一问题,提出了一种应用在分割网络解码器部分的路径聚合结构。该结构缩短了分割网络中低层信息到高层信息的传播路径并提供多尺度的上下文语义信息,使得分割网络能产生更为精细的边界分割结果。针对语义分割中常使用的Softmax交叉熵损失函数对外观相似样本区分能力不足的问题,对Softmax交叉熵损失函数进行改造,提出了双向交叉熵损失函数。本文提出的路径聚合扩张卷积网络结合新的损失函数方法在PASCAL VOC2012Aug数据集上获得了更好的效果,将mIoU值从78.77%提升到了80.44%。     

基于MGC法的非线性随机系统PDF控制

对于非线性随机系统,以均值、方差等低阶统计特征作为研究目标往往难以满足实际的控制要求,需要考虑更高阶的统计特征.概率密度函数(Probability density function,PDF)包含了完全统计特征,因此PDF控制能够实现各阶矩的有效控制.针对受高斯白噪声激励的非线性随机系统,将福克-普朗克-柯尔莫哥洛夫(Fokker-Planck-Kolmogrov,FPK)方程作为研究工具,提出一种基于多高斯闭合法(MGC)的PDF控制方法.首先,根据目标PDF的形状构造一个由多个高斯型PDF相叠加的PDF;然后,构造一个优化问题,使得该PDF逼近目标PDF;进一步,通过求解FPK方程得到被控系统的状态方程;最后,结合原始状态方程求得控制函数,实现对目标PDF的追踪控制.针对不同类型目标PDF进行的仿真结果表明了所提出方法的可行性和有效性.     

Remarks on the application of dynamic programming to the optimal path timing of robot manipulators

In this paper, we investigate the use of the dynamic programming approach in the solution of the optimal path timing problem in robotics. This problem is computationally feasible because the path constraint reduces the dimension of the state in the problem to two. The Hamilton–Jacobi–Bellman equation of dynamic programming, a nonlinear first order partial differential equation, is presented and is solved approximately using finite difference methods. Numerical solution of this results in the optimal policy which can then be used to define the optimal path timing by numerical integration. Issues relating to the convergence of the numerical schemes are discussed, and the results are applied to an experimental SCARA manipulator. © 1998 John Wiley & Sons, Ltd.     

Bionic autonomous learning control of a two-wheeled self-balancing flexible robot

This paper presents an OCPA (operant conditioning probabilistic automaton) bionic autonomous learning system based on Skinner’s operant conditioning theory for solving the balance control problem of a two-wheeled flexible robot. The OCPA learning system consists of two stages: in the first stage, an operant action is selected stochastically from a set of operant actions and then used as the input of the control system; in the second stage, the learning system gathers the orientation information of the system and uses it for optimization until achieves control target. At the same time, the size of the operant action set can be automatically reduced during the learning process for avoiding little probability event. Theory analysis is made for the designed OCPA learning system in the paper, which theoretically proves the convergence of operant conditioning learning mechanism in OCPA learning system, namely the operant action entropy will converge to minimum with the learning process. And then OCPA learning system is applied to posture balanced control of two-wheeled flexible self-balanced robots. Robot does not have posutre balanced skill in initial state and the selecting probability of each operant in operant sets is equal. With the learning proceeding, the selected probabilities of optimal operant gradually tend to one and the operant action entropy gradually tends to minimum, and so robot gradually learned the posture balanced skill.