基于改进Hopfield神经网络的图像特征点匹配算法

针对图像特征点匹配算法的运行时间呈指数增长的问题,提出了一种新的匹配算法NHop.该算法通过加入新的网络输入输出函数、点对间差异的度量和启发式选择目标点的方式,对传统的Hopfield神经网络进行了改进.新算法不仅解决了传统Hopfield神经网络运行时间长、能量函数易陷入局部极小点的问题,而且也有效地实现了图像特征点的匹配.实验结果表明,与传统的Hopfield神经网络相比,NHop算法的匹配速度更快、准确率更高,对于图像特征点的匹配效果更好.     

一种基于抗体网络的免疫算法

阐述了免疫系统抗体网络的机理和特点,深入分析了抗体网络与常用的免疫算法和Hopfield神经网络异同.通过不断更新输入模式(抗原)和采用最优保存策略,将基于克隆选择的竞争学习算子、自动生成网络结构、剪枝算子和低频变异用于进化操作,提出一种新的基于抗体网络的免疫算法,用于函数优化问题.实验结果表明新算法可行有效.与常用的免疫算法、Hopfield神经网络优化算法比较,新算法具有较好的全局搜索能力和较快收敛速度.     

基于Hopfield神经网络的作业车间生产调度方法

该文提出了基于Hopfield神经网络的作业车间生产调度的新方法.文中给出了作业车 间生产调度问题(JSP)的约束条件及其换位矩阵表示,提出了新的包括所有约束条件的计算能 量函数表达式,得到相应的作业车间调度问题的Hopfield神经网络结构与权值解析表达式,并 提出相应的Hopfield神经网络作业车间调度方法.为了避免Hopfield神经网络容易收敛到局部 极小,从而产生非法调度解的缺点,将模拟退火算法应用于Hopfield神经网络求解,使Hopfield 神经网络收敛到计算能量函数的最小值0,从而保证神经网络输出是一个可行调度方案.该文 改进了已有文献中提出的作业调度问题的Hopfield神经网络方法,与已有算法相比,能够保证 神经网络稳态输出为可行的作业车间调度方案.     

一种基于遗传Hopfield神经网络求解TSP问题的算法

针对Hopfield网络求解TSP问题时出现无效解和收敛性能差的问题,对约束条件能量函数进行改进,构造了一种求解TSP问题的遗传Hopfield神经网络算法,并与经典Hopfield神经网络求解TSP方法进行对比.实验结果表明,本文算法具有更好的整体求解性能.     

物流中心选址算法改进及其Hopfield神经网络设计

在分析物流中心选址传统算法的基础上,引入一种新的选址模型,该模型能减少决策变量和约束条件的个数.利用该模型设计了一种Hopfield神经网络,将约束合并进网络结构从而将罚函数从能量函数中消除,使得网络的运行时间显著降低.为物流中心选址优化提供了一种新的方法.     

基于Hopfield神经网络的数据分类

针对Hopfield神经网络的自联想特性,提出一种新的带有粒子群优化过程的Hopfield分类算法(PSO-HOP).该算法采用了Blatt-Vergin (BV)学习算法,一定程度上克服了传统Hopfield容量低的特点.与此同时,还提出了先测量后训练的方法来降低算法的复杂度,提高分类效率,并探讨了样本属性以及类标号在Hopfield神经网络的表示方法,使其能够很好地处理空缺值等噪声数据.通过采用离散型粒子群优化算法对Hopfield的拓扑结构进行优化,可以将多余的神经元分配给不同属性,使得属性在分类中的权重发生改变,从而提高分类精度,避免陷入局部最优值.从统计不同属性被分配神经元的次数中,可以反映出不同属性的重要程度.从大量实验结果可以看出,该算法具有较高的鲁棒性和分类准确度.     

用改进的竞争Hopfield神经网络求解多边形近似问题

多边形近似是提取曲线特征点和简化曲线描述的一种重要方法.提出一种改进的Hopfield神经网络多边形近似算法,该算法利用选择拐点策略减少了搜索空间,重新定义了神经网络的能量函数,使其更能反映优化目标;引?入合并拆分搜索策略,有效帮助神经网络脱离局部最小值.实验结果表明,提出的改进算法是有效的,比其它算法如关键点检测法、竞争Hopfield神经网络、混沌Hopfield神经网络、遗传算法等具有更优的性能.     

基于混沌神经网络的最短路径路由算法

飞速发展的计算机网络对路由算法的反应速度提出了更高的要求．神经网络作为一种新的组合优化计算工具。在网络路由方面的应用得到较大关注．与传统的采用串行执行方式的算法相比，神经网络路由算法以其固有的并行执行方式，以及潜在的硬件实施能力，将成为这一领域的有力竞争者．由此提出了一种基于混沌神经网络的最短路径路由算法．仿真结果表明，该算法能有效克服Hopfield神经网络易陷入局部最优解的缺点，并且在收敛速度方面有了很大改进．     

ReLU激活函数优化研究

门控循环单元(GRU)是一种改进型的长短期记忆模型(LSTM)结构,有效改善了LSTM训练耗时的缺点.在GRU的基础上,对激活函数sigmoid,tanh,ReLU等性能进行了比较和研究,详细分析了几类激活函数的优缺点,提出了一种新的激活函数双曲正切线性单元(TLU).实验证明:新的激活函数既能显著地加快深度神经网络的训练速度,又有效降低训练误差.     

基于NAS-RIF算法和神经网络的图像盲复原

在图像盲复原中,NAS-RIF算法在无噪情况下,能够得到较好的复原结果,但是对有观测噪声的图像复原效果不理想。而Hopfield神经网络有利于缓解图像复原过程中的震铃效应,但前提是知道退化图像的点扩展函数。将二者相结合提出一种基于NAS-RIF算法和神经网络的图像盲复原新算法,首先由NAS-RIF算法先估计出退化图像的点扩展函数,再利用Hopfield神经网络算法对其进行复原。实验结果表明,该算法具有较好的盲复原效果。     

神经网络在多机动目标跟踪中的应用

为解决多目标跟踪中数据关联与状态估计的问题引入神经网络方法.针对联合概率数据关联(JPDA)存在的计算组合爆炸问题,利用Hopfield网络解决TSP问题的思路找到神经网络联合概率数据关联(NJPDA)方法,并对其进行一定的改进解决其参数确定问题.基于以上关联方法得到的关联概率,利用简化的信息融合自适应滤波算法,实现对目标轨迹的状态滤波与预测.以上综合方法充分利用了神经网络的优点,保证了多目标的跟踪精度及实时性.     

基于神经网络的机动多目标数据关联问题研究

本文通过对联合概率数据关联的性能特征的分析,将其归结为一类约束组合优化问题,在此基础上,利用Hopfield神经网络求解典型的约束组合优化问题(旅行推销员问题)的方法,解决了传统的联合概率数据关联中出现的计算量组合爆炸现象,仿真结果表明,该方法效果良好,在密集多回波环境下,其优越性能更为突出。     

Achieving a More Robust Neural Network Model for Control of a MR Damper by Signal Sensitivity Analysis

Most neural network models can work accurately on their trained samples, but when encountering noise, there could be significant errors if the trained neural network is not robust enough to resist the noise. Sensitivity to perturbation in the control signal due to noise is very important for the prediction of an output signal. The goal of this paper is to provide a methodology of signal sensitivity analysis in order to enable the selection of an ideal Multi-Layer Perception (MLP) neural network model from a group of MLP models with different parameters, i.e. to get a highly accurate and robust model for control problems. This paper proposes a signal sensitivity which depends upon the variance of the output error due to noise in the input signals of a single output MLP with differentiable activation functions. On the assumption that noise arises from additive/multiplicative perturbations, the signal sensitivity of the MLP model can be easily calculated, and a method of lowering the sensitivity of the MLP model is proposed. A control system of a magnetorheological (MR) fluid damper, which is a relatively new type of device that shows the future promise for the control of vibration, is modelled by MLP. A large number of simulations on the MR damper’s MLP model show that a much better model is selected using the proposed method.     

Incorporating biological knowledge for construction of fuzzy networks of gene associations

Gene association networks have become one of the most important approaches to modelling of biological processes by means of gene expression data. According to the literature, co-expression-based methods are the main approaches to identification of gene association networks because such methods can identify gene expression patterns in a dataset and can determine relations among genes. These methods usually have two fundamental drawbacks. Firstly, they are dependent on quality of the input dataset for construction of reliable models because of the sensitivity to data noise. Secondly, these methods require that the user select a threshold to determine whether a relation is biologically relevant. Due to these shortcomings, such methods may ignore some relevant information.We present a novel fuzzy approach named FyNE (Fuzzy NEtworks) for modelling of gene association networks. FyNE has two fundamental features. Firstly, it can deal with data noise using a fuzzy-set-based protocol. Secondly, the proposed approach can incorporate prior biological knowledge into the modelling phase, through a fuzzy aggregation function. These features help to gain some insights into doubtful gene relations.The performance of FyNE was tested in four different experiments. Firstly, the improvement offered by FyNE over the results of a co-expression-based method in terms of identification of gene networks was demonstrated on different datasets from different organisms. Secondly, the results produced by FyNE showed its low sensitivity to noise data in a randomness experiment. Additionally, FyNE could infer gene networks with a biological structure in a topological analysis. Finally, the validity of our proposed method was confirmed by comparing its performance with that of some representative methods for identification of gene networks     

r-SVR中参数r与输入噪声间线性反比关系的仿真研究

为研究r范数-支持向量回归机r-SVR的鲁棒性,验证r-SVR中参数r与输入噪声方差之间的近似反比线性关系,对r-SVR进行了仿真.推导出了作为仿真的依据的r-SVR的解的形式和对其进行求解的牛顿迭代公式.仿真结果显示：输入噪声为高斯分布时,r-SVR中参数r与输入噪声方差之间存在近似线性反比关系;这一关系曲线随着信噪比增加而斜率减小、整个曲线下移.这一结果印证和丰富了现前的理论推导结果,为在已知输入高斯噪声方差时合理地选择r提供了更可信的依据.     

Sensitivity Analysis for Decision Boundaries

A novel approach is presented to visualize and analyze decision boundaries for feedforward neural networks. First order sensitivity analysis of the neural network output function with respect to input perturbations is used to visualize the position of decision boundaries over input space. Similarly, sensitivity analysis of each hidden unit activation function reveals which boundary is implemented by which hidden unit. The paper shows how these sensitivity analysis models can be used to better understand the data being modelled, and to visually identify irrelevant input and hidden units.     

Noise in integrate-and-fire neurons: from stochastic input to escape rates

We analyze the effect of noise in integrate-and-fire neurons driven by time-dependent input and compare the diffusion approximation for the membrane potential to escape noise. It is shown that for time-dependent subthreshold input, diffusive noise can be replaced by escape noise with a hazard function that has a gaussian dependence on the distance between the (noise-free) membrane voltage and threshold. The approximation is improved if we add to the hazard function a probability current proportional to the derivative of the voltage. Stochastic resonance in response to periodic input occurs in both noise models and exhibits similar characteristics.     

Regularizing the effect of input noise injection in feedforward neural networks training

Injecting input noise during feedforward neural network (NN) training can improve generalization performance markedly. Reported works justify this fact arguing that noise injection is equivalent to a smoothing regularization with the input noise variance playing the role of the regularization parameter. The success of this approach depends on the appropriate choice of the input noise variance. However, it is often not known a priori if the degree of smoothness imposed on the FNN mapping is consistent with the unknown function to be approximated. In order to have a better control over this smoothing effect, a loss function putting in balance the smoothed fitting induced by the noise injection and the precision of approximation, is proposed. The second term, which aims at penalizing the undesirable effect of input noise injection or controlling the deviation of the random perturbed loss, was obtained by expressing a certain distance between the original loss function and its random perturbed version. In fact, this term can be derived in general for parametrical models that satisfy the Lipschitz property. An example is included to illustrate the effectiveness of learning with this proposed loss function when noise injection is used.     

Nonlinear system identification using a Gabor/Hopfield network

This paper presents a method of nonlinear system identification using a new Gabor/Hopfield network. The network can identify nonlinear discrete-time models that are affine linear in the control. The system need not be asymptotically stable but must be bounded-input-bounded-output (BIBO) stable for the identification results to be valid in a large input-output range. The network is a considerable improvement over earlier work using Gabor basis functions (GBF's) with a back-propagation neural network. Properties of the Gabor model and guidelines for achieving a global error minimum are derived. The new network and its use in system identification are investigated through computer simulation. Practical problems such as local minima, the effects of input and initial conditions, the model sensitivity to noise, the sensitivity of the mean square error (MSE) to the number of basis functions and the order of approximation, and the choice of forcing function for training data generation are considered.     

鲁棒r-支持向量回归机中参数r 的选择研究

为使r-支持向量回归机更具鲁棒性．研究了r-支持向量回归机r-SVR中参数r与输入噪声之间的关系．将r-SVR的优化问题转换成最大后验估计问题．推导出r-SVR后验估计最大化的条件．得出了输入噪声为高斯分布时r-SVR中参数r与σ之间的近似线性反比关系．为已知输入高斯噪声方差σ时合理选择r提供了理论依据．