贝叶斯神经网络在误差修正中的应用研究

针对传统神经网络在实际应用中暴露的泛化能力不强的缺点,使用正则化方法通过修正神经网络的训练性能函数来提高神经网络的泛化能力。再将贝叶斯推理与神经网络相结合,可以在网络训练过程中自适应地调节正则化参数的大小,并使其达到最优。最后通过仿真方法验证贝叶斯神经网络在误差修正中的可行性。     

贝叶斯正则化BP神经网络在油气钻井成本预测的应用

对反映油田绩效的油气钻井成本进行准确预测,有助于做出科学的决策和评估。为了解决在运用BP神经网络进行油气钻井成本预测过程中,油气钻井成本影响因子确定难以及标准BP神经网络泛化能力差的问题,建立了基于主分量分析的贝叶斯正则化的BP神经网络油气钻井成本预测模型,并结合中国石油某公司各区块钻井成本数据,验证了该模型具有较高的预测精度及实用性。     

基于贝叶斯网络的电子商务质量评估研究

针对现有电子商务质量评估中存在的问题,提出了将静态质量概念模型与贝叶斯网络相结合的电子商务质量动态评估方法。并通过数据仿真建立了电子商务质量模型实例,对贝叶斯网络在B2C电子商务质量动态评估中的应用进行了探讨;最后将该评估方法与BP神经网络、决策树模型的评估结果进行比较,验证了使用该动态评估方法进行质量评估的准确性和优越性。     

正则化神经网络与提前终止迭代的比较分析

神经网络算法是一种非常经典的分类算法,然而神经网络的一个不足之处就是容易陷入过拟合。针对这种不足,正则化神经网路算法与提前终止迭代算法被提了出来。为了进一步研究这两种算法性能的差异,本文通过20个UCI标准数据集上对着这两种方法进行了性能测试。实验显示在分类准确率上正则化神经网路算法要更优秀一些,但是在分类速度上提前终止迭代算法更占优势。     

基于同伦正则化的最小二乘支持向量机

最小二乘支持向量机作为一种新颖的人工智能技术,已越来越广泛地运用于各个学科领域,其本质是一种正则化方法。近几年出现了一种新的正则化方法:同伦正则化,该方法已在多个方面得到了很好的应用。作者通过研究,将这种新的正则化方法应用于最小二乘支持向量机,建立了一种新的模型。该模型与最小二乘支持向量机相比,其最大的优点就是大大缩短了正则化参数的优化时间。     

一类非线性贝叶斯动态模型的算法研究

本文利用自主滤波与神经网络相结合的方法,对非线性贝叶斯动态模型进行了简单的探讨,从而给模型的短期预测以自适应,自纠正的智能化功能,为非线性贝叶斯动态系统的预测提供了一条可行的途径.     

基于贝叶斯网络的智能入侵检测模型的设计与实现

传统的入侵检测模型不能有效地检测未知入侵行为,缺乏智能性、自适应性.本文提出用贝叶斯网络数据挖掘技术构造智能入侵检测模型,利用贝叶斯网络的模糊识别能力提高入侵检测模型的智能性,最后通过实验证明模型的可行性及有较高的识别率.     

基于组合神经网络的聚合物质量预测

介绍了一种将组合神经网络用于聚合物质量预测的方法．由定量数据建立的单一神经网络模型往往缺乏泛化能力，而使用组合神经网络模型则可以显著改善模型的泛化能力．由于在建立组合神经网络模型过程中，合适的组合权重对模型是否具有良好预测性能是非常重要的，因此采用了岭回归方法来选择合适的组合权重．所提出的方法已成功应用于PVC颗粒特性的预测研究中。研究结果表明，与单一神经网络模型相比，组合神经网络模型具有更佳的模型预测精度和鲁棒性．     

预光滑子正则化求解的图像修复策略

由于传感器噪声或者拍摄抖动,容易导致数字图像含有噪声,所以必须对模糊图像进行修复处理,本文针对正则化模型在图像修复中还存在的抗噪性能较差的问题,提出了一种预光滑子正则化求解的图像修复策略。首先采用软阈值对正则化去噪模型进行最优化求解,然后构建基于离散小波的多重网格,然后为了得到最优正则化参数,采用预光滑子策略对其最粗层进行优化,并采用软阈值方法消除残留的高频信息。算法仿真实验结果表明,本文提出的方法在大多数噪声水平下比其它方法表现更优秀,并且计算时间明显比其它方法更少。     

一种基于BP神经网络的建筑防火检测改进方法

针对现有的建筑防火检测方法已经无法满足在实际检测中需求的问题,本文提出了一种基于建筑防火检测的改进BP神经网络模型。在传统BP神经网络的基础上,提出动态合并与删减规则,并且根据建筑防火检测的需求建立检测的指标,再根据改进BP神经网络和检测指标建立检测模型。仿真实验表明,基于建筑防火检测的改进BP神经网络模型实际操作性很强,可以应用于对建筑物的防火检测,值得推广使用。     

基于BP神经网络的参考咨询馆员素质评价模型

在建立参考咨询馆员素质评价指标体系的基础上,提出一种智能化的基于BP神经网络的参考咨询馆员素质评价方法。概述BP神经网络及其基本原理,并详述基于BP神经网络的参考咨询馆员素质评价模型的建立过程,包括神经网络结构的确定、网络训练,以及网络检验等。将该模型应用于实例检验,得到较满意的结果。     

基于集成学习的智能入侵检测系统

介绍了集成学习入侵检测系统设计的总体思路、总体结构和各模块功能,重点研究了基于遗传算法的集成学习分类引擎工作原理,通过仿真试验说明集成神经网络能克服单个神经网络的缺陷,具有高速数据处理与自学习功能.     

Input-to-state stability of impulsive inertial memristive neural networks with time-varying delayed

The property of input-to-state stability (ISS) of inertial memristor-based neural networks with impulsive effects is studied. Firstly, according to the characteristics of memristor and inertial neural networks, the inertial memristor-based neural networks are built. Secondly, based on the impulsive control theory, the average impulsive interval approach, Halanay differential inequality, Lyapunov method and comparison property, some sufficient conditions ensuring ISS of the inertial memristor-based neural networks under impulsive controller are derived. In this paper, we consider two types of impulse, stabilizing impulses and destabilizing impulses. When the inertial memristor-based neural networks are originally not ISS, by choosing a suitable lower bound of the average impulsive interval, the stabilizing impulses can be used to stabilize the inertial memristor-based neural networks. On the contrary, the inertial memristor-based neural networks are originally ISS, by restricting the upper bound of the average impulsive interval, the ISS of inertial memristor-based neural networks with destabilizing impulses can be ensured. Finally, numerical results are presented to illustrate the main results.     

GA优化神经网络思想

概述了神经网络的概念、神经网络的基本元素和BP算法的基本原理,同时在BP算法的基础上用遗传算法来优化BP算法的权值,并且改变了BP算法的传播方向。     

Deep forest

Current deep-learning models are mostly built upon neural networks, i.e. multiple layers of parameterized differentiable non-linear modules that can be trained by backpropagation. In this paper, we explore the possibility of building deep models based on non-differentiable modules such as decision trees. After a discussion about the mystery behind deep neural networks, particularly by contrasting them with shallow neural networks and traditional machine-learning techniques such as decision trees and boosting machines, we conjecture that the success of deep neural networks owes much to three characteristics, i.e. layer-by-layer processing, in-model feature transformation and sufficient model complexity. On one hand, our conjecture may offer inspiration for theoretical understanding of deep learning; on the other hand, to verify the conjecture, we propose an approach that generates deep forest holding these characteristics. This is a decision-tree ensemble approach, with fewer hyper-parameters than deep neural networks, and its model complexity can be automatically determined in a data-dependent way. Experiments show that its performance is quite robust to hyper-parameter settings, such that in most cases, even across different data from different domains, it is able to achieve excellent performance by using the same default setting. This study opens the door to deep learning based on non-differentiable modules without gradient-based adjustment, and exhibits the possibility of constructing deep models without backpropagation.     

混沌时间序列混合预测方法探索

根据Kolmogorov连续性定理,本文建立了混沌—神经网络(C-ANN)预测模型;提出了基于遗传算法和神经网络的混沌预测模型与方法(C-ANN-GA混合预测方法);解决了混沌时间序列的非解析式预测问题;使混沌时间序列预测方法得到了新的改进和发展。     

Quasi-synchronization of coupled neural networks with reaction-diffusion terms driven by fractional brownian motion

This paper investigates the quasi-synchronization of reaction-diffusion neural networks with hybrid coupling and parameter mismatches via sampled-data control technology. First, the models of neural networks with switching parameter and fraction Brownian motion are given. As a result of parameter mismatches, synchronization is normally not possible to realize directly, then the improved Halanay’s inequality is introduced, which is an important lemma to prove that the considered networks realize quasi-synchronization. Furthermore, based on stochastic theory, Lyapunov function method and inequality techniques, some sufficient conditions are derived to guarantee the quasi-synchronization of hybrid coupled neural networks with reaction-diffusion terms driven by fractional Brownian motion. Finally, two simulation examples are given to prove the efficiency of the developed criteria.     

Neural networks-based command filtering control for a table-mount experimental helicopter

This paper presents neural networks based on command filtering control method for a table-mount experimental helicopter which has three rotational degrees-of-freedom. First, the controller is designed based on backstepping technique, and further command filtering technique is used to solve the derivative of the virtual control, thereby avoiding the effects of signal noise. Secondly, the model uncertainty of the table-mount experimental helicopter’s system is estimated by using neural networks. And then, Lyapunov stabilization analysis proves the stability of the table-mount experimental helicopter closed-loop attitude tracking system. Finally, the experiment is carried out to clarify the effectiveness of the proposed method.     

Pinning impulsive stabilization for BAM reaction-diffusion neural networks with mixed delays

The exponential stabilization of BAM reaction-diffusion neural networks with mixed delays is discussed in this article. At first, a general pinning impulsive controller is introduced, in which the control functions are nonlinear and the pinning neurons are determined by reordering the state error. Next, based on the designed control protocol and the Lyapunov–Krasovskii functional approach, some novel and useful criteria, which depend on the diffusion coefficients and controlling parameters, are established to guarantee the global exponential stabilization of the considered neural networks. Finally, the effectiveness of the proposed control strategy is shown by two numerical examples.