基于SCFNN之PAM非线性信道均衡器成效研究

自组织型模糊类神经网络(SCFNN)可依据一定的法则自我构建神经网络的组织结构,从而适用于当前控制对象;多层神经元是传统的类神经网络,广泛应用于各个领域;倒传递学习法与最陡坡降法相结合,可使以上两种类神经网络进行有效的融合;目前,信道均衡器上的系统架构种类非常多,各种类神经网络应用于信道均衡器也颇为普遍;在研究SCFNN的基础上,将其应用于通道均衡器确实可行,效果良好;比较了SCFNN与MLP在通道均衡器的成效;仿真表明,在相同通道环境下,SCFNN的训练收敛速度、位错误率与系统敏感度优于MLP,完成结构学习后SCFNN的结构也颇为精简。     

Resolution of pattern recognition problems using a hybrid Genetic/Random Neural Network learning algorithm

Gelenbe has proposed a neural network, called a Random Neural Network, which calculates the probability of activation of the neurons in the network. In this paper, we propose to solve the patterns recognition problem using a hybrid Genetic/Random Neural Network learning algorithm. The hybrid algorithm trains the Random Neural Network by integrating a genetic algorithm with the gradient descent rule-based learning algorithm of the Random Neural Network. This hybrid learning algorithm optimises the Random Neural Network on the basis of its topology and its weights distribution. We apply the hybrid Genetic/Random Neural Network learning algorithm to two pattern recognition problems. The first one recognises or categorises alphabetic characters, and the second recognises geometric figures. We show that this model can efficiently work as associative memory. We can recognise pattern arbitrary images with this algorithm, but the processing time increases rapidly.     

Efficient MLP constructive training algorithm using a neuron recruiting approach for isolated word recognition system

This paper describes an efficient constructive training algorithm using a Multi Layer Perceptron (MLP) neural network dedicated for Isolated Word Recognition (IWR) systems. Incremental training procedure was employed and this approach was based on novel hidden neurons recruiting for a single hidden-layer. During Neural Network (NN) training phase, the number of pronunciation samples extracted from the Training Data (TD) was sequentially increased. Optimal structure of the NN classifier with optimized TD size was obtained using this proposed MLP constructive training algorithm.     

Normal and hypoacoustic infant cry signal classification using time-frequency analysis and general regression neural network

Crying is the most noticeable behavior of infancy. Infant cry signals can be used to identify physical or psychological status of an infant. Recently, acoustic analysis of infant cry signal has shown promising results and it has been proven to be an excellent tool to investigate the pathological status of an infant. This paper proposes short-time Fourier transform (STFT) based time-frequency analysis of infant cry signals. Few statistical features are derived from the time-frequency plot of infant cry signals and used as features to quantify infant cry signals. General Regression Neural Network (GRNN) is employed as a classifier for discriminating infant cry signals. Two classes of infant cry signals are considered such as normal cry signals and pathological cry signals from deaf infants. To prove the reliability of the proposed features, two neural network models such as Multilayer Perceptron (MLP) and Time-Delay Neural Network (TDNN) trained by scaled conjugate gradient algorithm are also used as classifiers. The experimental results show that the GRNN classifier gives very promising classification accuracy compared to MLP and TDNN and the proposed method can effectively classify normal and pathological infant cries.     

不确定混沌系统的差分进化小波神经网络控制

提出一种基于小生境自适应差分进化小波神经网络（NADE-WNN）的方法对不确定混沌系统进行控制。该方法利用小波神经网络学习未知模型混沌系统的动态特性并实施控制,为提高神经网络的学习精度和收敛速度,采用小生境自适应差分进化算法同时优化小波神经网络的结构和参数,简化网络结构,提高网络的学习精度和全局收敛性。仿真实验结果表明,在有外部干扰和参数摄动的情况下,NADE-WNN仍能对不确定混沌系统进行有效控制,且网络结构、控制精度和收敛速度都优于传统神经网络。     

神经网络在软件失效预测中的性能比较和研究

在软件开发的早期预测有失效倾向的软件模块,能够极大地提高软件的质量.软件失效预测中的一个普遍问题是数据中噪声的存在.神经网络具有鲁棒性而且对噪声有很强的抑制能力.不同结构的神经网络在训练算法和应用领域都有差异.该文主要就软件失效预测这个应用领域叙述几种适用的网络,并比较这几种网络在训练结果和性能上的差异.上述方法在SDH通信软件的失效预测中得到了成功的应用.试验结果显示虽然MLP、PNN、LVQ网络都能解决这类模式分类问题,但是只有MLP网络训练结果比较稳定,在不同的数据集上训练出的网络都有很好的预测效果.     

基于聚类关联网络的群组行为识别

目前群组行为识别方法没有充分利用群组关联信息而导致群组识别精度无法有效提升,针对这个问题,提出了基于近邻传播算法（AP）的层次关联模块的深度神经网络模型,命名为聚类关联网络（CRN）。首先,利用卷积神经网络（CNN）提取场景特征,再利用区域特征聚集提取场景中的人物特征。然后,利用AP的层次关联网络模块提取群组关联信息。最后,利用长短期记忆网络（LSTM）融合个体特征序列与群组关联信息,并得到最终的群组识别结果。与多流卷积神经网络（MSCNN）方法相比,CRN方法在Volleyball数据集与Collective Activity数据集上的识别准确率分别提升了5.39与3.33个百分点。与置信度能量循环网络（CERN）方法相比,CRN方法在Volleyball数据集与Collective Activity数据集上的识别准确率分别提升了8.7与3.14个百分点。实验结果表明,CRN方法在群体行为识别任务中拥有更高的识别准确精度。     

基于聚类关联网络的群组行为识别

目前群组行为识别方法没有充分利用群组关联信息而导致群组识别精度无法有效提升，针对这个问题，提出了基于近邻传播算法（AP）的层次关联模块的深度神经网络模型，命名为聚类关联网络（CRN）。首先，利用卷积神经网络（CNN）提取场景特征，再利用区域特征聚集提取场景中的人物特征。然后，利用AP的层次关联网络模块提取群组关联信息。最后，利用长短期记忆网络（LSTM）融合个体特征序列与群组关联信息，并得到最终的群组识别结果。与多流卷积神经网络（MSCNN）方法相比，CRN方法在Volleyball数据集与Collective Activity数据集上的识别准确率分别提升了5.39与3.33个百分点。与置信度能量循环网络（CERN）方法相比，CRN方法在Volleyball数据集与Collective Activity数据集上的识别准确率分别提升了8.7与3.14个百分点。实验结果表明，CRN方法在群体行为识别任务中拥有更高的识别准确精度。     

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。     

基于集成学习方法的实体关系抽取

基于神经网络的实体关系抽取模型已经被证明了它的有效性, 但使用单一的神经网络模型在不同的输入条件下, 会表现出不同的结果, 性能不太稳定. 因此本文提出一种利用集成学习思想将多个单一模型集成为一个综合模型的方法. 该方法主要使用MLP (MultiLayer Perceptron)将两个单一模型Bi-LSTM (Bi-directional Long Short-Term Memory)和CNN (Convolutional Neural Network)集成为一个综合模型, 该模型不仅可以充分利用两个单一模型的优势, 而且可以利用MLP的自学习能力与自动分配权重的优势. 本研究在SemEval 2010 Task 8数据集上取得了87.7％的F1值, 该结果优于其他主流的实体关系抽取模型.     

基于尺度自适应卷积神经网络的人群计数算法

为解决单幅图像中的人群遮挡和尺度变化问题,提出一种基于多列卷积神经网络的人群计数算法。利用具有不同尺寸感受野的卷积神经网络(CNN)和特征注意力模块自适应提取多尺度人群特征,引入可变形卷积增强CNN网络空间几何形变学习能力并优化特征图,从而生成高质量的密度图。Shanghai Tech和UCF_CC_50数据集上的实验结果表明,该算法能学习输入图和人群密度图之间的映射关系,且计数准确性高、鲁棒性强。     

Prediction of soil compression coefficient for urban housing project using novel integration machine learning approach of swarm intelligence and Multi-layer Perceptron Neural Network

In many engineering projects, the soil compression coefficient is an important parameter used for estimating the settlement of soil layers. The common practice of determining the soil compression coefficient via the oedometer test is time-consuming and expensive. This study proposes a machine learning solution to replace the conventional tests used for obtaining the coefficient of soil compression. The new approach is an integration of the Multi-Layer Perceptron Neural Network (MLP Neural Nets) and Particle Swarm Optimization (PSO). These two computational intelligence methods work synergistically to establish a prediction model of soil compression coefficient. The PSO metaheuristic is employed to optimize the MLP Neural Nets model structure. To train and validate the proposed method, named as PSO-MLP Neural Nets, a dataset of 154 soil samples featuring 12 influencing factors has been collected from the geotechnical investigation process of a high-rise building project. Experimental results show that the proposed PSO-MLP Neural Nets has attained the most accurate prediction of the soil compression coefficient performance with RMSE = 0.0267, MAE = 0.0145, and R² = 0.884. The result of the proposed model is significantly better than those obtained from other benchmark methods including the backpropagation neural network, the radial basis function neural network, the support vector regression, the random forest, and the Gaussian process. Based on the experimental results, the newly constructed PSO-MLP Neural Nets is very potential to be a new alternative to assist geotechnical engineers in design phase of civil engineering projects.     

基于深度卷积神经网络的图像检索算法研究

为解决卷积神经网络在提取图像特征时所造成的特征信息损失,提高图像检索的准确率,提出了一种基于改进卷积神经网络LeNet-L的图像检索算法。首先,改进LeNet-5卷积神经网络结构,增加网络结构深度。然后,对深度卷积神经网络模型LeNet-L进行预训练,得到训练好的网络模型,进而提取出图像高层语义特征。最后,通过距离函数比较待检图像与图像库的相似度,得出相似图像。在Corel数据集上,与原模型以及传统的SVM主动学习图像检索方法相比,该图像检索方法有较高的准确性。经实验结果表明,改进后的卷积神经网络具有更好的检索效果。     

Faster R-CNN模型在车辆检测中的应用

针对传统机器学习方法在车辆检测应用中易受光照、目标尺度和图像质量等因素影响,效率低下且泛化能力较差的问题,提出一种基于改进的较快的基于区域卷积神经网络（R-CNN）模型的车辆检测方法。该方法以Faster R-CNN模型为基础,通过对输入图像进行卷积和池化等操作提取车辆特征,结合多尺度训练和难负样本挖掘策略降低复杂环境的影响,利用KITTI数据集对深度神经网络模型进行训练,并采集实际场景中的图像进行测试。仿真实验中,在保证检测时间的情况下,相对原Faster R-CNN算法检测精确度提高了约8%。实验结果表明,所提方法能够自动地提取车辆特征,解决了传统方法提取特征费时费力的问题,同时提高了车辆检测精确度,具有良好的泛化能力和适用范围。     

基于有效神经元的自组织模糊神经网络算法

针对传统神经网络识别率低和泛化能力差的问题,提出了一种改进的自组织模糊神经网络（SOFNN）学习算法。以保存椭球基函数（EBF）层各个神经元的输出及输出之和为依据进行神经元的修改,删除和增加,进而得到网络的有效神经元,并减少样本训练的时间。用最小二乘法（RLSE）估计参数,用梯度下降法修改参数,保证网络收敛。与其他的模糊神经网络相比,在精确度、结构复杂性和抗干扰性方面的优越性,在真实数据集上得到了有效的验证。     

基于混合神经网络的恶意TLS流量识别研究

针对使用传统机器学习方法来识别恶意TLS流量受到专家经验的影响较大、识别与分类效果不理想的问题,提出了HNNIM（Hybrid Neural Network Identification Model）模型来进行识别与分类。模型由两层组成：第一层用于提取特征,第二层用于识别与分类。第一层中,提取的特征分为两部分,一部分特征由深度神经网络自动挖掘,另一部分特征根据专家经验选取,并由深度神经网络进一步筛选;第二层将第一层筛选出的特征进行聚合,采用全连接的深度神经网络进一步学习和拟合。通过分析大量TLS流量样本,最终选用TLS流量中的ClientHello与ServerHello消息报文与TCP协议交互信息这两部分来作为特征空间。实验的结果表明,HNNIM模型在恶意TLS流量的识别任务上关于恶意样本的F1值为0.989,较随机森林、SVM、XGBoost、卷积神经网络模型,在F1值上分别提升了0.016、0.016、0.019、0.043;在多分类任务上的平均准确率为89.28%,较随机森林、SVM、XGBoost、卷积神经网络模型分别提升了9.92%、9.09%、11.31%、7.03%。     

Extracting rules for classification problems: AIS based approach

Although Artificial Neural Network (ANN) usually reaches high classification accuracy, the obtained results in most cases may be incomprehensible. This fact is causing a serious problem in data mining applications. The rules that are derived from ANN are needed to be formed to solve this problem and various methods have been improved to extract these rules. In our previous work, a hybrid neural network was presented for classification (Kahramanli & Allahverdi, 2008). In this study a method that uses Artificial Immune Systems (AIS) algorithm has been presented to extract rules from trained hybrid neural network. The data were obtained from the University of California at Irvine (UCI) machine learning repository. The datasets are Cleveland heart disease and Hepatitis data. The proposed method achieved accuracy values 96.4% and 96.8% for Cleveland heart disease dataset and Hepatitis dataset respectively. It is been observed that these results are one of the best results comparing with results obtained from related previous studies and reported in UCI web sites.     

Learning Algorithm and Retrieval Process for the Multiple Classes Random Neural Network Model

Gelenbe has modeled neural networks using an analogy with queuing theory. This model (called Random Neural Network) calculates the probability of activation of the neurons in the network. Recently, Fourneau and Gelenbe have proposed an extension of this model, called multiple classes random neural network model. The purpose of this paper is to describe the use of the multiple classes random neural network model to learn patterns having different colors. We propose a learning algorithm for the recognition of color patterns based upon non-linear equations of the multiple classes random neural network model using gradient descent of a quadratic error function. In addition, we propose a progressive retrieval process with adaptive threshold values. The experimental evaluation shows that the learning algorithm provides good results.     

Fuzzy Clustering Using A Compensated Fuzzy Hopfield Network

Hopfield neural networks are well known for cluster analysis with an unsupervised learning scheme. This class of networks is a set of heuristic procedures that suffers from several problems such as not guaranteed convergence and output depending on the sequence of input data. In this paper, a Compensated Fuzzy Hopfield Neural Network (CFHNN) is proposed which integrates a Compensated Fuzzy C-Means (CFCM) model into the learning scheme and updating strategies of the Hopfield neural network. The CFCM, modified from Penalized Fuzzy C-Means algorithm (PFCM), is embedded into a Hopfield net to avoid the NP-hard problem and to speed up the convergence rate for the clustering procedure. The proposed network also avoids determining values for the weighting factors in the energy function. In addition, its training scheme enables the network to learn more rapidly and more effectively than FCM and PFCM. In experimental results, the CFHNN method shows promising results in comparison with FCM and PFCM methods.     

基于MLP改进型深度神经网络学习资源推荐算法

针对在线学习过程中出现的知识过载及传统推荐算法中存在的数据稀疏和冷启动问题,提出了一种基于多层感知机（MLP）的改进型深度神经网络学习资源推荐算法。该算法利用多层感知机对非线性数据处理的优势,将学习者特征和学习资源特征进行向量相乘的预测方式转换为输入多层感知机的方式,改进了DN-CBR神经网络推荐模型。为验证模型的有效性,以爱课程在线学习平台数据为样本构建数据集,通过对比实验表明,在该数据集上,改进后模型相较于DN-CBR模型在归一化折损累积增益和命中率指标上分别提升了1.2%和3%,有效地提高了模型的推荐性能。