一种新型的动态递归神经网络及其全自动设计算法

本文基于非线形自回归滑动平均模型NARMA模 型和前馈神经网络建模的思想,提出一种输入层与输出层神经元递归的动态递归神经网络; 基于进化计算中遗传算法和进化策略与自寻优BP算法的不同结合方式,提出两种动态递归神 经网络全自动高效设计算法,实现了网络结构、权重和自反馈增益同时优化学习,实例应用 表明所提网络结构及其设计算法的有效性．     

基于模糊递归小波神经网络的葡萄酒品质预测

针对葡萄酒品质预测模型难以建立的问题,提出一种基于模糊递归小波神经网络的葡萄酒品质预测模型。利用葡萄酒物理化学指标和品酒师打分作为模型的输入输出,采用梯度下降算法在线学习隶属函数层中心、宽度和小波函数平移因子、伸缩因子、自反馈权重因子以及输出层权值。仿真实验时,首先利用Mackey-Glass混沌时间序列进行了性能测试,然后利用UCI数据集葡萄酒品质数据对所建立的品质预测模型进行了验证。结果显示,与多层感知器、径向基函数神经网络等传统前馈神经网络相比,构建的模糊递归小波神经网络品质预测模型具有更高的预测精度,更加适合于葡萄酒的品质预测。     

动态模糊神经网络控制器在伺服系统中的应用

通过在ANFIS的归一化层与输出层之间加入递归,层提出了一种新型的动态模糊神经网络（DFNN）,将模糊推理系统、神经网络和Ⅲ型控制有机地结合起来。给出DFNN的网络结构,为基于收缩间距隶属函数和BP算法提供了参数调整方法。系统实验表明,DFNN控制器比PID＋前馈控制具有更好的动、静态响应,尤其在前馈信号难以取得的情况下具有更明显的优势。     

基于跨越连接的多层前馈神经网络结构分析

任何连接方式的神经网络总可以归结为跨越连接网络。在传统多层前馈神经网络算法的基础上,提出了完全全连接神经网络的概念,给出了基于跨越连接的多层前馈神经网络算法。通过分析多层前馈神经网络的误差函数,从理论上证明了：相对于无跨越连接网络,基于跨越连接的多层前馈神经网络能以更加简洁的结构逼近理想状态。最后,用一个隐层神经元解决了XOR问题。     

改进递归最小二乘RBF神经网络溶解氧预测

为提高溶解氧预测的准确性,将基于改进型递归最小二乘算法优化的径向基函数（ RBF）神经网络方法应用于溶解氧预测。利用K均值聚类算法进行隐层单元中心选择;利用改进型递归最小二乘算法优化RBF神经网络隐含层到输出层的权值。仿真结果表明：该方法对溶解氧的预测具有较好的非线性拟合能力,预测精度优于RBF神经网络和递归最小二乘算法优化的RBF神经网络。     

多层随机神经网络em算法

本文讨论了基于微分流形框架随机神经网络学习算法，称为ｅｍ学习算法；对于多层随机神经网络模型，我们从微分流形的角度分析它的对偶平坦流形结构，描述ｅｍ算法对于多层前馈随机神经网络模型学习算法实现和加速技术。     

基于自组织递归模糊神经网络的BOD软测量

生化需氧量是污水处理过程中评价水质的重要指标之一,神经网络软测量是解决其在线测量困难的主要方法。污水处理是一个动态的过程,而前馈神经网络由于缺乏动态性而难以保证对其的测量精度。本文提出了一种自组织递归模糊神经网络,建立了内部的反馈连接以增强网络动态性能,通过评估神经元的互信息关系和激活强度以增长或修剪规则层神经元,采用梯度下降学习算法进行参数更新,并结合自适应学习率以提高收敛精度。通过对实际污水厂数据的实验结果表明,本文提出的模型结构更紧凑,对出水生化需氧量的预测精度更高。     

基于神经网络监督控制的拥塞控制算法研究

提出了一个基于神经网络控制的主动队列管理（AQM）算法;研究了TCP/AQM拥塞控制系统的可逆性,并利用一种神经网络监督控制结构进行了AQM算法的设计。算法由一个三层前馈结构的神经网络控制器（neural network controller,NNC）和一个反馈控制器（feedback controller,FC）组成。NNC作为一个前馈控制器,通过FC产生的教师信号进行学习,以建立被控对象的逆动力学模型。仿真结果表明,提出的算法与PI（proportional-integral）算法相比,无论在瞬态性能     

基于Zernike矩和前馈神经网络的图像配准

提出一种基于Zernike矩和多级前馈神经网络的图像配准算法。利用低阶Zernike矩表征图像的全局几何特征,通过多级前馈神经网络学习图像所经历的旋转、缩放和平移等仿射变换参数,在一级前馈神经网络的基础上添加二级前馈网络,以提高参数估计精度。仿真结果表明,与基于DCT系数的神经网络算法相比,该算法旋转、缩放和平移估计精度较高,对噪声的鲁棒性较强。     

改进的新BRNN网络集成及其二级结构预测的研究

为提高二级结构预测精度,试用神经网络集成法预测.针对BRNN网络结构复杂、收敛时间长、参数多的缺点,本文提出一种改进的新BRNN网络,删除BRNN左、右子网络的隐层,直接将输入连接到状态层,并采用BP改进算法中的弹性算法训练.以90条蛋白质序列共15 377个氨基酸交叉验证,仿真结果表明新网络可以有效地缩短收敛时间,新BRNN集成预测二级结构效果较好.     

Towards designing modular recurrent neural networks in learning protein secondary structures

Precise prediction of protein secondary structures from the associated amino acids sequence is of great importance in bioinformatics and yet a challenging task for machine learning algorithms. As a major step toward predicting the ultimate three dimensional structures, the secondary structure assignment specifies the protein function. Considering a multilayer perceptron neural network, pruned for optimum size of hidden layers, as the reference network, advanced kinds of recurrent neural network (RNN) are devised in this article to enhance the secondary structure prediction. To better model the strong correlations between secondary structure elements, types of modular reciprocal recurrent neural networks (MRR-NN) are examined. Additionally, to take into account the long-range interactions between amino acids in formation of the secondary structure, bidirectional RNN are investigated. A multilayer bidirectional recurrent neural network (MBR-NN) is finally applied to capture the predominant long-term dependencies. Eventually, a modular prediction system based on the interactive combination of the MRR-NN and MBR-NN boosts the percentage accuracy (Q₃) up to 76.91% and augments the segment overlap (SOV) up to 68.13% when tested on the PSIPRED dataset. The coupling effects of the secondary structure types as well as the sequential information of amino acids along the protein chain can be well cast by the integration of the MRR-NN and the MBR-NN.     

An enhanced protein secondary structure prediction using deep learning framework on hybrid profile based features

Accurate protein secondary structure prediction (PSSP) is essential to identify structural classes, protein folds, and its tertiary structure. To identify the secondary structure, experimental methods exhibit higher precision with the trade-off of high cost and time. In this study, we propose an effective prediction model which consists of hybrid features of 42-dimensions with the combination of convolutional neural network (CNN) and bidirectional recurrent neural network (BRNN). The proposed model is accessed on four benchmark datasets such as CB6133, CB513, CASP10, and CAP11 using Q3, Q8, and segment overlap (Sov) metrics. The proposed model reported Q3 accuracy of 85.4%, 85.4%, 83.7%, 81.5%, and Q8 accuracy 75.8%, 73.5%, 72.2%, and 70% on CB6133, CB513, CASP10, and CAP11 datasets respectively. The results of the proposed model are improved by a minimum factor of 2.5% and 2.1% in Q3 and Q8 accuracy respectively, as compared to the popular existing models on CB513 dataset. Further, the quality of the Q3 results is validated by structural class prediction and compared with PSI-PRED. The experiment showed that the quality of the Q3 results of the proposed model is higher than that of PSI-PRED.     

双向循环神经网络在语音识别中的应用

针对前馈神经网络难以处理时序数据的问题,提出将双向循环神经网络（BiRNN）应用在自动语音识别声学建模中。首先,应用梅尔频率倒谱系数进行特征提取;其次,采用双向循环神经网络作为声学模型;最后,测试不同参数对系统性能的影响。在TIMIT数据集上的实验结果表明,与基于卷积神经网络和深度神经网络的声学模型相比,识别率分别提升了1.3%和4.0%,说明基于双向循环神经网络的声学模型具有更好的性能。     

Bidirectional segmented-memory recurrent neural network for protein secondary structure prediction

The formation of protein secondary structure especially the regions of β-sheets involves long-range interactions between amino acids. We propose a novel recurrent neural network architecture called segmented-memory recurrent neural network (SMRNN) and present experimental results showing that SMRNN outperforms conventional recurrent neural networks on long-term dependency problems. In order to capture long-term dependencies in protein sequences for secondary structure prediction, we develop a predictor based on bidirectional segmented-memory recurrent neural network (BSMRNN), which is a noncausal generalization of SMRNN. In comparison with the existing predictor based on bidirectional recurrent neural network (BRNN), the BSMRNN predictor can improve prediction performance especially the recognition accuracy of β-sheets.     

基于时域建模的自动语音识别

端到端神经网络能够根据特定的任务自动学习从原始数据到特征的变换,解决人工设计的特征与任务不匹配的问题。以往语音识别的端到端网络采用一层时域卷积网络作为特征提取模型,递归神经网络和全连接前馈深度神经网络作为声学模型的方式,在效果和效率两个方面具有一定的局限性。从特征提取模块的效果以及声学模型的训练效率角度,提出多时间频率分辨率卷积网络与带记忆模块的前馈神经网络相结合的端到端语音识别模型。实验结果表明,所提方法语音识别在真实录制数据集上较传统方法字错误率下降10%,训练时间减少80%。     

结合蛋白质二级结构信息预测蛋白质空间结构中的二硫键*

在蛋白质空间结构预测中,二硫键的确定可以大大减少蛋白质构象的搜索空间。为提高二硫键预测的准确率,对形成二硫键的半胱氨酸及其周围的氨基酸残基在蛋白质二级结构形成上的偏性进行了分析,并提出将蛋白质二级结构信息加入到BP神经网络预测模型的输入编码信息中。研究对象为从SWISS-PROT数据库中选取的252条蛋白质序列,随机均分4组,对预测准确率进行4-交叉验证。各项准确率均比未加入蛋白质二级结构信息前,有明显提高。结果表明,结合蛋白质二级结构信息的编码方式是可行且有效的。     

A multimodal neural network with single-state predictions for protein secondary structure

Prediction of protein secondary structure is considered to be an important step toward elucidating the three-dimensional structure and function of proteins. We have developed a multimodal neural network (MNN) to predict protein secondary structure. The MNN is composed of several subclassifiers for single-state predictions using neural networks and a decision neural network (DNN). Each subclassifier employs a number of subnetworks to predict the single-state of the secondary structure individually and produces the final results by majority decision. The DNN uses a three-layer neural network to produce the final overall prediction from the outputs of the single-state predictions. The MNN gives an overall accuracy of 71.1% with corresponding Matthews correlation coefficients of C_H = 0.62 and C_E = 0.53. The prediction test is based on a database of 126 nonhomologous protein sequences. This work was presented, in part, at the 8th International Symposium on Artificial Life and Robotics, Oita, Japan, January 24#x2013;26, 2003.     

A comparative study of neural network and Box-Jenkins ARIMA modeling in time series prediction

This paper aims to investigate suitable time series models for repairable system failure analysis. A comparative study of the Box-Jenkins autoregressive integrated moving average (ARIMA) models and the artificial neural network models in predicting failures are carried out. The neural network architectures evaluated are the multi-layer feed-forward network and the recurrent network. Simulation results on a set of compressor failures showed that in modeling the stochastic nature of reliability data, both the ARIMA and the recurrent neural network (RNN) models outperform the feed-forward model; in terms of lower predictive errors and higher percentage of correct reversal detection. However, both models perform better with short term forecasting. The effect of varying the damped feedback weights in the recurrent net is also investigated and it was found that RNN at the optimal weighting factor gives satisfactory performances compared to the ARIMA model.     

蛋白质二级结构预测方法研究

为提高蛋白质二级结构预测精度,提出一种新的网络模型和编码方法。首先利用基因表达式编程（GEP）的全局搜索能力同时进化设计神经网络的结构和连接权;其次,对神经网络输入层编码进行了改进,添加了氨基酸残基所处的疏水环境。用PDBSelect25中的36条蛋白质共6 122个残基进行测试,结果表明提出的网络模型和编码方法能有效提高蛋白质二级结构预测的精度。     

基于深度残差网络的自动调制识别方法研究

自动调制识别是电磁环境特性分析的关键问题,而传统方法多基于人工设计特征进行识别,数据特征表示和判别分析能力有限。为此提出一种新颖的深度神经网络特征表示方法进行调制识别任务。首先,利用递归神经网络结构对电磁信号序列进行表示,建立了基于多层双向GRU网络结构的识别方法。其次,从一维空间卷积表示序列的角度思考,建立了基于深度残差卷积网络的调制识别方法。最后,针对加性高斯白噪声信道的调制方式仿真数据集,将提出的方法与典型神经网络模型如多层感知器、卷积神经网络进行了对比实验。实验结果表明,提出的方法在自动调制识别方面具备更强的特征表示能力和竞争力,有利于推动深度学习在自动调制识别领域的应用。