Multi-step Learning Rule for Recurrent Neural Models: An Application to Time Series Forecasting

Multi-step prediction is a difficult task that has attracted increasing interest in recent years. It tries to achieve predictions several steps ahead into the future starting from current information. The interest in this work is the development of nonlinear neural models for the purpose of building multi-step time series prediction schemes. In that context, the most popular neural models are based on the traditional feedforward neural networks. However, this kind of model may present some disadvantages when a long-term prediction problem is formulated because they are trained to predict only the next sampling time. In this paper, a neural model based on a partially recurrent neural network is proposed as a better alternative. For the recurrent model, a learning phase with the purpose of long-term prediction is imposed, which allows to obtain better predictions of time series in the future. In order to validate the performance of the recurrent neural model to predict the dynamic behaviour of the series in the future, three different data time series have been used as study cases. An artificial data time series, the logistic map, and two real time series, sunspots and laser data. Models based on feedforward neural networks have also been used and compared against the proposed model. The results suggest than the recurrent model can help in improving the prediction accuracy.     

基于小波网络和多模块网络的数字识别

本文研究一种新的数字识别方法,这种方法用小波神经网络抽取特征、用多模块结构神经网络作模式分类器。小波分解的函数近似能力和人工神经网络的学习能力结合起来形成的小波神经网络,有着良好的特征描述性能,可用作特征抽取工具。多模块结构的神经网络将一个k类的模式分类问题转换为k个互相独立的2类分类问题。这种结构将一个复杂的分类问题化解为多个简单的分类问题,各个模块互相并联,各自负责一种模式的识别。用这种修改过的多模块结构网络的BP训练方法,可加速训练和提高训练精度,并且各模块可互相独立地进行训练。用美国NIST数字样本进行训练及测试,结果良好。这种方法可用于更广泛的平面图形识别。     

Neural modelling and control of a Diesel engine with pollution constraints

The paper describes a neural approach for modelling and control of a turbocharged Diesel engine. A neural model, whose structure is mainly based on some physical equations describing the engine behaviour, is built for the rotation speed and the exhaust gas opacity. The model is composed of three interconnected neural submodels, each of them constituting a nonlinear multi-input single-output error model. The structural identification and the parameter estimation from data gathered on a real engine are described. The neural direct model is then used to determine a neural controller of the engine, in a specialized training scheme minimising a multivariable criterion. Simulations show the effect of the pollution constraint weighting on a trajectory tracking of the engine speed. Neural networks, which are flexible and parsimonious nonlinear black-box models, with universal approximation capabilities, can accurately describe or control complex nonlinear systems, with little a priori theoretical knowledge. The presented work extends optimal neuro-control to the multivariable case and shows the flexibility of neural optimisers. Considering the preliminary results, it appears that neural networks can be used as embedded models for engine control, to satisfy the more and more restricting pollutant emission legislation. Particularly, they are able to model nonlinear dynamics and outperform during transients the control schemes based on static mappings.     

神经网络求解偏微分方程的分布式算法

本文提出了一种人工神经网络(Artificial Neural Networks,即ANN)求解偏微分方程的分布式的算法。文章介绍了这种算法的必要性,以分布式的人工神经网络完成了对热传导方程的求解为例,验证了这种算法的可行性,并给出了不同条件下的实验结果。     

Device-independent Color Correction for Multimedia Applications Using Neural Networks and Abductive Modeling Approaches

Although color has appeal for developers and consumers alike, color reproduction poses a major problem in many computer based applications including multimedia and desktop publishing. The problem arises because of the device-independence of color, and the way each device processes color. Matching the appearance of monitor and print images, and achieving satisfactory results is complex. Not only are there fundamental differences between computer screen (additive) and printers (subtractive), but subtractive color is in general more prone to errors due to dye inadequacies. In order to control the error in porting color, different techniques have been applied. In this paper, the utilization of artificial neural networks as well as abductive modeling approaches to color error reduction are introduced from an RGB (Red Green Blue) color model perspective. Analysis of the results and on-going research issues are discussed.     

一个基于神经网络模型的鲁棒自适应控制算法

对于具有不确定因素的离散非线性动态系统,通过校正神经网络预报器的输出,运用加权预报控制性能指标和网络辨识器模型局部线性化的思想,提出了一个间接鲁棒自适应神经网络控制算法,仿真研究证实了该控制策略的鲁棒性和有效性.     

MIGA, A Software Tool for Nonlinear System Modelling with Modular Neural Networks

This paper presents a software tool suitable for dynamic system modelling. The models generated by this tool are modular neural networks, see [1]. Each module behaves like a functional block and is connected to the other modules like in classical block diagrams. This tool allows the inclusion of a priori knowledge and, furthermore, to extract physical information from the models, once the system has learned. The modelling tool is capable of automatic model generation, parameter estimation and model validation.     

基于目标分解理论的全极化SAR图像神经网络分类方法

由于全极化合成孔径雷达(synthetic aperture radar)能够测量每一观测目标的全散射矩阵，即可合成包括线性极化、圆极化及椭圆极化在内的多种极化图像，因此与常规的单极化和多极化SAR相比，在雷达目标探测、识别，纹理特征和几何参数的提取等方面，全极化SAR均具有很多优点，但是由于地物分布的复杂性往往造成不同地物具有相似的后向散射信号特征，因而加大了地物信息提取的难度。同时由于这些极化合成图像具有较高的相关性，从而导致了图像分类精度的降低。为了提高全极化SAR图像的分类精度，基于新疆和田地区的SIR-CL波段全极化雷达数据，利用目标分解理论首先将地物回波的复杂散射过程分解为几种互不相关的单一的散射分量。由于这些单一的散射分量都对应于具有不同物理和几何特征以及分布特征的地物，从而提供了更加丰富的地表覆盖信息，这样就很大程度地改善了地物信息的分类精度；然后利用分解后单一散射分量数据结合传统的极化合成数据，可以得到更多的互不相关的数据源，再使用神经网络分类法对这些数据进行分类。分类结果表明，这种方法大幅度提高了全极化SAR数据用于实验区土地覆盖分类的精度。这种分类方法也可以广泛地用于SAR数据地表覆盖和土地利用动态监测和地表参数的提取。     

基于深度神经网络的语音驱动发音器官的运动合成

实现一种基于深度神经网络的语音驱动发音器官运动合成的方法,并应用于语音驱动虚拟说话人动画合成. 通过深度神经网络(Deep neural networks, DNN)学习声学特征与发音器官位置信息之间的映射关系,系统根据输入的语音数据估计发音器官的运动轨迹,并将其体现在一个三维虚拟人上面. 首先,在一系列参数下对比人工神经网络(Artificial neural network, ANN)和DNN的实验结果,得到最优网络; 其次,设置不同上下文声学特征长度并调整隐层单元数,获取最佳长度; 最后,选取最优网络结构,由DNN 输出的发音器官运动轨迹信息控制发音器官运动合成,实现虚拟人动画. 实验证明,本文所实现的动画合成方法高效逼真.     

模糊系统和神经网络的特征与比较

概述了模糊、神经网络 和人工智能技术之间的关系，尤其探讨了模糊系统和神经网络的特性；指出了模糊系统和神经网络的结合方式，分析了它们的特征。     

Computer Generated Holograms for Optical Neural Networks

While numerous artificial neural network (ANN) models have been electronically implemented and simulated by conventional computers, optical technology provides a far superior mechanism for the implementation of large-scale ANNs. The properties of light make it an ideal carrier of data signals. With optics, very large and high speed neural network architectures are possible. Because light is a predictable phenomenon, it can be described mathematically and its behavior can be simulated by conventional computers. A hologram is in essence a capture of the light field at a particular moment in time and space. Later, the hologram can be used to reconstruct the three dimensional light field carrying optical data. This makes a hologram an ideal medium for capturing, storing, and transmitting data in optical computers, such as optical neural networks (ONNs). Holograms can be created using conventional methods, but they can also be computer generated. In this paper, we will present an overview of optical neural networks, with emphasis on the holographic neural networks. We will take a look at the mathematical basis of holography in terms of the Fresnel Zone Plate and how it can be utilized in making computer generated holograms (CGHs). Finally, we will present various methods of CGH implementation in a two layer holographic ONN.     

大数据分析的无限深度神经网络方法

深度神经网络(deep neural networks, DNNs)及其学习算法,作为成功的大数据分析方法,已为学术界和工业界所熟知.与传统方法相比,深度学习方法以数据驱动、能自动地从数据中提取特征(知识),对于分析非结构化、模式不明多变、跨领域的大数据具有显著优势.目前,在大数据分析中使用的深度神经网络主要是前馈神经网络(feedforward neural networks, FNNs),这种网络擅长提取静态数据的相关关系,适用于基于分类的数据应用场景.但是受到自身结构本质的限制,它提取数据时序特征的能力有限.无限深度神经网络(infinite deep neural networks)是一种具有反馈连接的回复式神经网络(recurrent neural networks, RNNs),本质上是一个动力学系统,网络状态随时间演化是这种网络的本质属性,它耦合了“时间参数”,更加适用于提取数据的时序特征,从而进行大数据的预测.将这种网络的反馈结构在时间维度展开,随着时间的运行,这种网络可以“无限深”,故称之为无限深度神经网络.重点介绍这种网络的拓扑结构和若干学习算法及其在语音识别和图像理解领域的成功实例.     

随机形状变形生成的自适应神经网络控制法

将任意给定的确定性规则形状作为自适应神经网络的学习对象，用自适应学习的方法改变规则形状的形成规则，在没有加入随机移位扰动量的情况下，使规则形状从“规则”向“非规则”转变，改进了传统自适应神经网络的算法过程，增强了随机形状的局部和整体形态特征的可控性，拓宽了自适应线性神经网络在随机形状造型中的应用范围，且方法简明、易于实现。     

基于神经网络的传质系数模型建立及应用

采用神经网络解决传质系数模型建立的问题。在理论分析和实验的基础上，将其用于壁面散湿量的测试。结果表明本文建立的传质系数模型是可行的，具有较高的精度，同是实现了自动化测量。     

Part‐Based Mesh Segmentation: A Survey

This paper surveys mesh segmentation techniques and algorithms, with a focus on part‐based segmentation, that is, segmentation that divides a mesh (featuring a 3D object) into meaningful parts. Part‐based segmentation applies to a single object and also to a family of objects (i.e. co‐segmentation). However, we shall not address here chart‐based segmentation, though some mesh co‐segmentation methods employ such chart‐based segmentation in the initial step of their pipeline. Finally, the taxonomy proposed in this paper is new in the sense that one classifies each segmentation algorithm regarding the dimension (i.e. 1D, 2D and 3D) of the representation of object parts. The leading idea behind this survey is to identify the properties and limitations of the state‐of‐the‐art algorithms to shed light on the challenges for future work.     

一种求解TSP问题的新型人工神经网络方法

该文基于输出一阈值耦合神经网络的自动波现象,提出了一种用自动波方法求解TSP问题的方法。该方法具有鲁棒性和可靠性好、大规模并行计算等特点,可用于求解对称、非对称赋权图的TSP问题。与目前其它求解TSP问题的方法相比,自动波方法执行更为简单,不需要太多人为的选择参数等问题,且不存在局部极小点的问题,求得的解全部是最优解。其所需的计算量(迭代次数)主要取决于最短回路的长度,而与图的复杂程度、所存在的通路总数关系不大。最后文中给出了TSP求解的例子。     

基于有序神经网络的神经模糊预测控制

本文使用有序神经网络和改进的模糊控制器构成了一种新型的神经模糊预测控制方法，有序网络学习速度快，所需神经数目少，用事先训练好的有序网络代替传统的预测模型，以期增强输出预测的准确性；同时，用一种改进的模糊控制器原有的PID控制器，增强系统的鲁棒性。仿真结果表明，所提出的神经模糊预测控制方法可以获得理想的控制效果。     

神经网络集成

神经网络集成通过训练多个神经网络并将成结论进行合成，可以显著地提高学习系统的泛化能力。它不仅有助于科学家对机器学习和神经的深入研究，还有助于普通工程技术人员利用神经网络技术来解决真实世界中的问题。因此，它被视为一种广阔应用前景的工程化神经计算技术，已经成为机器学习和神经计算领域的研究热点。该文从实现方法、理论分析和应用成果等三个方面综述了神经网络集成的国际研究现状，并对该领域值得进一步研究的一些问题进行了讨论。     

Neural Net Based Hybrid Modeling of the Methanol Synthesis Process

A Hybrid modeling approach, combining an analytical model with a radial basis function neural network is introduced in this paper. The modeling procedure is combined with genetic algorithm based feature selection designed to select informative variables from the set of available measurements. By only using informative inputs, the model's generalization ability can be enhanced. The approach proposed is applied to modeling of the liquid–phase methanol synthesis. It is shown that a hybrid modeling approach exploiting available a priori knowledge and experimental data can considerably outperform a purely analytical approach.     

Artificial Neural Networks Manipulation Server: Research on the Integration of Databases and Artificial Neural Networks

This paper proposes a new whole and distributed integration approach between Artificial Neural Networks (ANNs) and Databases (DBs) taking into account the different stages of the former’s lifecycle (training, test and running). The integration architecture which has been developed consists of an ANN Manipulation Server (AMS) based on a client-server approach, which improves the ANNs’ manipulation and experimentation capabilities considerably, and also those of their training and test sets, together with their modular reuse among possibly remote applications. Moreover, the chances of integrating ANNs and DBs are analysed, proposing a new level of integration which improves the integration features considerably. This level has not been contemplated yet at full reach in any of the commercial or experimental tools analysed up to the present date. Finally, the application of the integration architecture which has been developed to the specific domain of Environmental Impact Assessments (EIAs) is studied. Thus, the versatility and efficacy of that architecture for developing ANNs is tested. The enormous complexity of the functioning of the patterns which rule the environment’s behaviour, and the great number of variables involved, make it the ideal domain for experimenting on the application of ANNs together with DBs.