Improved understorey bamboo cover mapping using a novel hybrid neural network and expert system

The giant panda is an obligate bamboo grazer. Therefore, the availability and abundance of understorey bamboo determines the quantity and quality of panda habitat. However, there is little or no information about the spatial distribution or abundance of bamboo underneath the forest canopy, due to the limitations of traditional remote sensing classification techniques. In this paper, a new method combines an artificial neural network and a GIS expert system in order to map understorey bamboo in the Qinling Mountains of south‐western China. Results from leaf‐off ASTER imagery, using a neural network and an expert system, were evaluated for their suitability to quantify understorey bamboo. Three density classes of understorey bamboo were mapped, first using a neural network (overall accuracy 64.7%, Kappa 0.45) and then using an expert system (overall accuracy 62.1%, Kappa 0.43). However, when using the results of the neural network classification as input into the expert system, a significantly improved mapping accuracy was achieved with an overall accuracy of 73.8% and Kappa of 0.60 (average z‐value = 3.35, p = 0.001). Our study suggests that combining a neural network with an expert system makes it possible to successfully map the cover of understorey species such as bamboo in complex forested landscapes (e.g. coniferous‐dominated and dense canopy forests), and with higher accuracy than when using either a neural network or an expert system.     

Reducing waste in casting with a predictive neural model

This paper describes an interactive neural network model that predicts the quality of cast ceramic products using multiple quantitative and qualitative inputs. This has been done to enable a major sanitary ware manufacturer to reduce product waste by better control of the slip casting process. The input variables describe the raw materials, ambient conditions and line settings for the ceramic casting process. The neural network predictive model assigns one of seven quality categories to the cast based on the input data. This prediction is used by the quality control engineer to make a priori adjustments to materials and line settings so that a good quality cast is produced without trial and error. The neural model can also be used to determine optimum settings of each adjustable input variable in the light of values of non-adjustable input variables.     

Appearance Flow Completion for Novel View Synthesis

Novel view synthesis from sparse and unstructured input views faces challenges like the difficulty with dense 3D reconstruction and large occlusion. This paper addresses these problems by estimating proper appearance flows from the target to input views to warp and blend the input views. Our method first estimates a sparse set 3D scene points using an off‐the‐shelf 3D reconstruction method and calculates sparse flows from the target to input views. Our method then performs appearance flow completion to estimate the dense flows from the corresponding sparse ones. Specifically, we design a deep fully convolutional neural network that takes sparse flows and input views as input and outputs the dense flows. Furthermore, we estimate the optical flows between input views as references to guide the estimation of dense flows between the target view and input views. Besides the dense flows, our network also estimates the masks to blend multiple warped inputs to render the target view. Experiments on the KITTI benchmark show that our method can generate high quality novel views from sparse and unstructured input views.     

Neural network‐based state prediction for strategy planning of an air hockey robot

We analyze a neural network implementation for puck state prediction in robotic air hockey. Unlike previous prediction schemes which used simple dynamic models and continuously updated an intercept state estimate, the neural network predictor uses a complex function, computed with data acquired from various puck trajectories, and makes a single, timely estimate of the final intercept state. Theoretically, the network can account for the complete dynamics of the table if all important state parameters are included as inputs, an accurate data training set of trajectories is used, and the network has an adequate number of internal nodes. To develop our neural networks, we acquired data from 1500 no‐bounce and 1500 one‐bounce puck trajectories, noting only translational state information. Analysis showed that performance of neural networks designed to predict the results of no‐bounce trajectories was better than the performance of neural networks designed for one‐bounce trajectories. Since our neural network input parameters did not include rotational puck estimates and recent work shows the importance of spin in impact analysis, we infer that adding a spin input to the neural network will increase the effectiveness of state estimates for the one‐bounce case. © 2001 John Wiley & Sons, Inc.     

Determining significance of input neurons for probabilistic neural network by sensitivity analysis procedure

In classical feedforward neural networks such as multilayer perceptron, radial basis function network, or counter‐propagation network, the neurons in the input layer correspond to features of the training patterns. The number of these features may be large, and their meaningfulness can be various. Therefore, the selection of appropriate input neurons should be regarded. The aim of this paper is to present a complete step‐by‐step algorithm for determining the significance of particular input neurons of the probabilistic neural network (PNN). It is based on the sensitivity analysis procedure applied to a trained PNN. The proposed algorithm is utilized in the task of reduction of the input layer of the considered network, which is achieved by removing appropriately indicated features from the data set. For comparison purposes, the PNN's input neuron significance is established by using the ReliefF and variable importance procedures that provide the relevance of the input features in the data set. The performance of the reduced PNN is verified against a full structure network in classification problems using real benchmark data sets from an available machine learning repository. The achieved results are also referred to the ones attained by entropy‐based algorithms. The prediction ability expressed in terms of misclassifications is obtained by means of a 10‐fold cross‐validation procedure. Received outcomes point out interesting properties of the proposed algorithm. It is shown that the efficiency determined by all tested reduction methods is comparable.     

Neural network training: Using untransformed or log‐transformed training data for the inversion of ocean colour spectra?

A bio‐optical model coupled with the radiative transfer model Hydrolight was used to create 18,000 synthetic ocean colour spectra corresponding to open ocean and coastal waters. The bio‐optical model took into account the optical properties of the three oceanic constituents, chlorophyll‐a, suspended non‐chlorophyllous particles and coloured dissolved organic matter (CDOM) as well as of normal seawater. The resulting spectra were input into multilayer perceptron neural network algorithms with the aim of computing the original concentrations of chlorophyll‐a, non‐chlorophyllous particles and CDOM initially input into the bio‐optical model. The process of training the neural networks is essential for the accuracy of the inversion the neural net performs on the coupled bio‐optical and radiative transfer models. The objective of this paper is to investigate the performance difference of a neural network trained with untransformed as opposed to logarithmically transformed data.     

PointProNets: Consolidation of Point Clouds with Convolutional Neural Networks

With the widespread use of 3D acquisition devices, there is an increasing need of consolidating captured noisy and sparse point cloud data for accurate representation of the underlying structures. There are numerous algorithms that rely on a variety of assumptions such as local smoothness to tackle this ill‐posed problem. However, such priors lead to loss of important features and geometric detail. Instead, we propose a novel data‐driven approach for point cloud consolidation via a convolutional neural network based technique. Our method takes a sparse and noisy point cloud as input, and produces a dense point cloud accurately representing the underlying surface by resolving ambiguities in geometry. The resulting point set can then be used to reconstruct accurate manifold surfaces and estimate surface properties. To achieve this, we propose a generative neural network architecture that can input and output point clouds, unlocking a powerful set of tools from the deep learning literature. We use this architecture to apply convolutional neural networks to local patches of geometry for high quality and efficient point cloud consolidation. This results in significantly more accurate surfaces, as we illustrate with a diversity of examples and comparisons to the state‐of‐the‐art.     

层次分析法和神经网络相融合的教学质量评价

为提高教学质量评价准确性,提出一种基于层次分析法和神经网络相融合的教学质量评价方法（AHP-BPNN）。采用层次分析法构建评价指标体系,筛选出对评价结果有重要影响的指标作为BP神经网络输入,采用神经网络建立教学质量评价模型。仿真结果表明,AHP-BPNN不仅简化神经网络的结构,而且提高了教学质量的评价精度和评价效率,是一种可行、有效的教学质量评价方法。     

Robust adaptive backstepping tracking control of stochastic nonlinear systems with unknown input saturation: A command filter approach

This paper studies the problem of adaptive observer‐based radial basis function neural network tracking control for a class of strict‐feedback stochastic nonlinear systems comprising an unknown input saturation, uncertainties, and unknown disturbances. To handle the issue of a non‐smooth saturation input signal, a smooth function is chosen to approximate the saturation function and the state observer is used to estimate unmeasured states. By the so‐called command filter method in the controller design procedure, the implementation complexity is reduced in the proposed backstepping method. Moreover, a radial basis function neural network is deployed to reconstruct the unknown nonlinear functions. In addition, the gains of all radial basis function neural networks are updated through one updating law leading to a minimal learning parameter which is independent of the number of neural nodes and the order of the system. Comparing with the existing results, the proposed approach can stabilize a constrained stochastic system more effectively and with less computational burden. Finally, a practical example shows the performance of the proposed controller design.     

基于知识和遥感图像的神经网络水质反演模型

为进一步提高遥感图像水质反演的精度，提出了一种基于知识和遥感图像相结合的神经网络水质反演模型。该模型利用遥感图像数据以及与水质相关的知识数据作为BP神经网络的输入，经训练后，确定神经网络的结构，在训练好的BP神经网络基础之上对水质进行反演。以中国太湖为例进行实证研究，实验中，使用的知识数据包括太湖的地理信息知识和先对太湖TM图像上的水域解译进而对水质进行分类的知识。实验结果表明，本文提出的水质反演模型较常规的线性回归模型和传统的神经网络模型有更高的反演精度。     

Direction of arrival estimation with antenna arrays based on fuzzy cerebellar model articulation controller neural network

Direction of arrival (DOA) estimation has been a challenging problem in many applications such as wireless communication, radar, sonar, and navigation. However, it is difficult to improve the angle resolution and reduce the computational complexity of super‐resolution methods. To solve these problems, the DOA estimation is viewed as a mapping problem, which can be modeled using a suitable artificial neural network trained with input‐output pairs. This article presents the use of a fuzzy cerebellar model articulation controller (FCMAC) neural network for the DOA estimation under a linear antenna array. The FCMAC neural network is a special feedforward neural network based on local approximation that can be adapted to solve the multidimensional nonlinear fitting problem. A new preprocessing scheme has been used in both training and test phase. It use magnitude and phase angles instead of the real and imaginary parts of the array covariance matrix to be the input of neural network. The proposed method avoids complex matrix eigen‐decomposition, such as multiple signal classification, and offers fast computation rate. The performance of FCMAC neural network is compared with the conventional subspace methods and the radial basis function neural network in the cases of noisy environment and coherent signal. Simulation results indicate that FCMAC neural network produces up to 61% lower error, 60% higher angle resolution, and 99% lower calculation time than other three methods, which indicates the superior performance of the proposed DOA estimation method under coherent signals and different noise levels.     

Wood defects classification using laws texture energy measures and supervised learning approach

Machine vision based inspection systems are in great focus nowadays for quality control applications. The proposed work presents a novel approach for classification of wood knot defects for an automated inspection. The proposed technique utilizes gray level co-occurrence matrix and laws texture energy measures as texture feature extractors and feed-forward back-propagation neural network as classifier. The proposed work involves the comparison of gray level co-occurrence matrix based features with laws texture energy measures based features. Firstly it takes contrast, correlation, energy and homogeneity as input parameters to a feed-forward back propagation neural network to predict wood defects and then it take energy calculated from laws texture energy measures based energy maps as input feature to a feed-forward back propagation neural network. Mean Square Error (MSE) for training data is found to be 0.0718 and 90.5% overall average classification accuracy is achieved when laws texture energy measures based features are used as input to the neural network as compared to gray level co-occurrence matrix based input features where MSE for training data is found to be 0.10728 and 84.3% overall average classification accuracy is achieved. The proposed technique shows promising results to classify wood defects using a feed forward back-propagation neural network.     

基于神经网络的丙酮产品质量分类挖掘

针对丙酮精制过程的特点,提出一种基于神经网络的丙酮产品质最分类挖掘方法。首先,讨论了数据挖掘中自变量筛选的方法,包括相关性分析、Fisher指数分析、主成分回归分析以及偏最小二乘回归分析等,综合各种疗法分析的结果,对丙酮精制过程中众多的工艺影响因素进行了重要性排序并据此筛选出重要的自变量;以选入的变量作为输入变量,构造基于神经网络的产品质量分类器。实验结果表明,训练后的神经网络分类器在丙酮产品质量分类挖掘中取得了良好的效果。     

Robust Adaptive Neural Network Control For A Class Of Multiple‐Input Multiple‐Output Nonlinear Time Delay System With Hysteresis Inputs And Dynamic Uncertainties

This paper studies an adaptive neural control for nonlinear multiple‐input multiple‐output systems with dynamic uncertainties, hysteresis input, and time delay. The studied systems are composed of N nonlinear time‐delay subsystems and the interconnection terms are contained in every equation of each subsystem. Adaptive neural control algorithms are developed by introducing a well‐defined smooth function. The unknown time‐varying delays and the unmodeled dynamics are dealt with by constructing appropriate Lyapunov–Krasovskii functions and introducing an available dynamic signal. The main advantage of the proposed controllers is that they contain fewer parameter estimates that need to be updated online. Consequently, the accuracy of ultimate tracking errors asymptotically approaches a pre‐defined bound, and all signals in the closed‐loop systems are also ensured to be uniformly ultimately bounded. Finally, a simulation example is provided to illustrate the effectiveness and merits of the proposed adaptive neural network control schemes.     

A new model of on‐chip inductors on ferrite film using KB‐FDSMN neural network

A new model of on‐chip planar inductors on ferrite film is developed by virtue of the knowledge‐based frequency‐dependent space‐mapping neural network (KB‐FDSMN). A modified π‐equivalent circuit is used to construct the KB‐FDSMN model for improving reliability in the model generalization. This new model makes use of empirical formulas to quickly estimate some circuit parameters for reducing the number of independent variables, whereas a three‐layer neural network is trained for the desirable accuracy and used to compute the rest of circuit parameters. This new approach provides an efficient scheme to model the on‐chip magnetic film inductors. In comparison with the conventional neural network model and the standalone modified π‐equivalent model, this new KB‐FDSMN model can map the input–output relationships with fewer hidden neurons yet better accuracy and higher reliability in the model generalization. © 2010 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2010.     

一种改进的前馈神经网络结构及其在复杂系统建模中的应用

针对包含多道加工工序、输入变量很多的复杂工业系统建模精度难以提高的问题,提出一种改进的前馈神经网络结构,输入变量不是由同一层输入,而是根据变量起作用的前后次序分别在网络的不同层输入,真实反映了大工业过程的各生产工序中的参数发生作用的时间顺序。同时由于输入变量在适当的时候输入网络,从而使网络的规模减小。该神经网络是处理高维问题,尤其是建立包含多道加工工序的大工业过程模型问题的强有力工具。将该神经网络用于热连轧产品质量建模,经过实测数据拟合与检验,仿真结果表明:提出的小波神经网络结构是可行的而且有很好的应用前景。     

Adaptive event‐triggered neural control for nonlinear uncertain system with input constraint

In this article, the issue of developing an adaptive event‐triggered neural control for nonlinear uncertain system with input delay is investigated. The radial basis function neural networks (RBFNNs) are adopted to approximate the uncertain terms, where the time‐varying approximation errors are considered into the approximation system. However, the RBFNNs' weight vector is extended, which may cause the computing burdens. To save network resource, the computing burden caused by the weight vector is handled with the developed adaptive control strategy. Furthermore, in order to compensate the effect of input delay, an auxiliary system is introduced into codesign. With the help of adaptive backstepping technique, an adaptive event‐triggered control approach is established. Under the proposed control approach, the effect of input delay can be compensated effectively while the considered system suffered network resource constraint, and all signals in the close‐loop system can be guarantee bounded. Finally, two simulation examples are given to verify the proposed control method's effectiveness.     

An improved procedure for detection of heart arrhythmias with novel pre‐processing techniques

The objective of this study is to develop an algorithm to detect and classify six types of electrocardiogram (ECG) signal beats including normal beats (N), atrial pre‐mature beats (A), right bundle branch block beats (R), left bundle branch block beats (L), paced beats (P), and pre‐mature ventricular contraction beats (PVC or V) using a neural network classifier. In order to prepare an appropriate input vector for the neural classifier several pre‐processing stages have been applied. Initially, a signal filtering method is used to remove the ECG signal baseline wandering. Continuous wavelet transform is then applied in order to extract features of the ECG signal. Next, principal component analysis is used to reduce the size of the data. A well‐known neural network architecture called the multi‐layered perceptron neural network is then utilized as the final classifier to classify each ECG beat as one of six groups of signals under study. Finally, the MIT‐BIH database is used to evaluate the proposed algorithm, resulting in 99.5% sensitivity, 99.66% positive predictive accuracy and 99.17% total accuracy.     

A classification technique based on radial basis function neural networks

In this paper, a new classification method is proposed based on the radial basis function (RBF) neural network architecture. The method is particularly useful for manufacturing processes, in cases where on-line sensors for classifying the product quality are not available. More specifically, the fuzzy means algorithm is employed on a set of training data, where the input data refer to variables that are measured on-line and the output data correspond to quality variables that are classified by human experts. The produced neural network model acts as an artificial sensor that is able to classify the product quality in real time. The proposed method is illustrated through an application to real data collected from a paper machine. The method produces successful results and outperforms a number of classifiers, which are based on the feedforward neural network (FNN) architecture.     

Cascaded neural network based small array synthesis with robustness to noise

A cascaded neural network approach has been presented in this paper to estimate the excitation for the desired field distribution using a radial basis function neural network (RBFNN). The article has employed an electromagnetic design example consisting of 5 × 5 and 6 × 6 planar antenna array of isotropic sources with inter element‐distance of 0.5λ to show the adaptation of the neural network model in estimating the desired output. A neural network is trained using a dataset of suitable excitation voltages and its corresponding radiation patterns, which proves to be efficient in predicting the excitation voltages required to generate the desired pattern. A set of techniques based on a cascaded neural network is adopted for pattern synthesis using magnitude and phase, magnitude only, and template‐based input data. The robustness of the method has also been tested by considering noise with different SNR levels. The results found in each case have a close fit with the desired pattern.