A new learning method using prior information of neural networks

In this paper, we present a new learning method using prior information for three-layered neural networks. Usually when neural networks are used for identification of systems, all of their weights are trained independently, without considering their interrelation of weight values. Thus the training results are not usually good. The reason for this is that each parameter has its influence on others during the learning. To overcome this problem, first, we give an exact mathematical equation that describes the relation between weight values given by a set of data conveying prior information. Then we present a new learning method that trains a part of the weights and calculates the others by using these exact mathematical equations. In almost all cases, this method keeps prior information given by a mathematical structure exactly during the learning. In addition, a learning method using prior information expressed by inequality is also presented. In any case, the degree of freedom of networks (the number of     

A fuzzy neighborhood-based training algorithm for feedforward neural networks

In this work we present a new hybrid algorithm for feedforward neural networks, which combines unsupervised and supervised learning. In this approach, we use a Kohonen algorithm with a fuzzy neighborhood for training the weights of the hidden layers and gradient descent method for training the weights of the output layer. The goal of this method is to assist the existing variable learning rate algorithms. Simulation results show the effectiveness of the proposed algorithm compared with other well-known learning methods.     

Learning and convergence analysis of neural-type structurednetworks

A class of feedforward neural networks, structured networks, has recently been introduced as a method for solving matrix algebra problems in an inherently parallel formulation. A convergence analysis for the training of structured networks is presented. Since the learning techniques used in structured networks are also employed in the training of neural networks, the issue of convergence is discussed not only from a numerical algebra perspective but also as a means of deriving insight into connectionist learning. Bounds on the learning rate are developed under which exponential convergence of the weights to their correct values is proved for a class of matrix algebra problems that includes linear equation solving, matrix inversion, and Lyapunov equation solving. For a special class of problems, the orthogonalized back-propagation algorithm, an optimal recursive update law for minimizing a least-squares cost functional, is introduced. It guarantees exact convergence in one epoch. Several learning issues are investigated.     

神经网络建模在热膨胀螺栓形变测量中的应用

基于神经网络建立热膨胀螺栓形变的非线性数学模型。神经网络的辨识采用变尺度二阶快速学习算法,利用二阶插值法来优化搜索学习速率。新方法具有很快的收敛速度和良好的收敛精度,克服了BP算法在神经网络的权值训练中收敛速度过慢的缺点。热膨胀螺栓的受热形变测量结果表明,该学习算法适用于非线性系统的建模与辨识。     

一种卷积神经网络和极限学习机相结合的人脸识别方法

卷积神经网络是一种很好的特征提取器,但却不是最佳的分类器,而极限学习机能够很好地进行分类,却不能学习复杂的特征,根据这两者的优点和缺点,将它们结合起来,提出一种新的人脸识别方法。卷积神经网络提取人脸特征,极限学习机根据这些特征进行识别。本文还提出固定卷积神经网络的部分卷积核以减少训练参 数,从而提高识别精度的方法。在人脸库ORL和XM2VTS上进行测试的结果表明,本文的结合方法能有效提高人脸识别的识别率,而且固定部分卷积核的方式在训练样本少时具有优势。     

采用资格迹的神经网络学习控制算法

强化学习是解决自适应问题的重要方法,被广泛地应用于连续状态下的学习控制,然而存在效率不高和收敛速度较慢的问题.在运用反向传播(back propagation,BP)神经网络基础上,结合资格迹方法提出一种算法,实现了强化学习过程的多步更新.解决了输出层的局部梯度向隐层节点的反向传播问题,从而实现了神经网络隐层权值的快速更新,并提供一个算法描述.提出了一种改进的残差法,在神经网络的训练过程中将各层权值进行线性优化加权,既获得了梯度下降法的学习速度又获得了残差梯度法的收敛性能,将其应用于神经网络隐层的权值更新,改善了值函数的收敛性能.通过一个倒立摆平衡系统仿真实验,对算法进行了验证和分析.结果显示,经过较短时间的学习,本方法能成功地控制倒立摆,显著提高了学习效率.     

Spatio–Temporal Adaptation in the Unsupervised Development of Networked Visual Neurons

There have been many computational models mimicking the visual cortex that are based on spatial adaptations of unsupervised neural networks. In this paper, we present a new model called neuronal cluster which includes spatial as well as temporal weights in its unified adaptation scheme. The “in-place” nature of the model is based on two biologically plausible learning rules, Hebbian rule and lateral inhibition. We present the mathematical demonstration that the temporal weights are derived from the delay in lateral inhibition. By training with the natural videos, this model can develop spatio–temporal features such as orientation selective cells, motion sensitive cells, and spatio–temporal complex cells. The unified nature of the adaption scheme allows us to construct a multilayered and task-independent attention selection network which uses the same learning rule for edge, motion, and color detection, and we can use this network to engage in attention selection in both static and dynamic scenes.      

A constructive algorithm to synthesize arbitrarily connected feedforward neural networks

In this work we present a constructive algorithm capable of producing arbitrarily connected feedforward neural network architectures for classification problems. Architecture and synaptic weights of the neural network should be defined by the learning procedure. The main purpose is to obtain a parsimonious neural network, in the form of a hybrid and dedicate linear/nonlinear classification model, which can guide to high levels of performance in terms of generalization. Though not being a global optimization algorithm, nor a population-based metaheuristics, the constructive approach has mechanisms to avoid premature convergence, by mixing growing and pruning processes, and also by implementing a relaxation strategy for the learning error. The synaptic weights of the neural networks produced by the constructive mechanism are adjusted by a quasi-Newton method, and the decision to grow or prune the current network is based on a mutual information criterion. A set of benchmark experiments, including artificial and real datasets, indicates that the new proposal presents a favorable performance when compared with alternative approaches in the literature, such as traditional MLP, mixture of heterogeneous experts, cascade correlation networks and an evolutionary programming system, in terms of both classification accuracy and parsimony of the obtained classifier.     

A review and analysis of backpropagation neural networks for classification of remotely-sensed multi-spectral imagery

A literature survey and analysis of the use of neural networks for the classification of remotely-sensed multi-spectral imagery is presented. As part of a brief mathematical review, the backpropagation algorithm, which is the most common method of training multi-layer networks, is discussed with an emphasis on its application to pattern recognition. The analysis is divided into five aspects of neural network classification: (1) input data preprocessing, structure, and encoding, (2) output encoding and extraction of classes, (3) network architecture, (4) training algorithms, and (5) comparisons to conventional classifiers. The advantages of the neural network method over traditional classifiers are its nonparametric nature, arbitrary decision boundary capabilities, easy adaptation to different types of data and input structures, fuzzy output values that can enhance classification, and good generalization for use with multiple images. The disadvantages of the method are slow training time, inconsistent results due to random initial weights, and the requirement of obscure initialization values (e.g., learning rate and hidden layer size). Possible techniques for ameliorating these problems are discussed. It is concluded that, although the neural network method has several unique capabilities, it will become a useful tool in remote sensing only if it is made faster, more predictable, and easier to use.     

Squeezing More Past Knowledge for Online Class-Incremental Continual Learning

Continual learning (CL) studies the problem of learning to accumulate knowledge over time from a stream of data. A crucial challenge is that neural networks suffer from performance degradation on previously seen data, known as catastrophic forgetting, due to allowing parameter sharing. In this work, we consider a more practical online class-incremental CL setting, where the model learns new samples in an online manner and may continuously experience new classes. Moreover, prior knowledge is unavailable during training and evaluation. Existing works usually explore sample usages from a single dimension, which ignores a lot of valuable supervisory information. To better tackle the setting, we propose a novel replay-based CL method, which leverages multi-level representations produced by the intermediate process of training samples for replay and strengthens supervision to consolidate previous knowledge. Specifically, besides the previous raw samples, we store the corresponding logits and features in the memory. Furthermore, to imitate the prediction of the past model, we construct extra constraints by leveraging multi-level information stored in the memory. With the same number of samples for replay, our method can use more past knowledge to prevent interference. We conduct extensive evaluations on several popular CL datasets, and experiments show that our method consistently outperforms state-of-the-art methods with various sizes of episodic memory. We further provide a detailed analysis of these results and demonstrate that our method is more viable in practical scenarios.      

Generalized Radial Basis Function Networks Trained with Instance Based Learning for Data Mining of Symbolic Data

The application of the Radial Basis Function neural networks in domains involving prediction and classification of symbolic data requires a reconsideration and a careful definition of the concept of distance between patterns. This distance in addition to providing information about the proximity of patterns should also obey some mathematical criteria in order to be applicable. Traditional distances are inadequate to access the differences between symbolic patterns. This work proposes the utilization of a statistically extracted distance measure for Generalized Radial Basis Function (GRBF) networks. The main properties of these networks are retained in the new metric space. Especially, their regularization potential can be realized with this type of distance. However, the examples of the training set for applications involving symbolic patterns are not all of the same importance and reliability. Therefore, the construction of effective decision boundaries should consider the numerous exceptions to the general motifs of classification that are frequently encountered in data mining applications. The paper supports that heuristic Instance Based Learning (IBL) training approaches can uncover information within the uneven structure of the training set. This information is exploited for the estimation of an adequate subset of the training patterns serving as RBF centers and for the estimation of effective parameter settings for those centers. The IBL learning steps are applicable to both the traditional and the statistical distance metric spaces and improve significantly the performance in both cases. The obtained results with this two-level learning method are significantly better than the traditional nearest neighbour schemes in many data mining problems.     

面向图像识别的多层脉冲神经网络学习算法综述

相较于第1代和第2代神经网络,第3代神经网络的脉冲神经网络是一种更加接近于生物神经网络的模型,因此更具有生物可解释性和低功耗性。基于脉冲神经元模型,脉冲神经网络可以通过脉冲信号的形式模拟生物信号在神经网络中的传播,通过脉冲神经元的膜电位变化来发放脉冲序列,脉冲序列通过时空联合表达不仅传递了空间信息还传递了时间信息。当前面向模式识别任务的脉冲神经网络模型性能还不及深度学习,其中一个重要原因在于脉冲神经网络的学习方法不成熟,深度学习中神经网络的人工神经元是基于实数形式的输出,这使得其可以使用全局性的反向传播算法对深度神经网络的参数进行训练,脉冲序列是二值性的离散输出,这直接导致对脉冲神经网络的训练存在一定困难,如何对脉冲神经网络进行高效训练是一个具有挑战的研究问题。本文首先总结了脉冲神经网络研究领域中的相关学习算法,然后对其中主要的方法:直接监督学习、无监督学习的算法以及ANN2SNN的转换算法进行分析介绍,并对其中代表性的工作进行对比分析,最后基于对当前主流方法的总结,对未来更高效、更仿生的脉冲神经网络参数学习方法进行展望。     

BP网络结构、参数及训练方法的设计与选择

根据已有的BP网络设计及其改进方案,对一个人工神经网络进行具体的设计,通过其详细的设计步骤与过程,对网络隐含层神经元数、初始权值、学习速率等参数在BP网络设计过程中的关系与影响,以及不同的改进算法在网络训练中所起的作用给予进一步揭示,使人们从中得到更多的启迪,以便使更多的人能够设计出效率更高、精度更好的神经网络。     

Improving information-theoretic competitive learning by accentuated information maximization

In this paper, we propose a new computational method for information-theoretic competitive learning. We have so far developed information-theoretic methods for competitive learning in which competitive processes can be simulated by maximizing mutual information between input patterns and competitive units. Though the methods have shown good performance, networks have had difficulty in increasing information content, and learning is very slow to attain reasonably high information. To overcome the shortcoming, we introduce the rth power of competitive unit activations used to accentuate actual competitive unit activations. Because of this accentuation, we call the new computational method “accentuated information maximization”. In this method, intermediate values are pushed toward extreme activation values, and we have a high possibility to maximize information content. We applied our method to a vowel–consonant classification problem in which connection weights obtained by our methods were similar to those obtained by standard competitive learning. The second experiment was to discover some features in a dipole problem. In this problem, we showed that as the parameter r increased, less clear representations could be obtained. For the third experiment of economic data analysis, much clearer representations were obtained by our method, compared with those obtained by the standard competitive learning method.     

Performance comparison between the training method and the numerical method of the orthogonal neural network in function approximation

The orthogonal neural network is a recently developed neural network based on the properties of orthogonal functions. It can avoid the drawbacks of traditional feedforward neural networks such as initial values of weights, number of processing elements, and slow convergence speed. Nevertheless, it needs many processing elements if a small training error is desired. Therefore, numerous data sets are required to train the orthogonal neural network. In the article, a least‐squares method is proposed to determine the exact weights by applying limited data sets. By using the Lagrange interpolation method, the desired data sets required to solve for the exact weights can be calculated. An experiment in approximating typical continuous and discrete functions is given. The Chebyshev polynomial is chosen to generate the processing elements of the orthogonal neural network. The experimental results show that the numerical method in determining the weights gives as good performance in approximation error as the known training method and the former has less convergence time. © 2004 Wiley Periodicals, Inc. Int J Int Syst 19: 1257–1275, 2004.     

面向机器博弈的即时差分学习研究

以六子棋机器博弈为应用背景,实现了基于即时差分学习的估值函数权值调整自动化.提出了一种新的估值函数设计方案,解决了先验知识与多层神经元网络结合的问题.结合具体应用对象的特性,提出了对即时差分序列进行选择性学习的方法,在一定程度上避免了无用状态的干扰.经过10020盘的自学习训练,与同一个程序对弈,其胜率提高了8%左右,具有良好的效果.     

Classical and superposed learning for quantum weightless neural networks

A supervised learning algorithm for quantum neural networks (QNN) based on a novel quantum neuron node implemented as a very simple quantum circuit is proposed and investigated. In contrast to the QNN published in the literature, the proposed model can perform both quantum learning and simulate the classical models. This is partly due to the neural model used elsewhere which has weights and non-linear activations functions. Here a quantum weightless neural network model is proposed as a quantisation of the classical weightless neural networks (WNN). The theoretical and practical results on WNN can be inherited by these quantum weightless neural networks (qWNN). In the quantum learning algorithm proposed here patterns of the training set are presented concurrently in superposition. This superposition-based learning algorithm (SLA) has computational cost polynomial on the number of patterns in the training set.     

一种分式过程神经元网络及其应用研究

针对带有奇异值复杂时变信号的模式分类和系统建模问题，提出了一种分式过程神经元网络．该模型是基于有理式函数具有的对复杂过程信号的逼近性质和过程神经元网络对时变信息的非线性变换机制构建的。其基本信息处理单元由两个过程神经元成对偶组成。逻辑上构成一个分式过程神经元，是人工神经网络在结构和信息处理机制上的一种扩展．分析了分式过程神经元网络的连续性和泛函数逼近能力，给出了基于函数正交基展开的学习算法．实验结果表明，分式过程神经元网络对于带有奇异值时变函数样本的学习性质和泛化性质要优于BP网络和一般过程神经元网络。网络隐层数和节点数可较大减少，且算法的学习性质与传统BP算法相同．     

The adaptive control using BP neural networks for a nonlinear servo-motor

The servo-motor possesses a strongly nonlinear property due to the effect of the stimulating input voltage, load-torque and environmental operating conditions. So it is rather difficult to derive a traditional mathematical model which is capable of expressing both its dynamics and steady-state characteristics. A neural network-based adaptive control strategy is proposed in this paper. In this method, two neural networks have been adopted for system identification （NNI） and control （NNC）, respectively. Then, the commonly-used specialized learning has been modified, by taking the NNI output as the approximation output of the servo-motor during the weights training to get sensitivity information. Moreover, the rule for choosing the learning rate is given on the basis of the analysis of Lyapunov stability. Finally, an example of applying the proposed control strategy on a servo-motor is presented to show its effectiveness.     

The adaptive control using BP networks for a nonlinear servo-motor

The servo-motor possesses a strongly nonlinear property due to the effect of the stimulating input voltageload-torque and environmental operating conditions.So it is rather diffcult to derive a traditional mathematical model which is capable of expressing both its dynamics and steady-state characteristics.A neural network-based adaptive control strategy is proposed in this paper.In this method,two neural networks have been adopted for system identification(NNI)and control(NNC),respectively.Then,the commonly-used specialized learning has been modified,by taking the NNI output as the approximation output of the servo-motor during the weights training to get sensitivity information.Moreover,the rule for choosing the learning rate is given on the basis of the analysis of Lyapunov stability.Finally,an example of applying the proposed control strategy on a servo-motor is presented to show its effectiveness.