Identifying Objects in Procedural Programs Using Clustering Neural Networks

This paper presents a general approach for the identification of objects in procedural programs. The approach is based on neural architectures that perform an unsupervised learning of clusters. We describe two such neural architectures, explain how to use them in identifying objects in software systems and briefly describe a prototype tool, which implements the clustering algorithms. With the aid of several examples, we explain how our approach can identify abstract data types as well as groups of routines which reference a common set of data. The clustering results are compared to the results of many other object identification techniques. Finally, several case studies were performed on existing programs to evaluate the object identification approach. Results concerning two representative programs and their generated clusters are discussed.     

基于自适应节的多媒体流QoS保证的研究

分布式多媒体计算机系统的一个关键问题是保证连续多媒体流的服务质量（QoS),大多数多媒体流要求的QoS是在一定范围内,即[QoSmin,QoSmax]。为了保证QoS,计算机系统和网络系统必须分配足够的CPU、I/O、内存和网络带宽资源。在多媒体流传输过程中,这些资源的可利用率是动态变化的,尤其是带宽资源,如Internet,文中提出了自适应节的概念,采用PID调节技术,在[QoSmin,QoSmax]内,动态调节QoS,自适应动态的可利用资源变化,既保证了多媒体流的QoS,又提高了资源的利用率。     

Feedforward networks in financial predictions: the future that modifies the present

The main goal of this paper is to show how relatively minor modifications of well-known algorithms (in particular, back propagation) can dramatically increase the performance of an artificial neural network (ANN) for time series prediction. We denote our proposed sets of modifications as the 'self-momentum', 'Freud' and 'Jung' rules. In our opinion, they provide an example of an alternative approach to the design of learning strategies for ANNs, one that focuses on basic mathematical conceptualization rather than on formalism and demonstration. The complexity of actual prediction problems makes it necessary to experiment with modelling possibilities whose inherent mathematical properties are often not well understood yet. The problem of time series prediction in stock markets is a case in point. It is well known that asset price dynamics in financial markets are difficult to trace, let alone to predict with an operationally interesting degree of accuracy. We therefore take financial prediction as a meaningful test bed for the validation of our techniques. We discuss in some detail both the theoretical underpinnings of the technique and our case study about financial prediction, finding encouraging evidence that supports the theoretical and operational viability of our new ANN specifications. Ours is clearly only a preliminary step. Further developments of ANN architectures with more and more sophisticated 'learning to learn' characteristics are now under study and test.     

一种改进的最近邻聚类学习算法

提出了一种自适应模糊逻辑系统（ＡＦＬＳ）的改进自适应学习算法,该算法将无导师学习算法与基于梯度信息的寻优学习算法相结和,并且在确定聚类过程中同时考虑样本输入与输出对聚类的影响。     

关于BP算法推导过程与讨论

本文通过对多层神经网络模型的BP算法的推导,讨论用BP算法对网络进行优化后误差精度、网络结构和学习次数之间的关系,最后指出BP算法的不足之处,提出加快学习过程的一些方法和技巧。     

Construction of a new adaptive wavelet network and its learning algorithm

A new adaptive learning algorithm for constructing and training wavelet networks is proposed based on the time-frequency localization properties of wavelet frames and the adaptive projection algorithm. The exponential convergence of the adaptive projection algorithm in finite-dimensional Hilbert spaces is constructively proved, with exponential decay ratios given with high accuracy. The learning algorithm can sufficiently utilize the time-frequency information contained in the training data, iteratively determines the number of the hidden layer nodes and the weights of wavelet networks, and solves the problem of structure optimization of wavelet networks. The algorithm is simple and efficient, as illustrated by examples of signal representation and denoising.     

重复传播网络的改进及其在遥感分类中的应用研究

对重复传播网络在权值初始化、网络学习进行修改,实现在一定警戒条件下竞争层节点的动
态分配,并在遥感分类上进行实验。重复传播网络中学习次数影响了竞争层节点数目和学习精度,
学习精度随学习次数的增加而有限度的增加;警戒参数决定了节点调整的搜索范围,随学习过程进
行,节点搜索范围变小,保证了分类精度和网络的稳定性;两层竞争层结构分类精度要优于单层,其
总精度和Kappa系数分别提高1.1%,0.02。实验结果表明该方法是一可行的遥感分类方法。     

基于矢量模型的语音线性预测研究

研究语音参数线性预测的并行处理问题。通过把语音源序列的相邻样本分组能够构成一个均方差平稳的语音向量自回归序列,在Hilbert空间中运用正交投影原理导出具有高度并行处理能力的一预测编码策略,由此可推出参数线性预测的并行处理自适应算法。同传统格型算法相比,这种算法的计算复杂度及存贮量有明显改善。最后通过仿真运算检测了算法的性能。     

Source number estimation and separation algorithms of underdetermined blind separation

Recently, sparse component analysis （SCA） has become a hot spot in BSS research. Instead of independent component analysis （ICA）, SCA can be used to solve underdetermined mixture efficiently. Two-step approach （TSA） is one of the typical methods to solve SCA based BSS problems. It estimates the mixing matrix before the separation of the sources. K-means clustering is often used to estimate the mixing matrix. It relies on the prior knowledge of the source number strongly. However, the estimation of the source number is an obstacle. In this paper, a fuzzy clustering method is proposed to estimate the source number and mixing matrix simultaneously. After that, the sources are recovered by the shortest path method （SPM）. Simulations show the availability and robustness of the proposed method.     

Adaptive guidance law design based on characteristic model for reentry vehicles

In this paper an adaptive guidance law based on the characteristic model is designed to track a reference drag acceleration for reentry vehicles like the Shuttle. The characteristic modeling method of linear constant systems is extended for single-input and single-output （SlSO） linear time-varying systems so that the characteristic model can be established for reentry vehicles. A new nonlinear differential golden-section adaptive control law is presented. When the coefficients belong to a bounded closed convex set and their rate of change meets some constraints, the uniformly asymptotic stability of the nonlinear differential golden-section adaptive control system is proved. The tracking control law, the nonlinear differential golden-section control law, and the revised logical integral control law are integrated to design an adaptive guidance law based on the characteristic model. This guidance law overcomes the disadvantage of the feedback linearization method which needs the precise model. Simulation results show that the proposed method has better performance of tracking the reference drag acceleration than the feedback linearizaUon one.