Classification error of multilayer perceptron neural networks

In subject classification, artificial neural networks (ANNS) are efficient and objective classification methods. Thus, they have been successfully applied to the numerous classification fields. Sometimes, however, classifications do not match the real world, and are subjected to errors. These problems are caused by the nature of ANNS. We discuss these on multilayer perceptron neural networks. By studying of these problems, it helps us to have a better understanding on its classification.     

数据中心设计标准TIA-942解读

着重从系统可靠性分级和电信基础分级两个方面入手．详细阐述TIA-942标准中数据中心的级别划分，以及各级别的技术指标及要求。     

A critical analysis of the fish hook effect in hydrocyclone classifiers

The infrequently reported fish hook in hydrocyclone classifiers has been ignored by many or disputed due to its random and sporadic occurrence. The imprecision of measurement of the actual efficiency and the ‘uncertainty principle’/observer effect appear to be main factors in its evolution and partial acceptance. Significantly, fundamental models to date do not predict a fish hook. Current theories to explain it based on size dependent bypass mechanism are mere mathematical transformations. It may be of theoretical/academic interest only as its exclusion in simulation models does not appear to affect cyclone performance prediction. This paper attempts to show that it is still a long way before it could be universally accepted as a scientifically significant physical effect.     

L0-norm sparse representation based on modified genetic algorithm for face recognition

The typical sparse representation for classification (SRC) exploits the training samples to represent the test samples, and classifies the test samples based on the representation results. SRC is essentially an L₀-norm minimization problem which can theoretically yield the sparsest representation and lead to the promising classification performance. We know that it is difficult to directly resolve L₀-norm minimization problem by applying usual optimization method. To effectively address this problem, we propose the L₀-norm based SRC by exploiting a modified genetic algorithm (GA), termed GASRC, in this paper. The basic idea of GASRC is that it modifies the traditional genetic algorithm and then uses the modified GA (MGA) to select a part of the training samples to represent a test sample. Compared with the conventional SRC based on L₁-norm optimization, GASRC can achieve better classification performance. Experiments on several popular real-world databases show the good classification effectiveness of our approach.     

广州荔湖城项目规划发展评析

本文结合项目设计管理经历与国外城市设计理论，对荔湖城的规划设计与发展等方面作评析.     

Classification performance comprehensive evaluation of an air classifier based on fuzzy analytic hierarchy process

The classification performance evaluation goal for an air classifier is usually limited to one of the classification performance indices including cut size, classification precision, Newton classification efficiency and degree of dispersion. This method hardly evaluates these performance indices of an air classifier comprehensively and suitably. In order to evaluate the classification performance truly and synthetically, Fuzzy Analytic Hierarchy Process is used to calculate the weights of the classification performance indices after determining the hierarchical model in the present paper. The dimensionless transformation eliminates the effect of the different dimensions. Then, the comprehensive evaluation value of the classification performance for each experiment is obtained using the linear weighted method. The maximum value corresponds to the best classification performance among these evaluation values. In the present study, a turbo air classifier is used as the classification system and talc powders are used as materials. The best classification performance indices are a cut size of 16.5 μm, a classification precision of 0.59, a Newton classification efficiency of 57%, and a degree of dispersion of 2.13. The corresponding optimal operational parameter combinations are: the feeding speed is 40 kg·h^–1, the air inlet velocity is 5 m·s^–1 and the rotor cage's rotary speed is 1200·min^–1. This assessment method avoids the limitation of evaluating a single classification performance index and the incomplete information derived from single factor experiments. Furthermore, the method also provides quantitative evaluation criteria for the classification performance of an air classifier. In the proposed method, the classification performance indices can be selected and the precedence relation matrix of Fuzzy Analytic Hierarchy Process can be set flexibly according to production requirements.     

Incorporation of Fuzziness in ID3 and Generation of Network Architerture

This article first describes a fuzzy version of ID3, called fuzzy ID3 by incorporating fuzziness at input, output and node levels. A fuzziness measure is computed at each node, in terms of class membership, to take care of the uncertainty arising from overlapping regions. The measure is such that in the crisp (non-overlapping) case, the algorithm boils down to the classical ID3. A confidence factor is estimated at the nodes for both making a decision and determining the rule base for network mapping. In the second part, we deal with a scheme of designing a fuzzy knowledge-based network by encoding an MLP with the rules generated using fuzzy ID3, whereby the network topology is automatically determined. The frequency of samples (representative of a rule) and the confidence factors of unresolved/ambiguous nodes are taken into consideration during mapping. The effectiveness of the system, in terms of recognition scores and speed of convergence, is demonstrated on two real life data sets.       

Operator error and system deficiencies: Analysis of 508 mining incidents and accidents from Queensland, Australia using HFACS

Historically, mining has been viewed as an inherently high-risk industry. Nevertheless, the introduction of new technology and a heightened concern for safety has yielded marked reductions in accident and injury rates over the last several decades. In an effort to further reduce these rates, the human factors associated with incidents/accidents needs to be addressed. A modified version of the Human Factors Analysis and Classification System was used to analyze incident and accident cases from across the state of Queensland to identify human factor trends and system deficiencies within mining. An analysis of the data revealed that skill-based errors were the most common unsafe act and showed no significant differences across mine types. However, decision errors did vary across mine types. Findings for unsafe acts were consistent across the time period examined. By illuminating human causal factors in a systematic fashion, this study has provided mine safety professionals the information necessary to reduce mine incidents/accidents further.     

Bayesian classification for bivariate normal gene expression

A Bayesian optimal screening method (BOSc) is proposed to classify an individual into one of two groups, based on the observation of pairs of covariates, namely the expression level of pairs of genes (previously selected by a specific method, among the thousands of genes present in the microarray) measured using DNA microarrays technology. The method is general and can be applied to any correlated pair of screening variables, either with a bivariate normal distribution or which can be transformed into a bivariate normal.¹ Results on microarray data sets (Leukemia, Prostate and Breast) show that BOSc performance is competitive with, and in some cases significantly better than, quadratic and linear discriminant analyses and support vector machines classifiers. BOSc provides flexible parametric decision rules. Finally, the screening classifier allows the calculation of operating characteristics while addressing information about the prevalence of the disease or type of disease, which is an advantage over other classification methods.     

基于模糊推理的条形域形状分类

基于单义域邻接图的扫描图样矢量化,要生成具有较强整体性的条形域,包括线段、圆弧、箭头和多义线。它们的分类特征和分类知识均有模糊性,采用模糊推理进行分类。在分类时,以条形域较小和较大的直线度、圆形度和等宽度为模糊集合,以计算较少的升降半梯形为相应的隶属函数。这样分类符合人类认知方法,而且算法简单易于实现。