Optimization of computational time in pattern recognizers

A sequential organization of the computations arising from pattern recognizers by absolute comparison is suggested in order to reduce the mean computational time involved. This optimization problem is solved by means of a supervising system which exploits the information obtained from the pattern-vector through a preclassifier: this information has the form of a conditional probability distribution of the classes to which the pattern-vector may belong. The results are extended to pattern recognizers by relative comparison.     

Further results on optimization of recognition time

Preceding studies on optimization of computational time in pattern recognizers started from strong hypotheses of separability of the classes by a known recognizer and from consideration of naive algorithms which implement the recognizer. Here, we consider weaker separability hypotheses, which allow for doubtful cases, and slightly more sophisticated algorithms. The expressions of the mean computational length and of its total variation actually valid are presented with their relation to the old ones. We give evidence for the fact that the old criterion for deciding about the optimality of an algorithm, by simple ordering of the class-probabilities, is still applicable in this new setting.     

基于差分法的灰度图像矩的快速算法

本文提出了一种新的对于灰度图像的几何矩的快速算法。首先运用图像差分法,将图像函数f（x,y）变换为图像函数d（x,y）。其次,从x^n（n=1,2,3）的递推求和得到一组数组。灰度图像的几何矩可以由该数组和函数d（x,y）计算获得。这种方法的优点在于：图像行（列）中具有相同像素值的连续部分,经差分后,除端点外的其它部分都为0,求矩无需考虑值为0的像素。所以,求矩计算量大大地降低了。文中给出了实验结果,和其它灰度图像求矩算法相比,文中算法在大多数情形下都极大地降低了计算复杂度。该算法乘法和加法的运算次数大约是Belkasim’s算法的47．4％和59．8％,大约是Yang’s算法的35％和51．8％。     

Determination of optimal dimensionality in statistical pattern classification

The effect of learning sample size on the optimal pattern recognition dimensionality is considered. Some procedures for determination of the optimal dimensionality are described and compared by a simulation method.     

基于Multi-memory机制的克隆选择算法及其在模式识别中的应用

人工免疫系统研究中大多借鉴克隆选择原理来构建免疫识别算法。描述了Castro提出的克隆选择算法CLONALG的整体框架,指出其在大规模模式识别问题中的不可收敛性。在CLONALG的基础上设计了Multi-memory机制,并以模式识别为应用背景提出了新的基于Multi-memory机制的克隆选择的免疫算法MCA,提出并深入分析了记忆抗体训练过程中的关键因素——变异概率的计算公式。实验表明,采用MCA的免疫系统具有更强的泛化能力、更高的抗原识别率以及更能适应大规模问题的优良特性。     

基于人工神经网络的板形模式识别方法

根据带钢板形控制的要求,运用人工神经网络理论,提出了一种新的板形识别方法,代替了传统的多项式最小二乘拟合法,该法具有很强的容错性和抗干扰能力,编制了板形模式识别软件,识别效果很好。     

A different approach to off-line handwritten signature verification using the optimal dynamic time warping algorithm

In this paper, a method for the automatic handwritten signature verification (AHSV) is described. The method relies on global features that summarize different aspects of signature shape and dynamics of signature production. For designing the algorithm, we have tried to detect the signature without paying any attention to the thickness and size of it. The results have shown that the correctness of our algorithm detecting the signature is more acceptable. In this method, first the signature is pre-processed and the noise of sample signature is removed. Then, the signature is analyzed and specification of it is extracted and saved in a string for the comparison. At the end, using adapted version of the dynamic time warping algorithm, signature is classified as an original or a forgery one.     

Genetic optimization of GRNN for pattern recognition without feature extraction

This paper describes an approach for pattern recognition using genetic algorithm and general regression neural network (GRNN). The designed system can be used for both 3D object recognition from 2D poses of the object and handwritten digit recognition applications. The system does not require any preprocessing and feature extraction stage before the recognition. In GRNN, placement of centers has significant effect on the performance of the network. The centers and widths of the hidden layer neuron basis functions are coded in a chromosome and these two critical parameters are determined by the optimization using genetic algorithms. Experimental results show that the optimized GRNN provides higher recognition ability compared with that of unoptimized GRNN.     

计算智能方法在KDD系统中的应用与研究

该文论述了KDD系统中计算智能方法的性质和特点,并且提出了神经网络和遗传算法技术在具体的数据挖掘过程中的应用模型,另外,从精确性、鲁棒性等方面对算法的优劣进行了比较。     

Analysis of global pattern features

A fundamental method for analysing the global features of visual patterns is presented. Pattern features are considered to be synonomous with non-random distributions in the pattern statistics. In particular, the statistics of the chords of the patterns are considered, leading to the histograms of the chord lengths and angles. Peaks in these histograms indicate the presence of structure in the pattern. It is demonstrated how the points of the pattern contributing to this structure can be enhanced and/or extracted.

It is argued that this is a more fundamental way of obtaining pattern features than other ad hoc methods.

The chord space approach allows analytic solutions to be easily obtained for idealized patterns such as circles, parallel lines, etc.

A method for implementing the algorithms on a parallel processor is indicated.     

A new algorithm for computing the minimum Hausdorff distance between two point sets on a line under translation

To determine the similarity of two point sets is one of the major goals of pattern recognition and computer graphics. One widely studied similarity measure for point sets is the Hausdorff distance. So far, various computational methods have been proposed for computing the minimum Hausdorff distance. In this paper, we propose a new algorithm to compute the minimum Hausdorff distance between two point sets on a line under translation, which outperforms other existing algorithms in terms of efficiency despite its complexity of O((m+n)lg(m+n)), where m and n are the sizes of two point sets.     

Hybrid evolutionary learning for synthesizing multi-class pattern recognition systems

This paper describes one aspect of a machine-learning system called HELPR that blends the best aspects of different evolutionary techniques to bootstrap-up a complete recognition system from primitive input data. HELPR uses a multi-faceted representation consisting of a growing sequence of non-linear mathematical expressions. Individual features are represented as tree structures and manipulated using the techniques of genetic programming. Sets of features are represented as list structures that are manipulated using genetic algorithms and evolutionary programming. Complete recognition systems are formed in this version of HELPR by attaching the evolved features to multiple perceptron discriminators. Experiments on datasets from the University of California at Irvine (UCI) machine-learning repository show that HELPR’s performance meets or exceeds accuracies previously published.     

Visible models for interactive pattern recognition

The exchange of information between human and machine has been a bottleneck in interactive visual classification. The visible model of an object to be recognized is an abstraction of the object superimposed on its picture. It is constructed by the machine but it can be modified by the operator. The model guides the extraction of features from the picture. The classes are rank ordered according to the similarities (in the hidden high-dimensional feature space) between the unknown picture and a set of labeled reference pictures. The operator can either accept one of the top three candidates by clicking on a displayed reference picture, or modify the model. Model adjustment results in the extraction of new features, and a new rank ordering. The model and feature extraction parameters are re-estimated after each classified object, with its model and label, is added to the reference database. Pilot experiments show that interactive recognition of flowers and faces is more accurate than automated classification, faster than unaided human classification, and that both machine and human performance improve with use.     

A net-structure learning system for pattern description

This paper discusses a computer program that recognizes and describes two-dimensional patterns composed of subpatterns. The program also recognizes all patterns in a scene consisting of several patterns.

Patterns are stored in a learned hierarchical, net-structure memory. Weighted links between memory nodes represent subpattern/pattern relationships. Both short term and permanent memories are used.

Pattern recognition is accomplished with a serial heuristic search algorithm, which attempts to search memory and compute input properties efficiently. Without special processing, the program can be asked to look for all occurrences of a specified pattern in a scene.     

Investigation of multidimensional data using the interactive pattern analysis system ispahan

In 1972, Kanal has enumerated some reasons for the desirability of interactive pattern analysis and classification systems (IPACS). One such system, ISPAHAN, was developed at the Department of Medical Informatics of the Free University in Amsterdam. Although this system is still expanding, it is now in operation. In order to obtain some experience with its properties three data sets, varying in complexity were analysed. Especially the possibility of applying various pattern recognition techniques in succession is shown to lead to an unsupervised classification scheme which may yield essentially the same results as the maximum likelihood decision rule applied to the labelled data set. The results of unsupervised methods are generally influenced by user chosen initial conditions. An objective criterion should therefore be used to compare different configurations. The use of one such criterion, measuring the compactness of the resulting clusters, is demonstrated in this paper.     

Invariant kernel functions for pattern analysis and machine learning

In many learning problems prior knowledge about pattern variations can be formalized and beneficially incorporated into the analysis system. The corresponding notion of invariance is commonly used in conceptionally different ways. We propose a more distinguishing treatment in particular in the active field of kernel methods for machine learning and pattern analysis. Additionally, the fundamental relation of invariant kernels and traditional invariant pattern analysis by means of invariant representations will be clarified. After addressing these conceptional questions, we focus on practical aspects and present two generic approaches for constructing invariant kernels. The first approach is based on a technique called invariant integration. The second approach builds on invariant distances. In principle, our approaches support general transformations in particular covering discrete and non-group or even an infinite number of pattern-transformations. Additionally, both enable a smooth interpolation between invariant and non-invariant pattern analysis, i.e. they are a covering general framework. The wide applicability and various possible benefits of invariant kernels are demonstrated in different kernel methods. Editor: Phil Long.     

Improved pattern recognition with complex artificial immune system

In this paper, we introduce the application of transformation pattern recognition based on a complex artificial immune system. The key feature of the complex artificial immune system is the introduction of complex data representation. We use complex numbers as the data representation instead of binary numbers used before, besides the weight between different layers. The complex partial autocorrelation coefficients of input antigen which are considered as the antigen presentation are calculated in major histocompatibility complex (MHC) layer of the complex artificial immune system. In the simulations, the transformation of patterns, such as translation, scale or rotation, are recognized in much higher accuracy, and it has obviously higher noise tolerance ability than traditional real artificial immune system and even the complex PARCOR model.     

A system for pattern recognition and pattern summarization in multi-band satellite images

A new system (C.T.R.F.’s LSGENSYS—Linguistic Summary Generation System) that has been developed for pattern recognition and summarization of patterns in multiband (RGB) satellite images is described in this paper. The system design is described in some detail. The system has been tested successfully with SPOT MS and LANDSAT images. It extracts, analyzes, and summarizes patterns such as land, island, water body, river, fire, and urban settlements from these images. The results are presented by allowing the system to automatically classify and interpret these images. Some elements of supervised classification are also introduced, and a comparison is made between the results in each case. The text was submitted by the author in English. Hema Nair. Date of Birth: November 21, 1965. Education: Hema Nair received her Bachelors Degree in Electrical Engineering from Government Engineering College, University of Calicut, Kerala, India, in 1986. She received her Masters Degree in Electrical Engineering from National University of Singapore in 1993. Ms. Nair received her Masters Degree in Computer Science from Clark Atlanta University, Atlanta, United States, in 1996. Membership: A member of IEEE (USA) and ACM (USA) since 1997. A member of the Institution of Engineers (India) since 1988. Awards: 1. Ms. Nair’s Masters Degree research in the United States was funded by a US Army Grant. 2. One of Ms. Nair’s publications was cited with the Abstract in NASA’s Scientific and Technical Information Program Reports of 2006. Work Experience: 1. Ms. Nair was employed as Senior Technical Associate II at AT and T, New Jersey, United States, between 1996 and 2000. Her work included research and leading AT&T Projects as Project Leader. 2. She also served as Faculty in Apple Information Technology, Ltd, Bangalore, India, between 1987 and 1990. 3. Ms. Nair worked on contract as a lecturer in Multimedia University, Malaysia, between 2001 and 2005. 4. Since 2005, she has been working as a Researcher at C.T.R.F., a research and education foundation in India. Research Interests: Ms. Nair’s research interests include Image Analysis, Pattern Recognition, Databases, Artificial Intelligence, and Data Mining. Publications: Ms. Nair has published several papers internationally. These include 7 International Conference Papers and 4 International Journals. Reviewer for LASTED International Conference 2004.