Feature Space

Spatial clustering, which groups similar spatial objects into classes, is an important component of spatial data mining [Han and Kamber, Data Mining: Concepts and Techniques, 2000]. Due to its immense applications in various areas, spatial clustering has been highly active topic in data mining researches, with fruitful, scalable clustering methods developed recently. These spatial clustering methods can be classified into four categories: partitioning method, hierarchical method, density-based method and grid-based method. Clustering large data sets of high dimensionality has always been a serious challenge for clustering algorithms. Many recently developed clustering algorithms have attempted to address either handling data with very large number of records or data sets with very high number of dimensions. This new clustering method GCHL (a Grid-Clustering algorithm for High-dimensional very Large spatial databases) combines a novel density-grid based clustering with axis-parallel partitioning strategy to identify areas of high density in the input data space. The algorithm work as well in the feature space of any data set. The method operates on a limited memory buffer and requires at most a single scan through the data. We demonstrate the high quality of the obtained clustering solutions, capability of discovering concave/deeper and convex/higher regions, their robustness to outlier and noise, and GCHL excellent scalability.

Recognizing human action from image sequences is an active area of research in computer vision. In this paper, we present a novel method for human action recognition from image sequences in different viewing angles that uses the Cartesian component of optical flow velocity and human body shape feature vector information. We use principal component analysis to reduce the higher dimensional shape feature space into low dimensional shape feature space. We represent each action using a set of multidimensional discrete hidden Markov model and model each action for any viewing direction. We performed experiments of the proposed method by using KU gesture database. Experimental results based on this database of different actions show that our method is robust.

We propose what we believe is a new approach to 3D reconstruction through the design of a 3D voxel volume, such that all the image information and camera geometry are embedded into one feature space. By customising the volume to be suitable for segmentation, the key idea that we propose is the recovery of a 3D scene through the use of globally optimal geodesic active contours. We also present an extension to this idea by proposing the novel design of a 4D voxel volume to analyse the stereo motion problem in multi-view image sequences.

Abstract In this paper, we introduce a new approach to learn dissimilarity for interactive search in content based image retrieval. In literature, dissimilarity is often learned via the feature space by feature selection, feature weighting or a parameterized function of the ...

In this paper, the effect of grey level quantization on the ability of co-occurrence probability statistics to classify natural textures is studied. Generally, as a function of increasing grey levels, many of the statistics demonstrate a decrease in classification ability while a few maintain constant classification accuracy. None of the individual statistics show increasing classification accuracy throughout all grey levels. Correlation analysis is used to rationalize a preferred subset of statistics. The preferred statistics set (contrast, correlation, and ...

We introduced a novel method employing a hierarchical domain ontology structure to extract features representing documents in our previous publication (Wang 2002). All raw words in the training documents are mapped to concepts in a concept hierarchy derived from the domain ontology. Based on these concepts, a concept hierarchy is established for the training document space, using is-a relationships defined in the domain ontology. An optimum concept set may be obtained by searching the concept hierarchy with an appropriate heuristic function. This may be used as the feature space to represent the training dataset. The proposed method aims to solve some drawbacks suffered by text classification algorithms and feature selection algorithms. In this paper, we conducted a series of experiments to compare our approach with other comparable feature-selection and feature-extraction methods. The results indicated that our approach has advantages in many aspects.

In this paper we report the application of techniques inspired by text retrieval research to the content-based query of image databases. In particular, we show how the use of an inverted le data structure permits the use of a feature space of O(10 4 ) dimensions, by restricting search to the subspace spanned by the features present in the query. A suitably sparse set of colour and texture features is proposed. A scheme based on the frequency of occurrence of features in both individual images and in the whole collection provides a means of weighting possibly incommensurate features in a compatible manner, and naturally extends to incorporate relevance feedback queries. The use of relevance feedback is shown consistently to improve system performance, as measured by precision and recall.

Abstract In this paper an automatic texture based volumetric region growing method for liver segmentation is proposed. 3D seeded region growing is based on texture features with the automatic selection of the seed voxel inside the liver organ and the automatic threshold value computation for the region growing stop condition. Co-occurrence 3D texture features are extracted from CT abdominal volumes and the seeded region growing algorithm is based on statistics in the features space. Each CT volume is composed by 230 slices, ...

High-dimensional data pose challenges to traditional clustering algorithms due to their inherent sparsity and data tend to cluster in different and possibly overlapping subspaces of the entire feature space. Finding such subspaces is called subspace mining. We present SCHISM, a new algorithm for mining interesting subspaces, using the notions of support and Chernoff-Hoeffding bounds. We use a vertical representation of the dataset, and use a depth-first search with backtracking to find maximal interesting subspaces. We test our algorithm on a number of highdimensional synthetic and real datasets to test its effectiveness.

In this study, we propose a novel method for medical problem, it is the integration of particle swarm optimization (PSO) and decision tree (C4.5) named PSO + C4.5 algorithm. To evaluate the effectiveness of PSO + C4.5 algorithm, it is implemented on 5 different data sets of life sciences obtained from UCI machine learning databases. Moreover, the results of PSO + C4.5 implementation are compared to logistic regression (LR), back propagation neural network (BPNN), support vector machine (SVM), and decision tree (C4.5). The accuracy index shows that PSO + C4.5 algorithm outperforms the other methods.

An automated approach to degradation analysis is proposed that uses a rotating machine's acoustic signal to determine Remaining Useful Life (RUL). High resolution spectral features are extracted from the acoustic data collected over the entire lifetime of the machine. A novel approach to the computation of Mutual Information based Feature Subset Selection is applied, to remove redundant and irrelevant features, that does not require class label boundaries of the dataset or spectral locations of developing defect to be known or pre-estimated. Using subsets of the feature space, multi-class linear and Radial Basis Function (RBF) Support Vector Machine (SVM) classifiers are developed and a comparison of their performance is provided. Performance of all classifiers is found to be very high, 85 to 98%, with RBF SVMs outperforming linear SVMs when a smaller number of features are used. As larger numbers of features are used for classification, the problem space becomes more linearly separable and the linear SVMs are shown to have comparable performance. A detailed analysis of the misclassifications is provided and an approach to better understand and interpret costly misclassifications is discussed. While defining class label boundaries using an automated k-means clustering algorithm improves performance with an accuracy of approximately 99%, further analysis shows that in 88% of all misclassifications the actual class of failure had the next highest probability of occurring. Thus, a system that incorporates probability distributions as a measure of confidence for the predicted RUL would provide additional valuable information for scheduling preventative maintenance.

Graph-based semi-supervised learning has recently emerged as a promising approach to data-sparse learning problems in natural language processing. All graph-based algorithms rely on a graph that jointly represents labeled and unlabeled data points. The problem of how to best construct this graph remains largely unsolved. In this paper we introduce a data-driven method that optimizes the representation of the initial feature space for graph construction by means of a supervised classifier. We apply this technique in the framework of label propagation and evaluate it on two different classification tasks, a multi-class lexicon acquisition task and a word sense disambiguation task. Significant improvements are demonstrated over both label propagation using conventional graph construction and state-of-the-art supervised classifiers.

In this paper we present denoising algorithms for enhancing noisy signals based on Local ICA (LICA), Delayed AMUSE (dAMUSE) and Kernel PCA (KPCA). The algorithm LICA relies on applying ICA locally to clusters of signals embedded in a high dimensional feature space of delayed coordinates. The components resembling the signals can be detected by various criteria like estimators of kurtosis or the variance of autocorrelations depending on the statistical nature of the signal. The algorithm proposed can be applied favorably to the problem of denoising multidimensional data. Another projective subspace denoising method using delayed coordinates has been proposed recently with the algorithm dAMUSE. It combines the solution of blind source separation problems with denoising efforts in an elegant way and proofs to be very efficient and fast. Finally, KPCA represents a non-linear projective subspace method that is well suited for denoising also. Besides illustrative applications to toy examples and images, we provide an application of all algorithms considered to the analysis of protein NMR spectra.

As most 'real-world' data is structured, research in kernel methods has begun investigating kernels for various kinds of structured data. One of the most widely used tools for modeling structured data are graphs. An interesting and important challenge is thus to investigate kernels on instances that are represented by graphs. So far, only very specific graphs such as trees and strings have been considered. This paper investigates kernels on labeled directed graphs with general structure. It is shown that computing a strictly positive definite graph kernel is at least as hard as solving the graph isomorphism problem. It is also shown that computing an inner product in a feature space indexed by all possible graphs, where each feature counts the number of subgraphs isomorphic to that graph, is NP-hard. On the other hand, inner products in an alternative feature space, based on walks in the graph, can be computed in polynomial time. Such kernels are defined in this paper.

Tensor-based techniques for learning allow one to exploit the structure of carefully chosen representations of data. This is a desirable feature in particular when the number of training patterns is small which is often the case in areas such as biosignal processing and chemometrics. However, the class of tensor based models is somewhat restricted and might suffer from limited discriminative power. On a different track, kernel methods lead to flexible nonlinear models that have been proven successful in many different contexts. Nonetheless, a naïve application of kernel methods does not exploit structural properties possessed by the given tensorial representations. The goal of this work is to go beyond this limitation by introducing non-parametric tensor based models. The proposed framework aims at improving the discriminative power of supervised tensor-based models while still exploiting the structural information embodied in the data. We begin by introducing a feature space formed by multilinear functionals. The latter can be considered as the infinite dimensional analogue of tensors. Successively we show how to implicitly map input patterns in such a feature space by means of kernels that exploit the algebraic structure of data tensors. The proposed tensorial kernel links to the MLSVD and features an interesting invariance property; the approach leads to convex optimization and fits into the same primal-dual framework underlying SVM-like algorithms.

Content-Based Image Retrieval (CBIR) systems are required to effectively harness infor- mation from ubiquitous image collections. Despite intense research efforts by the multidis- ciplinary CBIR community since early 1990s, apparently there is a mismatch between these advances and the one that is truly required to bring success to CBIR in the commercial mar- ket place. In this paper we provide

In this paper, a change detection approach based on an object-based classification of remote sensing data is introduced. The approach classifies not single pixels but groups of pixels that represent already existing objects in a GIS database. The approach is based on a supervised maximum likelihood classification. The multispectral bands grouped by objects and very different measures that can be derived from multispectral bands represent the n-dimensional feature space for the classification. The training areas are derived automatically from the geographical information system (GIS) database.

This paper presents an effective method for fingerprint verification based on a data mining technique called minutiae clustering and a graph-theoretic approach to analyze the process of fingerprint comparison to give a feature space representation of minutiae and to produce a lower bound on the number of detectably distinct fingerprints. The method also proving the invariance of each individual fingerprint by using both the topological behavior of the minutiae graph and also using a distance measure called Hausdorff distance.The method provides a graph based index generation mechanism of fingerprint biometric data. The self-organizing map neural network is also used for classifying the fingerprints.

We have proposed a new approach to pattern recognition in which not only a classifier but also a feature space of input variables is learned incrementally. In this paper, an extended version of Incremental Principal Component Analysis (IPCA) and Resource Allocating Network with Long-Term Memory (RAN-LTM) are effectively combined to implement this idea. Since IPCA updates a feature space incrementally by rotating the eigen-axes and increasing the dimensions, the inputs of a neural classifier must also be changed in their values and the number of input variables. To solve this problem, we derive an approximation of the update formula for memory items, which correspond to representative training samples stored in the longterm memory of RAN-LTM. With these memory items, RAN-LTM is efficiently reconstructed and retrained to adapt to the evolution of the feature space. This function is incorporated into our face recognition system. In the experiments, the proposed incremental learning model is evaluated over a self-compiled video clips of 24 persons. The experimental results show that the incremental learning of a feature space is very effective to enhance the generalization performance of a neural classifier in a realistic face recognition task.

This paper presents an effective method for fingerprint verification based on a data mining technique called minutiae clustering and a graph-theoretic approach to analyze the process of fingerprint comparison to give a feature space representation of minutiae and to produce a lower bound on the number of detectably distinct fingerprints. The method also proving the invariance of each individual fingerprint by using both the topological behavior of the minutiae graph and also using a distance measure called Hausdorff distance.The method provides a graph based index generation mechanism of fingerprint biometric data. The self-organizing map neural network is also used for classifying the fingerprints.

In this paper, the potential use tf spaceborne polarimettic synthetic aperture radar (SAB) data in mapping landcover tripes and monitoring d@restation in tropics is studied, ttere, the emphasis is placed on several clearing practices a~d forest regeneration that can be characteri:.ed by usi~,g the sensitivity of SAB channels to vegetation biomass and canopy str~lcture. A supervised Bayes-Jan classifier designed fi)r SAIl signal statistics is employed to separate five classes: primar*j fi)rest, second-ar~j fi)rest, pasture-crops, quebradao, and disturbed forest. The Land C-band polarimetric SAIl data acquired during the shuttle imaging radar-C (SIII-C)/X-SAB space-shuttle ntission in 1994 are used as input data to the class{tier, The results are verified by field observation and comparison with the LancLsat data acquired in August (f 1994. The SAB data can delineate these five classes with approxinmtely 72% accuracy. The confusion arises wher~ separating old secondary fi)rests from primary .tim'st aad the young ones from pasture-crops. It is shown that Landsat and SAg data camj contplenwntary information about the vegetation structure that, when used i1~ sy~u¢r~,ism, may increase the classification accuracq over secondar~j fi~rest regrowth. When the number (f land-co~)er types was reduced to three classes including primanj fi~rest, pasture-crops, and regrowth-disturbed fi~rest, the acc~tracy of classification increased to 87%. A dinwnsionality anahjsis of the classifier showed that the accurac~j can be fi~rther improved to 92% by reducing the feature space to L-band HH and HV chantwls. Comparison of SIIl-C data acquired in April (wet period) a~ld October (dr~j period) indicates that multi

Automatic discrimination of speech and music is an important tool in many multimedia applications. The paper presents a robust and effective approach for speech/music discrimination, which relies on a set of features derived from fundamental frequency (F0) estimation. Comparison between the proposed set of features and some commonly used timbral features is performed, aiming to assess the good discriminatory power of the proposed F0-based feature set. The classification scheme is composed of a classical Statistical Pattern Recognition classifier followed by a Fuzzy Rules Based System. Comparison with other well-proven classification schemes is also performed. Experimental results reveal that our speech/music discriminator is robust enough, making it suitable for a wide variety of multimedia applications.

Subspace selection approaches are powerful tools in pattern classification and data visualization. One of the most important subspace approaches is the linear dimensionality reduction step in the Fisher's linear discriminant analysis (FLDA), which has been successfully employed in many fields such as biometrics, bioinformatics, and multimedia information management. However, the linear dimensionality reduction step in FLDA has a critical drawback: for a classification task with c classes, if the dimension of the projected subspace is strictly lower than c - 1, the projection to a subspace tends to merge those classes, which are close together in the original feature space. If separate classes are sampled from Gaussian distributions, all with identical covariance matrices, then the linear dimensionality reduction step in FLDA maximizes the mean value of the Kullback-Leibler (KL) divergences between different classes. Based on this viewpoint, the geometric mean for subspace selection is studied in this paper. Three criteria are analyzed: 1) maximization of the geometric mean of the KL divergences, 2) maximization of the geometric mean of the normalized KL divergences, and 3) the combination of 1 and 2. Preliminary experimental results based on synthetic data, UCI Machine Learning Repository, and handwriting digits show that the third criterion is a potential discriminative subspace selection method, which significantly reduces the class separation problem in comparing with the linear dimensionality reduction step in FLDA and its several representative extensions.

Forecasting stock price or stock index is an important financial subject that has attracted researchers' attention for many years. In this paper, we put forward a new method called HLP as data preprocessing to process the stock data. By HLP method we can get the stock high low point with different frequency and amplitude. The extracted data describes the feature of stock price movement. After that we construct ANN models to forecast the stock movement direction and price. The HLP method and ANN models give assistance to investors.

In this work we consider the problem of soccer team discrimination. The approach we propose starts from the monocular images acquired by a still camera. The first step is the soccer player detection, performed by means of background subtraction. An algorithm based on pixels energy content has been implemented in order to detect moving objects. The use of energy information, combined with a temporal sliding window procedure, allows to be substantially independent from motion hypothesis. Colour histograms in RGB space are extracted from each player, and provided to the unsupervised classification phase. This is composed by two distinct modules: firstly, a modified version of the BSAS clustering algorithm builds the clusters for each class of objects. Then, at runtime, each player is classified by evaluating its distance, in the features space, from the classes previously detected. Algorithms have been tested on different real soccer matches of the Italian Serie A.

An on-line recursive algorithm for training support vector machines, one vector at a time, is presented. Adiabatic increments retain the Kuhn-Tucker conditions on all previously seen training data, in a number of steps each computed analytically. The incremental procedure is reversible, and decremental "unlearning" offers an efficient method to exactly evaluate leave-one-out generalization performance. Interpretation of decremental unlearning in feature space sheds light on the relationship between generalization and geometry of the data.

In any geostatistical study, an important consideration is the choice of an appropriate, repeatable, and objective search strategy that controls the nearby samples to be included in the location-specific estimation procedure. Almost all geostatistical software available in the market puts the onus on the user to supply search strategy parameters in a heuristic manner. These parameters are solely controlled by geographical coordinates that are defined for the entire area under study, and the user has no guidance as to how to choose these parameters. The main thesis of the current study is that the selection of search strategy parameters has to be driven by data-both the spatial coordinates and the sample values-and cannot be chosen beforehand. For this purpose, a genetic-algorithm-based ordinary kriging with moving neighborhood technique is proposed. The search capability of a genetic algorithm is exploited to search the feature space for appropriate, either local or global, search strategy parameters. Radius of circle/sphere and/or radii of standard or rotated ellipse/ellipsoid are considered as the decision variables to be optimized by GA. The superiority of GA-based ordinary kriging is demonstrated through application to the Wolfcamp Aquifer piezometric head data. Assessment of numerical results showed that definition of search strategy parameters based on both geographical coordinates and sample values improves cross-validation statistics when compared with that based on geographical coordinates alone. In the case of a variable search neighborhood for each estimation point, optimization of local search strategy parameters for an elliptical support domain-the orientation of which is dictated by anisotropic axes-via GA was able to capture the dynamics of piezometric head in west Texas/New Mexico in an efficient way.

A robust face detection technique along with mouth localization, processing every frame in real time (video rate), is presented. Moreover, it is exploited for motion analysis onsite to verify "liveness" as well as to achieve lip reading of digits. A methodological novelty is the suggested quantized angle features ("quangles") being designed for illumination invariance without the need for preprocessing (e.g., histogram equalization). This is achieved by using both the gradient direction and the double angle direction (the structure tensor angle), and by ignoring the magnitude of the gradient. Boosting techniques are applied in a quantized feature space. A major benefit is reduced processing time (i.e., that the training of effective cascaded classifiers is feasible in very short time, less than 1 h for data sets of order 10 4 ). Scale invariance is implemented through the use of an image scale pyramid. We propose "liveness" verification barriers as applications for which a significant amount of computation is avoided when estimating motion. Novel strategies to avert advanced spoofing attempts (e.g., replayed videos which include person utterances) are demonstrated. We present favorable results on face detection for the YALE face test set and competitive results for the CMU-MIT frontal face test set as well as on "liveness" verification barriers.

The overarching goal of this study was to map irrigated areas in the Ganges and Indus river basins using near-continuous time-series (8day), 500-m resolution, 7-band MODIS land data for [2001][2002]. A multitemporal analysis was conducted, based on a mega file of 294 wavebands, made from 42 MODIS images each of 7 bands. Complementary field data were gathered from 196 locations. The study began with the development of two cloud removal algorithms (CRAs) for MODIS 7-band reflectivity data, named: (a) blue-band minimum reflectivity threshold and (b) visible-band minimum reflectivity threshold.

This study investigates the efficacy of applying support vector machines (SVM) to bankruptcy prediction problem. Although it is a wellknown fact that the back-propagation neural network (BPN) performs well in pattern recognition tasks, the method has some limitations in that it is an art to find an appropriate model structure and optimal solution. Furthermore, loading as many of the training set as possible into the network is needed to search the weights of the network. On the other hand, since SVM captures geometric characteristics of feature space without deriving weights of networks from the training data, it is capable of extracting the optimal solution with the small training set size. In this study, we show that the proposed classifier of SVM approach outperforms BPN to the problem of corporate bankruptcy prediction.

Graphs are being increasingly used to model a wide range of scientific data. Such widespread usage of graphs has generated considerable interest in mining patterns from graph databases. While an array of techniques exists to mine frequent patterns, we still lack a scalable approach to mine statistically significant patterns, specifically patterns with low p-values, that occur at low frequencies. We propose a highly scalable technique, called GraphSig, to mine significant subgraphs from large graph databases. We convert each graph into a set of feature vectors where each vector represents a region within the graph. Domain knowledge is used to select a meaningful feature set. Prior probabilities of features are computed empirically to evaluate statistical significance of patterns in the feature space. Following analysis in the feature space, only a small portion of the exponential search space is accessed for further analysis. This enables the use of existing frequent subgraph mining techniques to mine significant patterns in a scalable manner even when they are infrequent. Extensive experiments are carried out on the proposed techniques, and empirical results demonstrate that GraphSig is effective and efficient for mining significant patterns. To further demonstrate the power of significant patterns, we develop a classifier using patterns mined by GraphSig. Experimental results show that the proposed classifier achieves superior performance, both in terms of quality and computation cost, over state-of-the-art classifiers.

The Support Vector Machine is a theoretically superior machine learning methodology with great results in classification of highdimensional datasets and has been found competitive with the best machine learning algorithms. In the past, SVMs have been tested and evaluated only as pixel-based image classifiers. Moving from pixel-based techniques towards object-based representation, the dimensions of remote sensing imagery feature space increases significantly. This results increasing complexity of the classification process, and causes problems to traditional sample-based classification schemes. The objective of this study was to evaluate SVMs for effectiveness and prospects for object-based image classification as a modern computational intelligence method. An SVM approach for multi-class classification was followed, based on primitive image objects produces by a multi-resolution segmentation algorithm. The segmentation algorithm produced primitive objects of variable sizes and shapes. Then, a feature selection step took place in order to provide the features for classification which involved spectral, texture and shape information. Contextual information was not used. Following the feature selection step, a module integrating an SVM classifier and the segmentation algorithm was developed in C++ and based on XML technology for feature representation. For training the SVM, sample image objects, derived from the segmentation procedure were used. The SVM procedure produced the final object classification results which were compared to the Nearest Neighbor classifier results, of the eCognition software, and were found satisfactory. The SVM approach seems very promising for Object Based Image Analysis and future work will focus on the integration SVM classifiers with rule-based classifiers.

Positioned as the backbone of service asset management console, a service registry has to enable real-time and offline service selection in an effective manner. This paper presents an analytic algorithm that is used to guide the architectural design of service exploration in a service registry. Service assets are proposed to be framed into a well-established categorical structure based on pattern recognition algorithm. This design aims to provide systematic methodology and enablement architecture for analyzing, clustering, and adapting heterogeneous services for dynamic application integration. The exploitation of pattern recognition algorithm maps a large amount of services into a manageable feature space, which consists of attributes that are related to static description and dynamic features, such as historical QoS and service-level agreement. The proposed architecture and associated service exploration methodology have been integrated into an industry strength service-oriented architecture solution design platform. We also present a case study using the developed platform to illustrate the proposed algorithm for business service clustering and selection.

Prediction of stock market returns is an important issue in finance. Artificial neural networks have been used in stock market prediction during the last decade. Studies were performed for the prediction of stock index values as well as daily direction of change in the index. In some applications it has been specified that artificial neural networks have limitations for learning the data patterns or that they may perform inconsistently and unpredictable because of the complex financial data used.

The paper compares semi-automated interpolation methods to produce soil-class maps from profile observations and by using multiple auxiliary predictors such as terrain parameters, remote sensing indices and similar. The Soil Profile Database of Iran, consisting of 4250 profiles, was used to test different soil-class interpolators. The target variables were soil texture classes and World Reference Base soil groups. The predictors were 6 terrain parameters, 11 MODIS EVI images and 17 physiographic regions (polygon map) of Iran. Four techniques were considered: (a) supervised classification using maximum likelihoods; (b) multinominal logistic regression; (c) regression-kriging on memberships; and (d) classification of taxonomic distances. The predictive capabilities were assessed using a control subset of 30% profiles and the kappa statistics as criterion. Supervised classification and multinominal logistic regression can lead to poor results if soil-classes overlap in the feature space, or if the correlation between the soil-classes and predictors is low. The two other methods have better predictive capabilities, although both are computationally more demanding. For both mapping of texture classes and soil types, the best prediction was achieved using regressionkriging of indicators/memberships (κ = 45%, κ = 54%). In all cases kappa was smaller than 60%, which can be explained by the preferential sampling plan, the poor definition of soil-classes and the high variability of soils. Steps to improve interpolation of soil-class data, by taking into account the fuzziness of classes directly on the field are further discussed. Published by Elsevier B.V.

In this paper, we present a new approach for color texture classification by use of Haralick features extracted from co-occurrence matrices computed from Local Binary Pattern (LBP) images. These LBP images, which are different from the color LBP initially proposed by Mäenpää and Pietikäinen, are extracted from color texture images, which are coded in 28 different color spaces. An iterative procedure then selects among the extracted features, those which discriminate the textures, in order to build a low dimensional feature space. Experimental results, achieved with the BarkTex database, show the interest of this method with which a satisfying rate of wellclassified images (85.6%) is obtained, with a 10-dimensional feature space.

In classic pattern recognition problems, classes are mutually exclusive by deÿnition. Classiÿcation errors occur when the classes overlap in the feature space. We examine a di erent situation, occurring when the classes are, by deÿnition, not mutually exclusive. Such problems arise in semantic scene and document classiÿcation and in medical diagnosis. We present a framework to handle such problems and apply it to the problem of semantic scene classiÿcation, where a natural scene may contain multiple objects such that the scene can be described by multiple class labels (e.g., a ÿeld scene with a mountain in the background). Such a problem poses challenges to the classic pattern recognition paradigm and demands a di erent treatment. We discuss approaches for training and testing in this scenario and introduce new metrics for evaluating individual examples, class recall and precision, and overall accuracy. Experiments show that our methods are suitable for scene classiÿcation; furthermore, our work appears to generalize to other classiÿcation problems of the same nature.

Remote sensing plays an important role in mineral exploration. One of its proven applications is extracting and locating alteration zones that are related to gold deposits. In this study, surface reflectance data derived from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) imagery are used to map hydrothermal alteration zones for gold exploration. A band ratio derived from the image spectra (4/8, 4/2, and 8/9 in RGB) and a mineral extraction method based on n-dimensional spectral feature space have been developed, and tested against other conventional methods, and known auriferous alteration zones. Spectral signatures used to construct the new band ratio are validated by a field study of the Abu-Marawat area in the Eastern Desert of Egypt. Results indicate that the method is promising for identifying alteration zones and is a useful tool for gold exploration in similar areas elsewhere.

In this paper we study a dual version of the Ridge Regression procedure. It allows us to perform non-linear regression by constructing a linear regression function in a high dimensional feature space. The feature space representation can result in a large increase in the number of parameters used by the algorithm. In order to combat this \curse of dimensionality", the algorithm allows the use of kernel functions, as used in Support Vector methods. We also discuss a powerful family of kernel functions which is constructed using the ANOVA decomposition method from the kernel corresponding to splines with an innite number of nodes. This paper introduces a regression estimation algorithm which is a combination of these two elements: the dual version of Ridge Regression is applied to the ANOVA enhancement of the in nitenode splines. Experimental results are then presented (based on the Boston Housing data set) which indicate the performance of this algorithm relative to other algorithms.

It is often necessary to have stage discharge curve extended (extrapolated) beyond the highest (and sometimes lowest) measured discharges, for river forecasting, flood control and water supply for agricultural/industrial uses. During the floods or high stages, the river may become inaccessible for discharge measurement. Rating curves are usually extended using log–log axes, which are reported to have a number of problems. This paper suggests the use of Support Vector Machine (SVM) in the extrapolation of rating curves, which works on the principle of linear regression on a higher dimensional feature space. SVM is applied to extend the rating curves developed at three gauging stations in Washington, namely Chehalis River at Dryad and Morse Creek at Four Seasons Ranch (for extension of high stages) and Bear Branch near Naselle (for extension of low stages). The results obtained are significantly better as compared with widely used logarithmic method and higher order polynomial fitting method. A comparison of SVM results with ANN (Artificial Neural Network) indicates that SVM is better suited for extrapolation.

The magnetic flux leakage (MFL) technique is commonly used for non-destructive testing of oil and gas pipelines. This testing involves the detection of defects and anomalies in the pipe wall, and the evaluation of the severity of these defects. The difficulty with the MFL method is the extent and complexity of the analysis of the MFL images. In this paper we show how modern machine learning techniques can be used to considerable advantage in this respect. We apply the methods of support vector regression, kernelization techniques, principal component analysis, partial least squares, and methods for reducing the dimensionality of the feature space. We demonstrate the adequacy of the performance of these methods using real MFL data collected from pipelines, with regard to the performance of both the detection of defects, and the accuracy in the estimation of the severity of the defects. We also show how low-dimensional latent variable structures can be effective for visualizing the clustering behaviour of the classifier.

Psycho-linguistic analysis Logistic regression Decision tree Support vector machine a b s t r a c t Text is still the most prevalent Internet media type. Examples of this include popular social networking applications such as Twitter, Craigslist, Facebook, etc. Other web applications such as e-mail, blog, chat rooms, etc. are also mostly text based. A question we address in this paper that deals with text based Internet forensics is the following: given a short text document, can we identify if the author is a man or a woman? This question is motivated by recent events where people faked their gender on the Internet. Note that this is different from the authorship attribution problem.

Rendón, The fuzzy classifier system: motivations and first results, Proc. First Intl. Conf. on Parallel Problem Solving from Nature-PPSN I, Springer, Berlin, 1991, pp. 330-334 (scatter Mamdani fuzzy rules for control/modeling problems) M. Valenzuela-Rendón, Reinforcement learning in the fuzzy classifier system, Expert Systems with Applications 14 (1998) 237-247 (scatter Mamdani fuzzy rules for control/modeling problems) J.R. Velasco, Genetic-based on-line learning for fuzzy process control, IJIS 13 (10-11) (1998) 891-903 (scatter Mamdani fuzzy rules for control problems)

Face detection is to find any face in a given image. Face recognition is a two-dimension problem used for detecting faces. The information contained in a face can be analysed automatically by this system like identity, gender, expression, age, race and pose. Normally face detection is done for a single image but it can also be extended for video stream. As the face images are normally upright, they can be described by a small set of 2-D characteristics views. Here the face images are projected to a feature space or face space to encode the variation between the known face images. The projected feature space or the face space can be defined as ‘eigenfaces’ and can be formed by eigenvectors of the face image set. The above process can be used to recognize a new face in unsupervised manner. This paper introduces an algorithm which is used for effective face recognition. It takes into consideration not only the face extraction but also the mathematical calculations which enable us to bring the image into a simple and technical form. It can also be implemented in real-time using data acquisition hardware and software interface with the face recognition systems. Face recognition can be applied to various domains including security systems, personal identification, image and film processing and human computer interaction.