Kernel Entropy Discriminant Analysis for Dimension Reduction

Kernel Entropy Component Analysis (KECA) is a newer method than Kernel Principle Component Analysis (KPCA) for data transformation and dimensionality reduction in case of face recognition. Although in almost all previous researches using KECA are shown to be more superior and more appropriate method compared to KPCA, here in this paper the significance of Kernel PCA in handling face pose in surveillance images is compared to KECA. Comparative analysis is made to signify the importance of Kernel Principle Component Analysis in terms of pose invariant face recognition in surveillance.

Context information is important in object representation. By embedding context cue of image attributes into kernel descriptors, we propose a set of novel kernel descriptors called Context Kernel Descriptors (CKD) for object classification and detection. The motivation of CKD is to use spatial consistency of image attributes or features defined within a neighboring region to improve the robustness of descriptor matching in kernel space. For feature selection, Kernel Entropy Component Analysis (KECA) is exploited to learn a subset of discriminative CKD. Different from Kernel Principal Component Analysis (KPCA) that only keeps features contributing mostly to image reconstruction, KECA selects the CKD that contribute mostly to the Renyi entropy of the image. These CKD are discriminative as they relate to the density distribution of the histogram of image attributes. We report superior performance of CKD for object classification on the CIFAR-10 dataset, and for detection on a challenging chicken feet dataset.

Projection pursuit and PCA associated with near and middle infrared hyperspectral images to investigate forensic cases of fraudulent documents

Effective information processing technology is one of the keys to improving detection accuracy. In this study, a feature reduction strategy is proposed for reducing the dimension of electronic nose (e-nose) sensor features, in combination with multiclassifiers to identify the origin of rice. Firstly, the time domain and time-frequency domain features were extracted from the detection data. Secondly, the kernel principal component analysis and kernel entropy component analysis (KECA) were introduced to reduce the dimension of the fusion features to obtain the kernel principal components (KPCs) and kernel entropy components (KECs). Finally, global discriminant analysis (GDA) was proposed in order to reduce the dimension of the KPCs and KECs to obtain the final features, respectively. The results indicated that the KECA-GDA achieved the dimensionality reduction of fusion features, effectively, the good classification accuracy of 97% and 93.29%, F 1-scores of 0.9697 and 0.9410, and Kappa coefficients of 0.9648 and 0.9210 were obtained by means of the random forest (RF) method in uncooked and cooked rice, respectively. This study shows that KECA-GDA-RF can be used as an effective tool in tracing the origin of rice. Moreover, it can provide a useful processing technique to improve the measurement accuracy of an e-nose.

In order to improve the correct identification rate of six types of Chinese spirits using electronic nose (E-nose), the Kernel entropy component analysis (KECA) identification method combined with Genetic algorithm (GA) was proposed. Firstly, integral value (INV), relative steady-state average value (RSAV) and wavelet energy value (WEV) were extracted and employed to represent the E-nose data. Secondly, radial basis function (RBF) was selected as the kernel function, then the kernel parameter η of RBF was optimized by the matrix similarity measurement method and the GA. The corresponding optimized kernel parameter η was 16.8608 (matrix similarity measurement) and 67.9039 (GA), respectively. When the first 125 kernel entropy components were selected for Fisher discriminant analysis (FDA), the correct identification rate of FDA (KECA + FDA) combined with GA were 97.62 % and 98.81 % for the training set and testing set, respectively; the correct identification rate of FDA (KECA + FDA) combined with matrix similarity measurement were 93.58 and 91.67 % for the training set and testing set, respectively. Therefore, the kernel parameter η determined by GA was significantly better than that of matrix similarity measurement. Finally, the correct identification rate of FDA and KECA + FDA was compared, and the results of FDA were only 82.14 % and 79.92 % for the training set and testing set, respectively. The identification results of FDA were far worse than that of KECA + FDA. The KECA + FDA method combined with GA was suitable for the identification of the six types of Chinese spirits by E-nose.

BackgroundType 1 diabetes (T1D) is an autoimmune disease resulting in the targeted destruction of pancreatic β-cells and permanent loss of insulin production. Proper glucose management results in better clinical outcomes for T1D and provides a strong rationale to identify non-invasive biomarkers indicative or predictive of glycemic control. Therefore, we investigated the association of salivary inflammation with HbA1c in a T1D cohort.MethodsUnstimulated saliva was collected from 144 subjects with T1D at the USF Diabetes Center. BMI, duration of diabetes, and HbA1c were recorded during clinical visit. Levels of interleukin (IL)-1β, -6, -8, -10, IFN-γ, TNF-α, MMP-3, -8, and -9 were measured using multiplexing immunoassay analysis. To account for smoking status, salivary cotinine levels were also determined.ResultsMultiple linear (HbA1c) and logistic (self-reported gingival condition) regression analyses were performed to examine the relationships between the Principal Component Analysis (PCA) components and HbA1c and gingival condition (adjusted for age, duration of diabetes, BMI, and sex; model for HbA1c also adjusted for gingival condition and model for gingival condition also adjusted for HbA1c). PCA components 1 (MMP-8 and MMP-9) and 3 (TNF-α) were significantly associated with HbA1c (β = 0.28 ±0.14, p = 0.045; β = 0.31 ±0.14, p = 0.029), while PCA component 2 (IL-6, IL-1β, and IL-8) was significantly associated with gingival condition (OR 1.60 95% CI 1.09–2.34, p = 0.016). In general, increased salivary inflammatory burden is associated with decreased glycemic control and self-reported gingival condition.ConclusionsThe saliva may represent a useful reservoir of novel noninvasive inflammatory biomarkers predictive of the progression and control of T1D.

While time‐of‐flight secondary ion mass spectrometry (ToF‐SIMS) image generation has been useful in the determination of the spatial distribution of chemistry in a broad application area, the amount of secondary ion signal available in each pixel remains small, hampering the use of multivariate analysis (MVA) approaches. The pre‐treatment of data commonly comprises two approaches to increase pixel signal intensity prior to MVA: mass channel summation and pixel summation. Recent advances in instrumentation have lead to much greater signal per image pixel such that a true spectrum can be discerned at each point across the sample. In this study, we determine the outcomes of these two pre‐treatments prior to principal components analysis (PCA) and maximum autocorrelation factor (MAF) analysis in comparison with high signal intensity data. Both PCA and MAF analyses of the high signal intensity data are presented with MAF being identified as the most effective approach. Image data was reduced in intensity to determine the effectiveness of MVA with lower spectral intensity. MAF was found to outperform PCA on such data, although both techniques were useful in the identification of the chemistry present. The data were then mass summed, using a number of different approaches, to reduce mass resolution, leading to a detrimental effect on PCA analysis, but little discernable change to MAF output. Reducing the spatial resolution by summing spectra from adjacent pixels, however, lead to a severe blurring of the MAF image structure, with PCA performing well. Conclusions and recommendation for pre‐treatment and MVA technique selection are presented. Copyright © 2009 John Wiley & Sons, Ltd.

A novel Gabor-Kernel face recognition method is proposed in this paper. This involves convolving a face image with a series of Gabor wavelets at different scales, locations, and orientations and extracting features from resulting Gabor filtered images. kernel discriminant analysis (KDDA) is then applied to the feature vectors for dimension reduction as well as class separability enhancement. A database of 600 frontal-view face images from the FERET face database is used to test the method. Experimental results demonstrate the advantage of KDDA over other Kernel methods such as kernel principal component analysis (KPCA) and general discriminant analysis (GDA). Significant improvements are also observed when features are extracted from Gabor filtered images instead of the original images. A 94% accuracy has been observed for the novel Gabor + KDDA method on the FERET database using a simple classifier, which could be further improved by employing a more complex classifier and distance measurer.

The identification of between-group differences and changes in gait mechanics are useful for injury prevention. Previous studies suggest the differences in gait biomechanical variables may interact in a complex non-linear fashion rather than a simple linear fashion. A non-linear multivariate analysis technique is therefore necessary to unravel the inherent structure in the gait data. Kernel principal component analysis (KPCA) is an extension of principal component analysis (PCA), the most widely used method in this field, and can provide inside into non-linear relationships of the variables of interest. Despite the growing use of PCA in running gait research, no prior studies have evaluated the utility of KPCA in determining the between-group differences in running gait patterns. Therefore, the objective of this study was to compare the performance of KPCA and PCA in identifying differences between male and female runners and as well between young and older runners in running gait kinematics. Running kinematic data were analysed on a gender group (female, n = 50; male, n = 50) and an age group (young, n = 50; older, n = 50) for discrete variables and waveform data. The following key results were obtained: (1) the performance of the KPCA for the identification of gender and age differences in running gait kinematics increased as compared to using the linear PCA; (2) the Gaussian function often performed better than the polynomial function for these two experiments; (3) there was no consistent optimal value of either the width σ of the Gaussian kernel or the degree d of the polynomial kernel for different data types and experiments. These results suggest that non-linear features extracted by KPCA could be potentially useful in identifying and discriminating between-group differences and changes in running gait patterns, and could provide useful information about the intrinsic non-linear dynamics of running movement.

Multiple phases with phase to phase transitions are important characteristics of many batch processes. The linear characteristics between phases are taken into consideration in the traditional algorithms while nonlinearities are neglected, which can lead to inaccuracy and inefficiency in monitoring. The focus of this paper is nonlinear multi-phase batch processes. A similarity metric is defined based on kernel entropy component analysis (KECA). A KECA similarity-based method is proposed for phase division and fault monitoring. First, nonlinear characteristics can be extracted in feature space via performing KECA on each preprocessed time-slice data matrix. Then phase division is achieved with the similarity variation of the extracted feature information. Then, a series of KECA models and slide-KECA models are established for steady and transitions phases respectively, which can reflect the diversity of transitional characteristics objectively and preferably deal with the stage-transition monitoring problem in multistage batch processes. Next, in order to overcome the problem that the traditional contribution plot cannot be applied to the kernel mapping space, a nonlinear contribution plot diagnosis algorithm is proposed, which is easier, more intuitive and implementable compared with the traditional one. Finally, simulations are performed on penicillin fermentation and industrial application. Specifically, the proposed method detects the abnormal agitation power and the abnormal substrate supply at 47 h and 86 h, respectively. Compared with traditional methods, it has better real-time performance and higher efficiency. Results demonstrate the ability of the proposed method to detect faults accurately and effectively in practice.

This paper presents a novel algorithm - robust kernel principal component analysis (robust KPCA), on the basis of the research of kernel principal component analysis (KPCA) and robust principal component analysis (RPCA). First, this algorithm sets the radius of the images of the training samples in the feature space using kernel tricks, then determines whether the samples are outliers or not, and finally analyzes the training samples which have eliminated the outliers using KPCA algorithm. The improved KPCA algorithm not only retains the non-linearity property of KPCA algorithm but also gets better robustness. Because the effects of outliers are eliminated, robust KPCA algorithm gets higher accuracy than KPCA algorithm for data analysis. The simulation experiments show that the robust KPCA algorithm developed is better than the KPCA algorithm.

Kernel discriminant analysis (KDA) and the kernel principal component analysis (KPCA), which are the extension of the linear discriminant analysis (LDA) and the principal component analysis (PCA), respectively, from linear domain to nonlinear domain via the kernel trick, are two very popular nonlinear feature extraction methods. In this paper, we present a new feature extraction algorithm by combing KDA and KPCA, and then apply it to the face recognition task. The experimental results on Yale face dataset show that the proposed method can significantly improve the performance both KDA and KPCA.

The combination of advanced techniques with multivariate analysis methods are being used with great confidence by forensics experts due to their high precision and accuracy. The prime objective of present work was to explore the potential of chemometric tools in discriminating and predicting the blue pen inks when applied on Attenuated total reflectance Fourier transform infrared (ATR-FTIR) data of pen strokes of different brands and models commercially available in Pakistani markets. The pen strokes on fine A4 paper and recycled A4 paper were obtained from 86 samples including ballpoints, gel, oil- gel and roller ball pens. ATR-FTIR spectra were first characterized for different functional groups present in ink pigments however, they were apparently indistinguishable having some common functional groups. After spectra preprocessing the principal component analysis (PCA) and agglomerative hierarchical cluster analysis (AHC) showed that pen strokes could be distinguished. Both PCA and AHC analyses showed that ballpoints and gel pen could be discriminated and grouped separately. PCA analyses also showed the discrimination between local and foreign brands of ballpoints. Finally, classification method partial least squares-discriminant analysis (PLS-DA) was successfully applied for the classification and subsequently for the identification of type of pen used in personnel’s signature to mimic a forensic case. To evaluate PLS-DA classification performance, the classification parameters such as sensitivity, specificity, accuracy and precision were also derived from confusion matrix. Overall, this work emphasized the worth of multivariate analysis in classification and prediction of type of writing pen in a document as a forensic study.

A rapid and accurate method of UFLC-Q-TOF-MS/MS combined with multivariate statistical analysis was established for the identification of Ainsliaea fragrans from different origins in this study. The A. fragrans from different producing areas of Jiangxi, Yunnan, Henan and Jiangsu were determined by UFLC-Q-TOF-MS/MS in the negative ion mode. And the data of the study were analyzed by the Markerview and other software for the PCA and OPLS-DA cluster analysis as well as t test. The results of the principal component analysis(PCA)showed that the main components from different origins were well distinguished. And the results of multivariate statistical showed the differences and similarities between different producing areas. Besides, 40 different compounds were identified in the negative ion mode. This method for identifying A. fragrans from different producing areas has the advantages of rapid accuracy and simplicity, which laid the foundation for the evaluation of the quality of the A. fragrans.

Rhubarb (Rheum ribes L.) is used globally as both a food supplement and a vegetable. The lipophilic fingerprints of wild rhubarb of Turkiye origin were investigated by gas chromatography-mass spectrometry (GC-MS) combined with chemometrics, including hierarchical cluster analysis (HCA) and principal component analysis (PCA). GC-MS results identified a total of 72 lipophi6lic compounds, including fatty acids, fatty acid esters, simple hydrocarbons and other organics, in all parts of the plant. PCA and HCA analyses revealed intraspecific lipophilic component variation among six wild rhubarb populations, as well as significant variation among lipophilic fingerprints in leaves, fruits, flowers, and roots. Among the compounds, palmitic acid, α-linolenic acid, ethyl palmitate, ethyl oleate, linoleic acid ethyl ester and n-tetracosane were important for the overall lipophilic fingerprints of all parts of the rhubarb. These findings highlight the potential of using HCA and PCA analysis in combination with GC-MS fingerprints as an effective analytical approach.