Locality Preserving Projection Method Research Articles

Locality Preserving Projections-Based Spatiotemporal Modeling of the Temperature Distribution of Lithium-Ion Batteries

The temperature distribution is one of the key factors affecting the safety, capacity fade, and optimal use of lithium-ion batteries (LIBs). However, it is a challenging task to develop an accurate approximation model for estimating this distribution, owing to the inherent nature of thermal processes, including time/space coupling, infinite dimensions, and strong nonlinearity. Therefore, this paper proposes a novel locality preserving projections (LPP)-based spatiotemporal modeling method to address this problem. First, an orthogonal enhanced LPP (OELPP) method is proposed for time/space decoupling. With this novel reduction dimension technique, the spatiotemporal temperature variables are separated into spatial basis functions (SBFs) and low-dimensional time variables. Then, to handle the temporal nonlinearity, a broad learning system with kernel-based manifold features (BLS-KMF) is designed to construct a dynamic temporal model. To improve the modeling performance, the kernel trick and the intrinsic manifold information are introduced to the proposed BLS-KMF to learn the mapped features. Finally, with the time/space synthesis, the battery temperature distribution can be reconstructed. Experimental results on a 32 Ah ternary lithium-ion battery illustrate the superior performance and effectiveness of the proposed method.

Locality Preserving Projections with Autoencoder

Locality Preserving Projections (LPP) is a popular dimensionality reduction method in the manifold learning field. However, LPP and all its variants only consider the one-way mapping from the high-dimensional space to the low-dimensional space and have no reverse verification, resulting in inaccurate low-dimensional embeddings. In this paper, we propose a new LPP method, called LPPAE (Locality Preserving Projections with Autoencoder), based on the linear Autoencoder. It constructs a two-way mapping: at the encoding stage, the conventional projection of LPP is viewed as a mapping from the high-dimensional space to the low-dimensional space. At the decoding stage, the low-dimensional embeddings are mapped back to the original high-dimensional space. The main contributions of the new method are: (1) This design not only preserves the neighborhood relationship of the data but more importantly, ensures that the low-dimensional embeddings can more accurately ”represent” the original data, thus significantly improving the performance of LPP. Experimental results on Handwritten Alphadigits, COIL-20, Yale, AR datasets show that the recognition rates of LPPAE are 26.06, 10.09, 5.40, and 8.86% higher than those of the original LPP respectively. On the MNIST dataset, compared to some of the latest improvements of LPP, including LPPMDC, LAPP, LPP+TR, and DNLPP, the recognition rate of LPPAE has been improved by 12.50, 38.10, 9.10, and 2.61%, respectively. (2) LPPAE regards the conventional LPP as an encoder, which is a new perspective. The idea of LPPAE can be used as a general framework and then extended to other manifold learning methods, and then a series of new methods can be developed.

Multi-omics data integration for hepatocellular carcinoma subtyping with multi-kernel learning.

Hepatocellular carcinoma (HCC) is a leading malignant liver tumor with high mortality and morbidity. Patients at the same stage can be defined as different molecular subtypes associated with specific genomic disorders and clinical features. Thus, identifying subtypes is essential to realize efficient treatment and improve survival outcomes of HCC patients. Here, we applied a regularized multiple kernel learning with locality preserving projections method to integrate mRNA, miRNA and DNA methylation data of HCC patients to identify subtypes. We identified two HCC subtypes significantly correlated with the overall survival. The patient 3-years mortality rates in the high-risk and low-risk group was 51.0% and 23.5%, respectively. The high-risk group HCC patients were 3.37 times higher in death risk compared to the low-risk group after adjusting for clinically relevant covariates. A total of 196 differentially expressed mRNAs, 2,151 differentially methylated genes and 58 differentially expressed miRNAs were identified between the two subtypes. Additionally, pathway activity analysis showed that the activities of six pathways between the two subtypes were significantly different. Immune cell infiltration analysis revealed that the abundance of nine immune cells differed significantly between the two subtypes. We further applied the weighted gene co-expression network analysis to identify gene modules that may affect patients prognosis. Among the identified modules, the key module genes significantly associated with prognosis were found to be involved in multiple biological processes and pathways, revealing the mechanism underlying the progression of HCC. Hub gene analysis showed that the expression levels of CDK1, CDCA8, TACC3, and NCAPG were significantly associated with HCC prognosis. Our findings may bring novel insights into the subtypes of HCC and promote the realization of precision medicine.

Geometrical structure preservation joint with self-expression maintenance for adaptive graph learning

Locality preserving projection (LPP) is a subspace learning method that uses pairwise distance to measure the similarity between data points. However, when data points from different clusters are adjacent, the pairwise distance may not accurately reflect the similarity between data points. In this study, a novel graph learning model is proposed to alleviate this problem; it is called adaptive graph learning with geometrical structure preservation and self-expression maintenance (GEAGL), in which a discriminative LPP method is used to extract the geometrical structure of data. By integrating self-expressive learning into the LPP method, a similarity graph that preserves the geometrical structure and self-expressive properties of data can be adaptively learned. The learned similarity graph tends to consist of a block diagonal structure when data points are extracted from independent linear subspaces, which alleviates the cluster interweaving problem. In this study, the relationship between the proposed method and k-means clustering was revealed, and the geometrical structure preservation property of the proposed method was theoretically analyzed. Finally, a graph-based clustering method was developed based on the similarity graph produced by GEAGL. Experimental results of benchmark datasets demonstrate the superiority of the proposed method in comparison with state-of-the-art methods.

A Novel Discriminant Locality Preserving Projections Method

Locality preserving projections (LPP) is a popular unsupervised dimensionality reduction method based on manifold learning. As a supervised version of the LPP method, discriminant locality preserving projections (DLPP) method has been recently proposed and paid much attention to by researchers. However, the DLPP method has the small-sample-size (SSS) problem. In this paper, in the view of the eigenvalues of scattering matrices of DLPP, they are first mapped to the new values by two polynomial functions, and with the properties of the matrix function of the two polynomial functions, the criterion of the DLPP method is reconstructed; thus, a novel dimensionality reduction method, named polynomial discriminant locality preserving projections (PDLPP) method, is proposed. The proposed PDLPP method has two advantages: one is that it addresses the SSS problem of DLPP, and the other is that, with the nonlinear mapping implied by PDLPP, the distance between inter-class samples is much enlarged and then the better performance of pattern classification is achieved. The experiments are conducted on the COIL-20 database, ORL, Georgia Tech, and AR face datasets, and the results show that the PDLPP is superior to state-of-the-art methods.

Local class-specific discriminant analysis with variable weighting and its application in fault diagnosis

In this paper, local class-specific discriminant analysis with variable weighting (VW-LCSDA) model was developed for fault classification in industrial processes. Although FDA has been extensively studied in the field of fault diagnosis as a classic method of dimension reduction and classifying, the effectiveness of the FDA is degraded in many cases due to data distribution and number of classes. The proposed method weights the variable vectors by evaluating the impact of the fault on each feature. Besides, the in-class and out-of-class scatter matrixes of class-specific discriminant analysis (CSDA) are used for dimension reduction and classification. In addition, the local structure of the sample data is preserved with reference to the LPP(Locality preserving projection) method, which is used to optimize the scatter matrix. The proposed fault classification method is validated by the Tennessee-Eastman process. The simulation results show that the VW-LCSDA method is a fault diagnosis method superior to FDA, CSDA and variable-weighted CSDA (VW-CSDA).

Recognition of similar characters using gradient features of discriminative regions

One important and challenging issue in handwritten character recognition is the discrimination of visually similar characters. In this paper, we propose a character recognition method for distinguishing similar characters by augmenting commonly used image feature with gradient features from potentially discriminative image regions. The discriminative regions of similar characters sets are automatically detected by analysing the weight vectors of the sparsity promoting logistic fused Lasso method. The histogram of oriented gradients is adopted to compactly represent the gradient features. Additionally, the locality preserving projection method is employed to alleviate the high dimensional nature of the resulting feature vectors. Experimental results on handwritten Lanna Dhamma and Thai characters datasets demonstrate the capability of the proposed method in discriminating visually similar characters. The method also outperforms existing character recognition methods by considerable margins. It has a great potential for character recognition of other alphabets.

Fault detection based on improved local entropy locality preserving projections in multimodal processes

AbstractRecently, local modelling methods have drawn increasing attention for fault detection in industrial processes. However, current local modelling adopts the basic multivariate statistical process monitoring method without deep mining the internal statistical information of the process data. Locality preserving projection (LPP) is a typical local fault detection method. However, industrial processes typically have complex multimodal characteristics. In these situations, the traditional LPP method performs poorly due to its assumption that the process data are Gaussian and unimodal. To improve fault detection performance of LPP in multimodal industrial processes and mine the internal system information, this paper proposes a new fault detection method based on improved local entropy LPPs (ILELPPs). ILELPP first eliminates the multimodal and non‐Gaussian characteristics of the original data using the improved local entropy method. Then, LPP is applied to the preprocessed data, neglecting the influences of multimodal and non‐Gaussian characteristics. A multimodal numerical example and an industrial application in a semiconductor manufacturing process are used to verify the effectiveness of the proposed method. The simulation results demonstrate that the proposed method has better fault detection performance than the LPP, global entropy LPP, and local entropy LPP methods.

Multiple Feature Learning Based on Edge-Preserving Features for Hyperspectral Image Classification

The classification of hyperspectral images is the basis and hotspot in the research of hyperspectral images. In this paper, a classification algorithm of hyperspectral image based on multiple edge-preserving features and multiple feature learning (MFL) is proposed. First, aiming to eliminate the high correlation between adjacent bands and to remove the noise in the image, a new band clustering algorithm is employed to reduce the dimensions, where the dimension-reduced image can be used as spectral information features to extract linearly separable classes. Then, spatial information features are obtained by applying the multiple edge-preserving filter on the reduced-dimensional image. This filter is used to acquire more comprehensive spatial information features of the image for extraction of nonlinearly separable classes. Following that, the locality preserving projections method is applied to retain the representative spatial information from the extracted spatial information for classification accuracy. Finally, the spectral information features and spatial information features are combined for classification using the MFL. The experiments are conducted to verify the validity of the proposed algorithm on three universally adopted hyperspectral datasets.

EEG İşaretlerinde Boyut İndirgeme Algoritmalarının Karşılaştırılması

Like in all classification applications, the most important process which increases classification success of electroencephalography (EEG) applications is to choose the proper features for signals. Since there is not certain feature extraction method for data classification applications, used feature matrix size can be redundantly large and this state effect the system's speed and success negatively. In this study Data Set III of BCI competition 2003 was used. We extract features using this data set and then dimension of feature matrix size reduced by using Principal Component Analysis, Kernel Principal Component Analysis and Locality Preserving Projections method which is alternative to Principal Component Analysis. As a result, the best success rate is obtained as 83.28% when Linearity Preserving Projections algorithm with Chebycev distance measuring method is used.

Nuclear Norm-Based 2DLPP for Image Classification

Two-dimensional locality preserving projections (2DLPP) that use 2D image representation in preserving projection learning can preserve the intrinsic manifold structure and local information of data. However, 2DLPP is based on the Euclidean distance, which is sensitive to noise and outliers in data. In this paper, we propose a novel locality preserving projection method called nuclear norm-based two-dimensional locality preserving projections (NN-2DLPP). First, NN-2DLPP recovers the noisy data matrix through low-rank learning. Second, noise in data is removed and the learned clean data points are projected on a new subspace. Without the disturbance of noise, data points belonging to the same class are kept as close to each other as possible in the new projective subspace. Experimental results on six public image databases with face recognition, object classification, and handwritten digit recognition tasks demonstrated the effectiveness of the proposed method.

Supervised bilateral two-dimensional locality preserving projection algorithm based on Gabor wavelet

Bilateral two-dimensional locality preserving projection (B2DLPP) is an effective method for unsupervised linear dimensionality reduction, which directly extracts face features from image matrices based on locality criterion. Motivated by B2DLPP, this paper proposes a supervised bilateral two-dimensional locality preserving projection (SB2DLPP). Different from B2DLPP, the proposed method takes into account the class information when constructing the similarity matrix. It increases inter-class distance in the projection space so that better right and left-projection matrices are obtained. Furthermore, a Gabor-based supervised bilateral two-dimensional locality preserving projection method is proposed for face recognition. Gabor wavelet representations are adopted for face images to make the proposed method robust to illumination variations and facial expression changes. Then, SB2DLPP is applied to reduce feature dimension. The performance of the proposed method is evaluated and compared with other traditional face recognition schemes on the FERET, Yale and JAFFE databases. The experiment results demonstrate the effectiveness and superiority of the proposed approach.

Affinity Propagation Algorithm Based on Locality Preserving Projections and Particle Swarm Optimization

Affinity propagation algorithm is a new powerful and effective clustering method. One of the major problems in clustering is the determination of the optimal number of clusters. In this paper, the particle swarm optimization algorithm is utilized to cope with this problem by using the parameter p as each particle and Silhouette index as the fitness, which can search for the optimal value of p and determine the optimal number of clusters automatically. Moreover, the problem of information overlap is the main drawback of affinity propagation algorithm in dealing with complex structure or high dimensional data for clustering. Hence the enhanced Locality preserving projections method is proposed to integrate with affinity propagation algorithm to reduce the dimension of the data as a processing step. As the result of experiment shows, the proposed method can simultaneously obtain the optimal number of clusters accurately and improve the clustering accuracy by eliminating the redundant information of the data without losing the internal nonlinear structure.

High-level hyperspectral image classification based on spectro-spatial dimensionality reduction

Spectro-spatial dimensionality reduction in HyperSpectral Images (HSI) classification is a challenging task due to the problem of curse dimensionality, i.e. the high number of spectral bands and the heterogeneity of data. In this context, many dimensionality reduction methods have been developed to overcome the high correlation between bands and the redundancy of information in order to improve the classification accuracy. Most of these methods represent the original HSI as a set of vectors. Therefore, they only exploit spectral properties, neglecting the spatial information, i.e. the spatial rearrangement is lost. To jointly take advantage of spatial and spectral information, HSI has been recently represented as a tensor. In order to preserve the spatial and spectral information, we develop a hybrid method using both the Tensor Locality Preserving Projections method (TLPP) projecting the original data into a lower subspace and the Constrained Band Selection method (CBS) to select the relevant bands. These two methods will be jointly used to get high-level quality classification. Moreover, since the two obtained classifications are uncertain and imprecise, we propose to fuse them using the Dempster–Shafer’s Theory (DST) to obtain an accurate classification preserving the spectro-spatial information. The proposed approach has been applied on real HSI showing its efficiency compared with conventional dimensionality reduction methods.

Quality-Related Statistical Process Monitoring Method Based on Global and Local Partial Least-Squares Projection

A novel quality-related statistical process monitoring method based on global and local partial least-squares projection (QGLPLS) is proposed in this paper. The main idea of the QGLPLS method is to integrate the advantages of locality-preserving projections (LPP) and partial least squares (PLS) and extract meaningful low-dimensional representations of high-dimensional process and quality data. QGLPLS can exploit the underlying geometrical structure that contains both global and local information pertaining to the sampled data, including the process variable and quality variable measurements. It is well-known that the PLS method can find only the global structure information and ignores the local features of data sets and that the LPP method can preserve local features of data sets well without considering the product quality variables. The capacity for the preservation of global and local projections of the proposed method is compared to that of the PLS and LPP methods; the comparison results demonstrate ...

English

Offline Handwritten Word Recognition (HWR) plays a major role in the field of image processing and pattern recognition. Compared to online recognition, handwritten words cannot be identified easily because of the variations in the handwriting styles, type of paper used, quality of the scanner etc. In our paper we have focused on the Kannada handwritten word recognition. Large number of characters present in the Kannada language makes it as a open problem for the researchers. Major steps in offline Kannada HWR are preprocessing, feature extraction, and classification. Locality Preserving Projections (LPP) method is used here for the feature extraction. For the classification Support Vector Machines (SVM) is used. Result is compared with the K-Means classifier. Experimental results show that SVM is better than K-Means classifier for our data set.

Learning Robust Locality Preserving Projection via p-Order Minimization

Locality preserving projection (LPP) is an effective dimensionality reduction method based on manifold learning, which is defined over the graph weighted squared L2-norm distances in the projected subspace. Since squared L2-norm distance is prone to outliers, it is desirable to develop a robust LPP method. In this paper, motivated by existing studies that improve the robustness of statistical learning models via L1-norm or not-squared L2-norm formulations, we propose a robust LPP (rLPP) formulation to minimize the p-th order of the L2-norm distances, which can better tolerate large outlying data samples because it suppress the introduced biased more than the L1-norm or not squared L2-norm minimizations. However, solving the formulated objective is very challenging because it not only non-smooth but also non-convex. As an important theoretical contribution of this work, we systematically derive an efficient iterative algorithm to solve the general p-th order L2-norm minimization problem, which, to the best of our knowledge, is solved for the first time in literature. Extensive empirical evaluations on the proposed rLPP method have been performed, in which our new method outperforms the related state-of-the-art methods in a variety of experimental settings and demonstrate its effectiveness in seeking better subspaces on both noiseless and noisy data.

Discriminant Locality Preserving Projections Based on Neighborhood Maximum Margin

This paper took a research about the small size sample problem of the discriminant locality preserving projections method, and proposed the discriminant locality preserving projections method based on neighborhood maximum margin (NMMDLPP). Firstly, the training sample structured a weighted of K-nearest neighbor graph, and gave the weight parameter to each side of the nearest neighbor graph for obtaining the intraclass neighbors and interclass neighbors local geometry information of each point; then reduce the interval between the intraclass neighbors and increase the interval between the interclass neighbors with the result of transfer matrix, and applied the neighbor point optimal refactoring coefficient of the data to the objective function. This method chose the difference between the locality preserving between-class scatter and the locality preserving within-class scatter as the objective function to avoid of calculating the inversion of matrix. This method has conducted an experiment on the UMIST face database and Yale face database. Experimental results show that the NMMDLPP algorithm is superior to other algorithms in recognition rate. The recognition rate can reach more than 91.4%.

Data mining-based flatness pattern prediction for cold rolling process with varying operating condition

Data-rich environments in modern rolling processes provide a great opportunity for more effective process control and more total quality improvement. Flatness is a key geometrical feature of strip products in a cold rolling process. In order to achieve good flatness, it is necessary to reveal the factors that often influence the flatness quality, to develop a general flatness pattern prediction model that can handle the varying operating condition during the rolling of products with different specifications and to realize an effective flatness feedback control strategy. This paper develops a practical data mining-based flatness pattern prediction method for cold rolling process with varying operating condition. Firstly, the high-dimensional process measurements are projected onto a low-dimensional space (i.e., the latent variable space) using locality preserving projection method; at the same time, the Legendre orthogonal polynomials are used to extract the basic flatness patterns by projecting the high-dimensional flatness measurements into several flatness characteristic coefficients. Secondly, a mixture probabilistic linear regression model is adopted to describe the relationships between the latent variables and the flatness characteristic coefficients. Case study is conducted on a real steel rolling process. Results show that the developed method has not only the satisfactory prediction performance, but good potentials to improve process understanding and strip flatness quality.

Quaternion-based improved LPP method for color face recognition

In recent years, pattern recognition and computer vision have increasingly become the focus of research. Locality preserving projection (LPP) is a very important learning method in these two fields and has been widely used. Using LPP to perform face recognition, we usually can get a high accuracy. However, the face recognition application of LPP suffers from a number of problems and the small sample size is the most famous one. Moreover, though the face image is usually a color image, LPP cannot sufficiently exploit the color and we should first convert the color image into the gray image and then apply LPP to it. Transforming the color image into the gray image will cause a serious loss of image information. In this paper, we first use the quaternion to represent the color pixel. As a result, an original training or test sample can be denoted as a quaternion vector. Then we apply LPP to the quaternion vectors to perform feature extraction for the original training and test samples. The devised quaternion-based improved LPP method is presented in detail. Experimental results show that our method can get a higher classification accuracy than other methods.