Class Variance Research Articles

Feature selection based on star coordinates plots associated with eigenvalue problems

Feature selection consists of choosing a smaller number of variables to work with when analyzing high-dimensional data sets. Recently, several visualization tools, techniques, and feature relevance measures have been developed in order to help users carry out the feature selection. Some of these approaches are based on radial axes methods, where analysts perform backward feature elimination by discarding features that have a low impact on the visualizations. Similarly, in this paper, we propose a new feature relevance measure for star coordinates plots associated with the class of linear dimensionality reduction mappings defined through the solutions of eigenvalue problems, such as linear discriminant analysis or principal component analysis. We show that the approach leads to enhanced feature subsets for class separation or variance maximization in the plots for numerous data sets of the UCI repository. Lastly, in practice, the tool allows analysts to decide which features to discard by examining their relevance and by taking into account previous domain knowledge.

Image Segmentation of Brain MR Images Using Otsu’s based Hybrid WCMFO Algorithm

In this study, a novel hybrid Water Cycle Moth-Flame Optimization (WCMFO) algorithm is proposed for multilevel thresholding brain image segmentation in Magnetic Resonance (MR) image slices. WCMFO constitutes a hybrid between the two techniques, comprising the water cycle and moth-flame optimization algorithms. The optimal thresholds are obtained by maximizing the between class variance (Otsu’s function) of the image. To test the performance of threshold searching process, the proposed algorithm has been evaluated on standard benchmark of ten axial T₂-weighted brain MR images for image segmentation. The experimental outcomes infer that it produces better optimal threshold values at a greater and quicker convergence rate. In contrast to other state-of-the-art methods, namely Adaptive Wind Driven Optimization (AWDO), Adaptive Bacterial Foraging (ABF) and Particle Swarm Optimization (PSO), the proposed algorithm has been found to be better at producing the best objective function, Peak Signal-to-Noise Ratio (PSNR), Standard Deviation (STD) and lower computational time values. Further, it was observed thatthe segmented image gives greater detail when the threshold level increases. Moreover, the statistical test result confirms that the best and mean values are almost zero and the average difference between best and mean value 1.86 is obtained through the 30 executions of the proposed algorithm.Thus, these images will lead to better segments of gray, white and cerebrospinal fluid that enable better clinical choices and diagnoses using a proposed algorithm.

Ellipsoidal Subspace Support Vector Data Description

In this paper, we propose a novel method for transforming data into a low-dimensional space optimized for one-class classification. The proposed method iteratively transforms data into a new subspace optimized for ellipsoidal encapsulation of target class data. We provide both linear and non-linear formulations for the proposed method. The method takes into account the covariance of the data in the subspace; hence, it yields a more generalized solution as compared to Subspace Support Vector Data Description for a hypersphere. We propose different regularization terms expressing the class variance in the projected space. We compare the results with classic and recently proposed one-class classification methods and achieve better results in the majority of cases. The proposed method is also noticed to converge much faster than recently proposed Subspace Support Vector Data Description.

An improved image segmentation algorithm based on the maximum class variance method

Image segmentation is an important part of image processing. The result of image segmentation directly affects the effect of subsequent image processing. However the efficiency of the traditional maximum class variance method is low. This paper uses the cuckoo algorithm to optimize the traditional maximum class variance method to achieve a better segmentation effect. This image segmentation method combined with optimization theory can achieve the purpose of finding the optimal segmentation.

Maximum Likelihood Classification based on Classified Result of Boundary Mixed Pixels for High Spatial Resolution of Satellite Images

Maximum Likelihood Classification: MLC based on classified result of boundary Mixed Pixels (Mixel) for high spatial resolution of remote sensing satellite images is proposed and evaluated with Landsat Thematic Mapper: TM images. Optimum threshold indicates different results for TM and Multi Spectral Scanner: MSS data. This may since the TM spatial resolution is 2.7 times finer than MSS, and consequently, TM imagery has more spectral variability for a class. The increase of the spectral heterogeneity in a class and the higher number of channels being used in the classification process may play significant role. For example, the optimum threshold for classifying an agricultural scene using MSS data is about 2.5 standard deviations, while that for TM corresponds to more than four standard deviations. This paper compares the optimum threshold between MSS and TM and suggests a method of using unassigned boundary pixels to determine the optimum threshold. Further, it describes the relationship of the optimum threshold to the class variance with a full illustration of TM data. The experimental conclusions suggest to the user some systematic methods for obtaining an optimal classification with MLC.

Whale Optimization Algorithm Based CME-CV Thresholding for Soil Porous Medium Extraction

Simulated Soil CT image is one of the rich boundary limitation of all the informations of soil pore, pebble and solid medium. Ground truth of soil skeleton is more useful to examine porosity and void ratio. With the help of the above parameters are used to recognize the other physical natures of soil. Soil management system is a well-known identified tool to enhance the field of work to do. Image processing is the major methods to examine and delivers the outcomes. Primary process of image processing is thresholding the given input samples and limits the boundary of the various objects. Thresholding is under the trustworthy of most powerful two traditional algorithms such as Otsu inter class variance and Kapur’s maximum entropy. Mixture of both the traditional methods are given all the informations of soil samples. In proposed towards novelty, combine some Bio- Inspired optimization with the traditional mixtures. Whale Optimization Algorithm (WOA) enhance the efficiency and accuracy of the segmentation throughput. It compares between certain conventional Bio-inspired Optimization.

Segmentation of Low Contrast Satellite and Medical Image s Based on Level Set Function with Harmony Search Optimization Algorithm

The main aim of digital image segmentation for portioned the image in to its constituents parts for getting information regarding features of image also used to get pathological details from medical images. The literature available from last two decades the important scheme for image segmentation is with Level Set technique, multilevel thresholding of gray scale on histogram of image is also a traditional method of image segmentation. In this paper low contrast images from medical and satellite images considered for image segmentation to extract features. This paper puts forward a novel image segmentation method via Level Set Function along with Bi-Histogram Equalization based on Harmony Search Algorithm(LSFBHEHS). The Selective Binary and Gaussian Filtering Regularised Level Set (SBGFRLS) is efficient novel region based Active Contour Model, it uses a novel region-based signed pressure force (SPF) function, it can adeptly halt the contours at blurred edges and weak edges. Other important advantage is internal and external boundaries can be distinguished by fixing the initial contour may be anyplace in the considered image. This method is resourceful but requires more time and inefficient for segmentation of low contrast images. This problem is rectified by applying bi-histogram equalization(BHE) image enhancement method prior to Level Set, it can be treated as pre-processing. In BHE technique of image enhancement, the image histogram is partitioned into two divisions based optimized gray level threshold , and equalize each part of histogram separately and combined later. To find the optimized threshold level to slice the histogram into two parts, Otsu’s multilevel thresholding method used to find threshold level, to find optimized thresholding level Harmony Search Algorithm(HSA) is implemented to maximize inert class variance as objective function. For evaluating the proposed method and SBGFRLS, the qualitative measured used like Dice similarity index, Measure of Enhancement(EME) and time required, for experimentation numerous low contrast satellite and medical images are considered, results clarified that the proposed method is more efficient for low contrast and inhomogeneous images.

Quality grading of soybean seeds using image analysis

<span>Image processing and machine learning technique are modified to use the quality grading of soybean seeds. Due to quality grading is a very important process for the soybean industry and soybean farmers. There are still some critical problems that need to be overcome. Therefore, the key contributions of this paper are first, a method to eliminate shadow noise for segment soybean seeds of high quality. Second, a novel approach for color feature which robust for illumination changes to reduces problem of color difference. Third, an approach to discover a set of feature and to form classifier model to strengthen the discrimination power for soybean classification. This study used background subtraction to reduce shadow appearing in the captured image and proposed a method to extract color feature based on robustness for illumination changes which was H components in HSI model. We proposed classifier model using combination of the color histogram of H components in HSI model and GLCM statistics to represent the color and texture features to strengthen the discrimination power of soybean grading and to solve shape variance in each soybean seeds class. SVM classifiers are generated to identify normal seeds, purple seeds, green seeds, wrinkled seeds, and other seed types. We conducted experiments on a dataset composed of 1,320 soybean seeds and 6,600 seed images with varies in brightness levels. The experimental results achieved accuracies of 99.2%, 97.9%, 100%, 100%, 98.1%, and 100% for overall seeds, normal seeds, purple seeds, green seeds, wrinkled seeds, and other seeds, respectively</span>

Attack-Aware Dynamic Upstream Bandwidth Assignment Scheme for Passive Optical Network

Abstract Network security is an important component of today’s networks to combat the security attacks. The passive optical network (PON) works at the medium access layer (MAC). A distributed denial of service (DDOS) attack may be launched from the network and transport layers of an Optical Network unit (ONU). Although there are various security techniques to mitigate its impact, however, these techniques cannot mitigate the impact on the MAC Layer of the PON and can cause an ONU to continuously drain too much bandwidth. This will result in reduced bandwidth availability to other ONUs and, thus, causing an increase in US delays and delay variance. In this work we argue that the impact of a DDOS attack can be mitigated by improving the Dynamic bandwidth assignment (DBA) scheme which is used in PON to manage the US bandwidth at the optical line terminal (OLT). The present DBA schemes do not have the capability to combat a security attack. Thus, this study, uses a machine learning approach to learn the ONU traffic demand patterns and presents a security aware DBA (SA-DBA) scheme that detects a rogue (attacker) ONU from its traffic demand pattern and limits its illegitimate bandwidth demand and only allows it the bandwidth assignment to it as per the agreed service level agreement (SLA). The simulation results show that the SA-DBA scheme results in up to 53%, 55% and 90% reduced US delays and up to 84%, 76% and 95% reduced US delay variance of T2, T3 and T4 traffic classes compared to existing insecure DBA schemes.

Efficient Counterfactual Learning from Bandit Feedback

What is the most statistically efficient way to do off-policy optimization with batch data from bandit feedback? For log data generated by contextual bandit algorithms, we consider offline estimators for the expected reward from a counterfactual policy. Our estimators are shown to have lowest variance in a wide class of estimators, achieving variance reduction relative to standard estimators. We then apply our estimators to improve advertisement design by a major advertisement company. Consistent with the theoretical result, our estimators allow us to improve on the existing bandit algorithm with more statistical confidence compared to a state-of-theart benchmark.

Utilizing Class Information for Deep Network Representation Shaping

Statistical characteristics of deep network representations, such as sparsity and correlation, are known to be relevant to the performance and interpretability of deep learning. When a statistical characteristic is desired, often an adequate regularizer can be designed and applied during the training phase. Typically, such a regularizer aims to manipulate a statistical characteristic over all classes together. For classification tasks, however, it might be advantageous to enforce the desired characteristic per class such that different classes can be better distinguished. Motivated by the idea, we design two class-wise regularizers that explicitly utilize class information: class-wise Covariance Regularizer (cw-CR) and classwise Variance Regularizer (cw-VR). cw-CR targets to reduce the covariance of representations calculated from the same class samples for encouraging feature independence. cw-VR is similar, but variance instead of covariance is targeted to improve feature compactness. For the sake of completeness, their counterparts without using class information, Covariance Regularizer (CR) and Variance Regularizer (VR), are considered together. The four regularizers are conceptually simple and computationally very efficient, and the visualization shows that the regularizers indeed perform distinct representation shaping. In terms of classification performance, significant improvements over the baseline and L1/L2 weight regularization methods were found for 21 out of 22 tasks over popular benchmark datasets. In particular, cw-VR achieved the best performance for 13 tasks including ResNet-32/110.

A novel relative homogeneity thresholding method with optimization strategy

Determining thresholds by measuring class variance is highly effective for image segmentation. Otsu’s method and its derivatives are common approaches that are both simple and adaptable. In spite of these methods’ excellent segmentation performance, images with particular gray distributions cause a thresholding bias that limits their usefulness. We explore the limitations of Otsu’s method and apply other evaluation criteria. In particular, we determine the relative homogeneity between the object and the background and then use it as a classification criterion along with a new binary thresholding method. Our method employs a histogram-smoothing method to improve valley-point selection, establishes a uniformity measure to identify the region with the best homogeneity, and identifies an optimization function for obtaining the best values for the adjustable parameters and threshold value. We also introduce a multilevel thresholding criterion based on a binary thresholding approach. Experiments using real and ground truth test images confirm the validity of our proposed method. Our method also offers a denoising ability when configured to use neighborhood information.

IMPLEMENTATION OF EEG SIGNAL PROCESSING AND DECODING FOR TWO-CLASS MOTOR IMAGERY DATA

This work decodes two-class motor imagery (MI) based on four main processing steps: (i) Raw electroencephalographic (EEG) signal is decomposed to single trials and spatial filters are estimated for each trial by common spatial filtering (CSP) method; (ii) features are extracted by taking the log transformation (normal distribution) of the spatially filtered EEG signal; (iii) optimal channel selection algorithm is proposed to reduce the number of EEG channels, such approach is regarded as key technological advantage in the implementation of brain–computer interface (BCI) to reduce the system processing time; (iv) finally, support vector machine (SVM) is employed to discriminate two classes of left and right hand MI. Two variations of SVM were proposed: polynomial function kernel and radial-based function RBF kernel. The results revealed that CSP succeeded in removing the strong correlation bound between the EEG samples by maximizing the variance of class 2 samples while minimizing the variance of class 1 samples. The channel selection algorithm achieved its goal to reduce the data dimension by selecting two channels out of three having the lowest variance entropies of 0.239 and 0.261 for channel 1 and channel 2, respectively. The features vector was divided into 80% train and 20% test with five-fold cross validation. The classification performance of SVM-polynomial kernel was 87.86% while it is 95.72% for SVM-RBF kernel as average accuracy of five-folds for both. Thus SVM-RBF is superior to SVM-Poly in the proposed framework.

Timber species identification from chemical fingerprints using direct analysis in real time (DART) coupled to Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS): comparison of wood samples subjected to different treatments

Abstract Timber genus identification based on the anatomical features of wood is well established in botany. However, species-level wood identification is not always possible based on traditional wood morphology techniques alone. To compensate for the deficiencies of traditional methods, direct analysis in real time coupled to Fourier transform ion cyclotron resonance mass spectrometry (DART-FTICR-MS) was used to obtain the mass spectral fingerprints of different timber species. Using heartwood samples of two morphologically similar species, Pterocarpus santalinus and Pterocarpus tinctorius, subjected to different treatments, i.e. solvent extractions and powdered samples as well as air-dried samples and samples dried at low and high temperatures, we observed distinct chemical signatures for the wood samples from the two species, enabling rapid species-level identification when multivariate statistical analysis was adopted. The supervised orthogonal partial least squares discriminant analysis (OPLS-DA) models for samples subjected to different treatments all exhibited accurate differentiation performance of the explained fraction of variance of classes (R2Y = 0.936–0.987) and the cross-validated fraction of variance of classes (Q2 = 0.857–0.949). Compared with solvent types and the physical form of the sample, the drying treatment method had a greater impact on the chemical fingerprint from DART-FTICR-MS. Air-dried wood chips were the optimal samples for the DART-FTICR-MS method coupled with statistical analysis.

Operationally Accessible Bounds on Fluctuations and Entropy Production in Periodically Driven Systems.

For periodically driven systems, we derive a family of inequalities that relate entropy production with experimentally accessible data for the mean, its dependence on driving frequency, and the variance of a large class of observables. With one of these relations, overall entropy production can be bounded by just observing the time spent in a set of states. Among further consequences, the thermodynamic efficiency both of isothermal cyclic engines like molecular motors under a periodic load and of cyclic heat engines can be bounded using experimental data without requiring knowledge of the specific interactions within the system. We illustrate these results for a driven three-level system and for a colloidal Stirling engine.

Maximum Inter Class Variance Segmentation Algorithm Based on Decision Tree

In image segmentation, there are always some false targets which remain in the segmented image. As the grayscale values of these false targets are quite similar to the grayscale values of the targets of interest, it is very difficult to split them out. And because these false targets exist in the original image, which are not caused by noise or traditional filtering methods, such as median filtering, they cannot be eliminated effectively. It is important to analyze the characteristics of false targets, so the false targets can be removed. In addition, it should be noted that the targets of interest cannot be affected when the false targets are removed. In order to overcome above problems, a maximum inter-class variance segmentation algorithm based on a decision tree is proposed. In this method, the decision tree classification algorithm and the maximum inter-class variance segmentation algorithm are combined. First, the maximum inter-class variance algorithm is used to segment the image, and then a decision tree is constructed according to the attributes of regions in the segmented image. Finally, according to the decision tree, the regions of the segmented image are divided into three categories, including large target regions, small target regions and false target regions, so that the false target regions are removed. The proposed algorithm can eliminate the false targets and improve the segmentation accuracy effectively. In order to demonstrate the effectiveness of the algorithm proposed in this article, the proposed method is compared with some frequently used false target removal approaches. Experimental results show that the proposed algorithm can achieve better results than other algorithms.

Rational selection of reverse phase columns for high throughput LC-MS lipidomics.

Natural lipidomes are characterized by extremely high complexity and dynamic range of lipid concentrations. Furthermore, high diversity of lipid physicochemical properties requires high resolving powers for both chromatographic and mass spectrometric analytical platforms. Reverse-phase chromatography coupled with data-dependent MS/MS acquisition is one of the most popular techniques in untargeted lipidomics. Optimal method should provide good chromatographic separation and resolution, reproducibility, selectivity and sensitivity. Here, we developed and set-up a RPLC-MS/MS workflow capable of resolving complex mixtures of lipids in 32 min of analysis. Human blood plasma was chosen as a representative complex natural lipidome with large variance of lipid classes, species and lipid concentrations. Lipids were separated by RPLC on five different reverse phase columns with different types of stationary phase particles, size and chemistry. High mass accuracy MS analysis and data-dependent MS/MS analysis were performed using a Q Exactive™ HF Hybrid Quadrupole-Orbitrap™ Mass Spectrometer to identify individual lipid molecular species. This workflow was applied to evaluate the separation capability of each column and to identify the lipidomics profile in highly complex biological samples. As a result, we report more than 600 lipid species covering 18 lipid classes in human blood plasma and provide suggestions to the selection of the appropriate reverse phase column for the analysis of specific lipidomes.

Transferred Knowledge Aided Positioning via Global and Local Structural Consistency Constraints

The fluctuation of received signal strength (RSS) induced by changing environment is the main hindrance from practical applications of the fingerprint-based indoor positioning methods. Transfer learning can mitigate the fluctuation of RSS by transferring knowledge from a source domain (off-line RSS data) to a target domain (online RSS data). However, the existing transfer learning approaches do not fully take into account the full constraints in Global and LOcal Structural conSistency (GLOSS), thus resulting in insufficient knowledge transfer. To overcome the above drawback, we propose a Transferred knowlEdge-Aided POsiTioning (TEAPOT) approach via GLOSS constraints in this paper. TEAPOT imposes the global structural consistency by minimizing the differences between the marginal and conditional distributions of the source and target domains and maximizing the samples variance in a latent subspace. Simultaneously, it also imposes the local structural consistency by minimizing within class variance and maximizing between class variance to retain the source discriminative information and preserving the local neighborhood relationship by using manifold regularization. Furthermore, a nonlinear TEAPOT is derived to improve the ability of TEAPOT to alleviate the limitation of linear projection. Compared with the existing methods, two of our proposed TEAPOT approaches via GLOSS constraints show higher accuracy and better ability in handling out-of-sample generalization. The experimental results verify that the proposed method significantly outperforms the existing methods.

A Study on Detection and Recognition of Facial Area Using Linear Discriminant Analysis

We propose a more stable robust recognition algorithm which detects faces reliably even in cases where there are changes in lighting and angle of view, as well it satisfies efficiency in calculation and detection performance. We propose detects the face area alone after normalization through pre-processing and obtains a feature vector using (PCA). The feature vector is applied to LDA and using Euclidean distance of intra-class variance and inter class variance in the 2nd dimension, the final analysis and matching is performed. Experimental results show that the proposed method has a wider distribution when the input image is rotated 45 ° left / right. We can improve the recognition rate by applying this feature value to a single algorithm and complex algorithm, and it is possible to recognize in real time because it does not require much calculation amount due to dimensional reduction.

Segmentation of Rice Seedlings Using the YCrCb Color Space and an Improved Otsu Method

Rice seedling segmentation is a fundamental process of extracting the guidance line for automated rice transplanters with a visual navigation system, which can provide crop row information to ensure the transplanter plants seedlings along the crop row without damaging seedlings. However, obtaining accurate rice seedling segmentation in paddy fields is still a challenging task. In this paper, a rice seedling segmentation method in paddy fields is proposed. The method mainly consists of two steps: image graying and threshold segmentation. In the procedure of image graying, the RGB (Red Green Blue) seedling image is first converted into the YCrCb color space and a Cg component is constructed. A color-index 2Cg-Cb-Cr is then constructed for image graying based on the excess green index (2G-R-B), which can reduce the influence of illumination variation on the equality of image graying. For the second step, an improved Otsu method is proposed to segment rice seedlings. With respect to the improved Otsu method in this research, the background variance of within class variance is weighted by a probability parameter to ensure that the method works well for both bimodal and near-unimodal histogram images, and the search range of gray levels is constrained to reduce the time to search the segmentation threshold. Experimental results indicate that the proposed method achieves better segmentation results and reduces the computational cost compared with the traditional Otsu method and other improved Otsu methods.