Precision Matrix Research Articles

Automated Detection of Floating Wood Logs on River Surfaces using YOLOv5 Model for Flood Warning System

In Sabah and Sarawak, Malaysia, log transportation by rivers poses risks due to log fragmentation. This can obstruct water flow, causing navigation problems and flood risks during heavy rainfall. Current monitoring methods involve personnel at checkpoints but are slow. This project proposes an AI-based system using YOLO-v5 to detect intact logs and fragments on river surfaces. A dataset will be created by scraping websites and using Google Colab commands to download relevant keywords. Preprocessing includes data augmentation, contrast adjustment, noise reduction, and resolution standardization. The model is trained in Google Colab and integrated into a warning system using Thonny IDE. Performance metrics like precision, recall, F1 score, and confusion matrix are generated. By automating monitoring through AI, this project aims to improve safety and sustainability in Malaysian river log transportation.

Classification and Prediction Evaluation of a Fuzzy Rule-Based Model for Diabetes Diagnosis, Level of Care and Pima Indians Diabetes Datasets

— Classification and prediction of diseases is crucial in decision-making for the healthcare sectors, especially diabetes, a chronic disease. The availability and accessibility of diabetes datasets assists medical practitioners in the diagnosis process as well as researchers in various fields and these datasets are valuable sources. However, diabetes datasets are exposed to vagueness and uncertainty issues. This research work improved an interrelated decision-making model for diabetes by proposing a fuzzy rule-based model to handle the vagueness and uncertainty issues. The research methodology starts with pre-processing of the simulated diabetes diagnosis and level of care datasets that were validated by medical experts, as well as the Pima Indians Diabetes Dataset (PIDD). This is followed by the design of the fuzzy model and the construction of the fuzzy rules. Next, the testing of the fuzzy model using six supervised machine learning algorithms namely J48, Logistic, Naive Bayes Updateable, Random Tree, Bayes Net and AdaBoostM1. Lastly, the evaluation of the fuzzy model in terms of the accuracy, precision, recall, F1-Score and confusion matrix. Experimental results show 100% accuracy for the diabetes diagnosis fuzzy model, for all the five machine learning algorithms mentioned except AdaBoostM1 with 79.8165% accuracy. In addition, for the level of care fuzzy model, the highest accuracy produced is 97.1098% for J48 algorithm and the lowest accuracy is 93.1049% for Naive Bayes Updateable and Bayes Net algorithms. Furthermore, for the PIDD fuzzy model, the highest accuracy obtained is 74.8698% for J48 and AdaBoostM1 algorithms and the lowest accuracy is 70.1823% for Random Tree algorithm. Overall, the proposed fuzzy model produced a good accuracy and working as expected associated to the previous interrelated decision-making model.

Crop leaf disease detection for beans using ensembled-convolutional neural networks

Abstract Crops’ health is affected by a varied range of diseases. Convenient and precise diagnosis plays a substantial role in preventing the loss of crop quality. In the past decade, deep learning (DL), particularly Convolutional Neural Networks (CNNs), has presented extraordinary performance for diverse applications involving crop disease (CD) detection. In this study, a comparison is drawn for the three pre-trained state-of-art architectures, namely, EfficientNet B0, ResNet50, and VGG19. An ensembled CNN has also been generated from the mentioned CNNs, and its performance has been evaluated over the original coloured, grey-scale, and segmented dataset. K-means clustering has been applied with six clusters to generate the segmented dataset. The dataset is categorized into three classes (two diseased and one healthy class) of bean crop leaves images. The model performance has been assessed by employing statistical analysis relying on the accuracy, recall, F1-score, precision, and confusion matrix. The results have shown that the performance of ensembled CNNs’ has been better than the individual pre-trained DL models. The ensembling of CNNs gave an F1-score of 0.95, 0.93, and 0.97 for coloured, grey-scale, and segmented datasets, respectively. The predicted classification accuracy is measured as: 0.946, 0.938, and 0.971 for coloured, grey-scale, and segmented datasets, respectively. It is observed that the ensembling of CNNs performed better than the individual pre-trained CNNs.

Hyper Markov law in undirected graphical models with its applications

By exploring the prime decomposition of undirected graphs, this work investigates the hyper Markov property within the framework of arbitrary undirected graph, which can be seen as the generalization of that for decomposable graphical models proposed by Dawid and Laurizten. The proposed hyper Markov properties of this article can be used to characterize the conditional independence of a distribution or a statistical quantity and helpful to simplify the likelihood ratio functions for statistical test for two different graphs obtained by removing or adding one edge. As an application of these properties, the G-Wishart law is introduced as a prior law for graphical Gaussian models for Bayesian posterior updating, and a hypothesis test for precision matrix is designed to determine the model structures. Our simulation experiments are implemented by using the Gibbs Sampler algorithm and the results show that it performs better in convergence speed.

Joint semiparametric kernel network regression.

Variable selection and graphical modeling play essential roles in highly correlated and high-dimensional (HCHD) data analysis. Variable selection methods have been developed under both parametric and nonparametric model settings. However, variable selection for nonadditive, nonparametric regression with high-dimensional variables is challenging due to complications in modeling unknown dependence structures among HCHD variables. Gaussian graphical models are a popular and useful tool for investigating the conditional dependence between variables via estimating sparse precision matrices. For a given class of interest, the estimated precision matrices can be mapped onto networks for visualization. However, the limitation of Gaussian graphical models is that they are only applicable to discretized response variables and for the case when , where is the number of variables and is the sample size. They are necessary to develop a joint method for variable selection and graphical modeling. To the best of our knowledge, the methods for simultaneously selecting variable selection and estimating networks among variables in the semiparametric regression settings are quite limited. Hence, in this paper, we develop a joint semiparametric kernel network regression method to solve this limitation and to provide a connection between them. Our approach is a unified and integrated method that can simultaneously identify important variables and build a network among those variables. We developed our approach under a semiparametric kernel machine regression framework, which can allow for nonlinear or nonadditive associations and complicated interactions among the variables. The advantages of our approach are that it can (1) simultaneously select variables and build a network among HCHD variables under a regression setting; (2) model unknown and complicated interactions among the variables and estimate the network among these variables; (3) allow for any form of semiparametric model, including non-additive, nonparametric model; and (4) provide an interpretable network that considers important variables and a response variable. We demonstrate our approach using a simulation study and real application on genetic pathway-based analysis.

Advancing Brain Tumor Classification through Fine-Tuned Vision Transformers: A Comparative Study of Pre-Trained Models.

This paper presents a comprehensive study on the classification of brain tumor images using five pre-trained vision transformer (ViT) models, namely R50-ViT-l16, ViT-l16, ViT-l32, ViT-b16, and ViT-b32, employing a fine-tuning approach. The objective of this study is to advance the state-of-the-art in brain tumor classification by harnessing the power of these advanced models. The dataset utilized for experimentation consists of a total of 4855 images in the training set and 857 images in the testing set, encompassing four distinct tumor classes. The performance evaluation of each model is conducted through an extensive analysis encompassing precision, recall, F1-score, accuracy, and confusion matrix metrics. Among the models assessed, ViT-b32 demonstrates exceptional performance, achieving a high accuracy of 98.24% in accurately classifying brain tumor images. Notably, the obtained results outperform existing methodologies, showcasing the efficacy of the proposed approach. The contributions of this research extend beyond conventional methods, as it not only employs cutting-edge ViT models but also surpasses the performance of existing approaches for brain tumor image classification. This study not only demonstrates the potential of ViT models in medical image analysis but also provides a benchmark for future research in the field of brain tumor classification.

Estimation of banded time-varying precision matrix based on SCAD and group lasso

A new banded time-varying precision matrix estimator is proposed for high-dimensional time series. The estimator utilizes the modified Cholesky decomposition, and the two factors in the decomposition are dynamically estimated by applying the GARCH model to the innovation variance and the Kalman filter on the Cholesky factor. The SCAD penalty and group lasso penalty are imposed on the Cholesky factor to estimate the banded structure. An efficient algorithm based on the alternating direction method of multipliers (ADMM), local linear approximation (LLA), and blockwise coordinate descent (BCD) algorithms is developed. The convergence of the algorithm is proven theoretically, and the estimator is guaranteed to be banded. Simulation and real-data analysis demonstrate the favorable performance of the proposed algorithm compared to other methods.

Stochastic PDE representation of random fields for large-scale Gaussian process regression and statistical finite element analysis

The efficient representation of random fields on geometrically complex domains is crucial for Bayesian modelling in engineering and machine learning, including Gaussian process regression and statistical finite element analysis. Today’s prevalent random field representations are either intended for unbounded domains or are too restrictive in terms of possible field properties. Because of these limitations, new techniques leveraging the historically established link between stochastic PDEs (SPDEs) and random fields have been gaining interest in the statistics and engineering literature. The SPDE representation is especially appealing for engineering applications with complex geometries which already have a finite element discretisation for solving the physical conservation equations. In contrast to the dense covariance matrix of a random field, its inverse, the precision matrix, is usually sparse and equal to the stiffness matrix of an elliptic SPDE. In this paper, we use the SPDE representation to develop a scalable framework for large-scale statistical finite element analysis and Gaussian process (GP) regression on geometrically complex domains. The statistical finite element method (statFEM) introduced by Girolami et al. (2022) is a novel approach for synthesising measurement data and finite element models. In both statFEM and GP regression, we use the SPDE formulation to obtain the relevant prior probability densities with a sparse precision matrix. The properties of the priors are governed by the parameters and possibly fractional order of the SPDE so that we can model on bounded domains and manifolds anisotropic, non-stationary random fields with arbitrary smoothness. We use for assembling the sparse precision matrix the same finite element mesh used for solving the physical conservation equations. The observation models for statFEM and GP regression are such that the posterior probability densities are Gaussians with a closed-form mean and precision. The expressions for the mean vector and the precision matrix do not contain dense matrices and can be evaluated using only sparse matrix operations. We demonstrate the versatility of the proposed framework and its convergence properties with one and two-dimensional Poisson and thin-shell examples.

Classification and prediction of deformed steel and concrete bond-slip failure modes based on SSA-ELM model

Bond-slip failure modes between reinforcement and concrete can seriously affect the load-bearing performance and safety of reinforced concrete (RC) structures. The pull-out test is widely used to analyze RC bond-slip and failure modes because of its simplicity. Therefore, a prediction model based on Sparrow Search Algorithm (SSA) Optimized Extreme Learning Machine (SSA-ELM) is proposed in this study to quickly and effectively evaluate the failure forms of RC structures. The bond-slip test samples containing 399 sets of deformed bar and concrete deformation pull-out tests were first collected as a database, and the various characteristic parameters were preprocessed. Then, the features significantly influencing the failure pattern were screened out using random forest, and the plausibility tests were performed using the Pearson correlation coefficient and mutual information. Subsequently, a classification prediction model of the pull-out failure pattern was developed based on standard machine learning (ML) algorithms. In addition, nine well-known ML algorithms (LR, K-NN, DT, SVC, BPNN, NB, RF, AdaBoost, and XGBoost) are considered. This model's accuracy and generalization ability are compared based on the Precision and Recall confusion matrix. Finally, the optimized ELM was trained by SSA using the training set data with optimized ELM weights and thresholds. The results showed that the concrete protective layer thickness ratio to reinforcement diameter had the highest sensitivity to the failure pattern of RC-drawn specimens. The accuracy of the prediction results of the SSA-ELM model was better than other ML algorithms (Accuracy = 95.8%), such as BP Neural Network. In addition, SSA has better stability and higher accuracy than Gray Wolf Optimization (GWO) Algorithm, Particle Swarm Optimization (PSO) Algorithm, Genetic Algorithm (GA), Ant Colony Optimization (ACO), and other ML classification algorithms.

Sparse Quadratic Approximation for Graph Learning.

Learning graphs represented by M-matrices via an l1-regularized Gaussian maximum-likelihood method is a popular approach, but also one that poses computational challenges for large scale datasets. Recently proposed methods cast this problem as a constrained optimization variant of precision matrix estimation. In this paper, we build on a state-of-the-art sparse precision matrix estimation method and introduce two algorithms that learn M-matrices, that can be subsequently used for the estimation of graph Laplacian matrices. In the first one, we propose an unconstrained method that follows a post processing approach in order to learn an M-matrix, and in the second one, we implement a constrained approach based on sequential quadratic programming. We also demonstrate the effectiveness, accuracy, and performance of both algorithms. Our numerical examples and comparative results with modern open-source packages reveal that the proposed methods can accelerate the learning of graphs by up to 3 orders of magnitude, while accurately retrieving the latent graphical structure of the data. Furthermore, we conduct large scale case studies for the clustering of COVID-19 daily cases and the classification of image datasets to highlight the applicability in real-world scenarios.

The Effect of Spatially Correlated Errors on Sea Surface Height Retrieval from SWOT Altimetry

The upcoming technology of wide-swath altimetry from space will enable monitoring the ocean surface at 4–5 times better spatial resolution and 2–3 times better accuracy than traditional nadir altimeters. This development will provide a chance to directly observe submesoscale sea surface height (SSH) variations that have a typical magnitude of a few centimeters. Taking full advantage of this opportunity requires correct treatment of the correlated SSH errors caused by uncertainties in environmental conditions beneath the satellite and in the geometry and orientation of the on-board interferometer. These observation errors are highly correlated both along and across the surface swath scanned by the satellite, and this presents a significant challenge for accurate processing. In particular, the SWOT precision matrix has off-diagonal elements that are too numerous to allow standard approaches to remain tractable. In this study, we explore the utility of a block-diagonal approximation to the SWOT precision matrix in order to reconstruct SSH variability in the region east of Greenland. An extensive set of 2dVar assimilation experiments demonstrates that the sparse approximation proposed for the precision matrix provides accurate SSH retrievals when the background-to-observation error ratio ν does not exceed 3 and significant wave height is below 2.5 m. We also quantify the range of ν and significant wave heights over which the retrieval accuracy of the exact spatially correlated SWOT error model will outperform the uncorrelated model. In particular, the estimated range is found to be substantially wider (ν<10 with significant wave heights below 8–10 m), indicating the potential benefits of further improving the accuracy of approximations for the SWOT precision matrix.

A Hybridized- Logistic Regression and Deep Learning-based Approaches for Precise Anomaly Detection in Cloud

Anomaly Detection plays a pivot role in determining the abnormal behaviour in the cloud domain. The objective of the manuscript is to present two approaches for Precise Anomaly Detection Approaches by hybridizing RBM with LR and SVM models. The various phases in the present approach are (a) Data collection (b) Pre-processing and normalization; OneHot Encoder for converting categorical values to numerical values followed by encoding the binary features through normalization (c) training the data (d) Building the Feedforward Deep Belief Network (EDBN) using hybridizing Restricted Boltzmann Machine (RBM) with Logistic Regression (LR) and Support Vector Machine (SVM); In the first approach, RBM model is trained through unsupervised pre-training followed by fine-tuning using LR model. In the later approach, RBM model is trained through unsupervised pre-training followed by fine-tuning using SVM model; both the approaches adopt unsupervised pre-training followed by supervised-fine-tuning operations (e) Model Evaluation using the significant parameters such as Precision, Recall, Accuracy, F1-score and Confusion Matrix. The experimental evaluations concluded the effective anomaly detection techniques by integrating the RBM with LR and SVM for capturing the intricate patterns and complex relationships among the data. The proposed approaches paves a path to improved anomaly detection technique, thereby enhancing the security features and anomaly monitoring systems across distinct domains.

Noise cleaning the precision matrix of short time series.

We present a comparison between various algorithms of inference of covariance and precision matrices in small data sets of real vectors of the typical length and dimension of human brain activity time series retrieved by functional magnetic resonance imaging (fMRI). Assuming a Gaussian model underlying the neural activity, the problem consists of denoising the empirically observed matrices to obtain a better estimator of the (unknown) true precision and covariance matrices. We consider several standard noise-cleaning algorithms and compare them on two types of data sets. The first type consists of synthetic time series sampled from a generative Gaussian model of which we can vary the fraction of dimensions per sample q and the strength of off-diagonal correlations. The second type consists of time series of fMRI brain activity of human subjects at rest. The reliability of each algorithm is assessed in terms of test-set likelihood and, in the case of synthetic data, of the distance from the true precision matrix. We observe that the so-called optimal rotationally invariant estimator, based on random matrix theory, leads to a significantly lower distance from the true precision matrix in synthetic data and higher test likelihood in natural fMRI data. We propose a variant of the optimal rotationally invariant estimator in which one of its parameters is optimzed by cross-validation. In the severe undersampling regime (large q) typical of fMRI series, it outperforms all the other estimators. We furthermore propose a simple algorithm based on an iterative likelihood gradient ascent, leading to very accurate estimations in weakly correlated synthetic data sets.

Extremely sparse models of linkage disequilibrium in ancestrally diverse association studies.

Linkage disequilibrium (LD) is the correlation among nearby genetic variants. In genetic association studies, LD is often modeled using large correlation matrices, but this approach is inefficient, especially in ancestrally diverse studies. In the present study, we introduce LD graphical models (LDGMs), which are an extremely sparse and efficient representation of LD. LDGMs are derived from genome-wide genealogies; statistical relationships among alleles in the LDGM correspond to genealogical relationships among haplotypes. We published LDGMs and ancestry-specific LDGM precision matrices for 18 million common variants (minor allele frequency >1%) in five ancestry groups, validated their accuracy and demonstrated order-of-magnitude improvements in runtime for commonly used LD matrix computations. We implemented an extremely fast multiancestry polygenic prediction method, BLUPx-ldgm, which performs better than a similar method based on the reference LD correlation matrix. LDGMs will enable sophisticated methods that scale to ancestrally diverse genetic association data across millions of variants and individuals.

Prediction of Monkeypox on the Skin Lesion with the Siamese Deep Learning Model

One of the viral diseases that started to cause concern in various parts of the world after the COVID-19 pandemic is the monkeypox virus, which has recently emerged. The virus, which was known in previous years and mostly seen in the Western and Central parts of the African continent, has recently begun to affect different human populations in different ways. Monkeypox is transmitted to humans from an animal infected with the virus or from another human being infected with monkeypox. Among the most basic symptoms are high fever, back and muscle aches, chills, and blisters on the skin. These blisters seen on the skin are sometimes confused with chickenpox and measles, and this causes the diagnosis and, accordingly, the treatment process to be wrong. Therefore, the need for computer-aided systems has increased and the need for more robust and reliable approaches has arisen. In this study, using the deep learning model, the distinction of the blisters seen in the body was made and it was decided whether the disease was monkeypox or another disease (chickenpox and measles). The study consisted of three stages. In the first stage, data were obtained and images of both chickenpox and other diseases were used. In the second stage, the Siamese deep learning model was used, and data were classified. In the last stage, the performance of the classifier was evaluated and accordingly accuracy, precision, recall, F1-score, and confusion matrix were used. At the end of the study, an accuracy score of 91.09% was obtained. This result showed that the developed deep learning-based model can be used in this field.

Combined two-level guided wave structural health monitoring strategy using multifeature integration and machine learning: application to early-age grouted sleeves

Structural health monitoring of grouted sleeves is one of the assembly industry’s huge challenges. In this study, a combined two-level damage detection was introduced. It comprises defect classification (healthy, rebar eccentricity, and grout leakage) and severity evaluation for early-age grouted sleeves using guided waves. Multiple features (MF) from time-, frequency-, and time-frequency domains were extracted and defined according to the diverse defects and ages of grouted sleeves to represent complex damage characteristics. Moreover, the egret swarm algorithm optimization–extreme learning machine (ESAO-ELM) models were proposed to avoid the influence of subjective experience and judgment from experts. ESAO optimized the initial random parameters (input weights and hidden layer bias) of ELM. Then, two MF-ESAO-ELM models were trained for two-level damage detection on the experimental dataset. The performance of the proposed models was comprehensively evaluated using accuracy, recall, precision, and confusion matrix. MF-ESAO-ELM performs better than ELM and PSO-ELM in accuracy. In this strategy, the defect classification model works in the outer layer to distinguish the state and types of defects of grouted sleeves (healthy, eccentric, or leakage). In comparison, the inner layer starts predicting the severity only if the defect type is leakage. MF-ESAO-ELM offers advantages in terms of accuracy, strategy, and calculation time.

The cluster D-trace loss for differential network analysis

ABSTRACT A growing literature suggests that gene expression can be greatly altered in disease conditions, and identifying those changes will improve the understanding of complex diseases such as cancers or diabetes. A prevailing direction in the analysis of gene expression studies the changes in gene pathways which include sets of related genes. Therefore, introducing structured exploration to differential analysis of gene expression networks may lead to meaningful discoveries. The topic of this paper is differential network analysis, which focuses on capturing the differences between two or more precision matrices. We discuss the connection between the thresholding method and the D-trace loss method on differential network analysis in the case that the precision matrices share the common connected components. Based on this connection, we further propose the cluster D-trace loss method which directly estimates the differential network and achieves model selection consistency. Simulation studies demonstrate its improved performance and computational efficiency. Finally, the usefulness of our proposed estimator is demonstrated by a real-data analysis on non-small cell lung cancer.

Partial correlation graphical LASSO

AbstractStandard likelihood penalties to learn Gaussian graphical models are based on regularizing the off‐diagonal entries of the precision matrix. Such methods, and their Bayesian counterparts, are not invariant to scalar multiplication of the variables, unless one standardizes the observed data to unit sample variances. We show that such standardization can have a strong effect on inference and introduce a new family of penalties based on partial correlations. We show that the latter, as well as the maximum likelihood, and logarithmic penalties are scale invariant. We illustrate the use of one such penalty, the partial correlation graphical LASSO, which sets an penalty on partial correlations. The associated optimization problem is no longer convex, but is conditionally convex. We show via simulated examples and in two real datasets that, besides being scale invariant, there can be important gains in terms of inference.

Development of a robust UPLC-MS/MS method for the quantification of riluzole in human plasma and its application in pharmacokinetics.

Introduction: The aim of the present study was to establish a simple method for the determination of riluzole in human plasma by ultraperformance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) and apply it for the determination of riluzole in amyotrophic lateral sclerosis (ALS) patients. Methods: Samples were prepared by protein precipitation and were then gradient-eluted on a column of ACQUITY UPLC® HSS T3 by using 0.1% formic acid acetonitrile and 0.1% formic acid water as the mobile phase. Detection was performed on a Xevo TQ-S tandem mass spectrometer in the multiple-reaction monitoring mode using positive electrospray ionization. Validation was performed in the range of 5-800ng/mL. Results and discussion: Three batches of precision accuracy, selectivity, matrix effects, extraction recovery, and stability were also verified and met the requirements. The results showed that the method was reliable and successfully applied to the pharmacokinetics study of riluzole in Chinese amyotrophic lateral sclerosis patients. Meanwhile, in comparison with other prior published methods, our method has the advantages of simple sample preparation, relatively short running time, and small plasma sample consumption, which represented a high-throughput sample determination potential.

TSM-CV: Twitter Sentiment Analysis for COVID-19 Vaccines Using Deep Learning

The coronavirus epidemic has imposed a devastating impact on humans around the globe, causing profound anxiety, fear, and complex emotions and feelings. Vaccination against the new coronavirus has started, and people’s feelings are becoming more diverse and complicated. In the presented work, our goal is to use the deep learning (DL) technique to understand and elucidate their feelings. Due to the advancement of IT and internet facilities, people are socially connected to explain their emotions and sentiments. Among all social sites, Twitter is the most used platform among consumers and can assist scientists to comprehend people’s opinions related to anything. The major goal of this work is to understand the audience’s varying sentiments about the vaccination process by using data from Twitter. We have employed both the historic (All COVID-19 Vaccines Tweets Kaggle dataset) and real (tweets) data to analyze the people’s sentiments. Initially, a preprocessing step is applied to the input samples. Then, we use the FastText approach for computing semantically aware features. In the next step, we apply the Valence Aware Dictionary for sentiment Reasoner (VADER) method to assign the labels to the collected features as being positive, negative, or neutral. After this, a feature reduction step using the Non-Negative Matrix Factorization (NMF) approach is utilized to minimize the feature space. Finally, we have used the Random Multimodal Deep Learning (RMDL) classifier for sentiment prediction. We have confirmed through experimentation that our work is effective in examining the emotions of people toward the COVID-19 vaccines. The presented work has acquired an accuracy result of 94.81% which is showing the efficacy of our strategy. Other standard measures like precision, recall, F1-score, AUC, and confusion matrix are also reported to show the significance of our work. The work is aimed to improve public understanding of coronavirus vaccines which can help the health departments to stop the anti-vaccination leagues and motivate people to a booster dose of coronavirus.