In the fields of environmental science and medicine it is increasingly common to have access to data collected on subjects over time. Given a sufficiently dense sampling, these data can often be smoothed and analyzed as functional variables. While functional variables can be used as covariates in regression models, traditional methods, such as the functional linear model, impose constraints that limit the usefulness of functional covariates as predictors. In this paper we introduce Basis-Expanded Bayesian Additive Regression Trees (BBART), an adaptation to the original BART model that allows for the inclusion of functional covariates. By leveraging the BART model, BBART inherits many attractive features, including requiring no assumption of additivity or smooth effects and enabling posterior inference with MCMC samples.
Monitoring a group of animals would greatly benefit from automated animal re-identification from video. Multiple Object Tracking alone does not provide a sufficiently good re-identification, hence we propose to augment the process by further clustering the output tracks. Unlike datasets for person and vehicle identification, existing animal datasets are not substantial enough to train an advanced model for conventional clustering. In this paper, we present a Classification-Based Clustering method (CBC) which employs track labels and temporal constraints to train a bespoke model for each video dataset. Our proposed method works better than using the tracks alone as animal identities. It also outperforms 13 alternative clustering methods applied to the tracking results.
Tabular data, prevalent in relational databases and spreadsheets, is fundamental across fields like healthcare, engineering, and finance. Despite significant advances in tabular data learning, critical challenges remain: handling missing values, addressing class imbalance, enabling transfer learning, and facilitating feature incremental learning beyond traditional supervised paradigms. We introduce TabMixer, an innovative model that enhances the multilayer perceptron (MLP) mixer architecture to address these challenges. TabMixer incorporates a self-attention mechanism, making it versatile across various learning scenarios including supervised learning, transfer learning, and feature incremental learning. Extensive experiments on eight public datasets demonstrate TabMixer's superior performance over existing state-of-the-art methods. Notably, TabMixer achieved substantial improvements in ANOVA AUC across all scenarios: a 4% increase in supervised learning (0.840 to 0.881), 8% in transfer learning (0.803 to 0.872), and 4% in feature incremental learning (0.806 to 0.843). TabMixer demonstrates high computational efficiency and scalability through reduced floating-point operations and learnable parameters. Moreover, it exhibits strong resilience to missing values and class imbalances through both its architectural design and optional preprocessing enhancements. These results establish TabMixer as a promising model for tabular data analysis and a valuable tool for diverse applications.
In this paper, a subsequence time-series clustering algorithm is proposed to identify the strongly coupled aftershocks sequences and Poissonian background activity from earthquake catalogs of active regions. The proposed method considers the inter-event time statistics between the successive pair of events for characterizing the nature of temporal sequences and observing their relevance with earthquake epicenters and magnitude information simultaneously. This approach categorizes the long-earthquake time series into the finite meaningful temporal sequences and then applies the clustering mechanism to the selective sequences. The proposed approach is built on two phases: (1) a Gaussian kernel-based density estimation for finding the optimal subsequence of given earthquake time-series, and (2) inter-event time ( t ) and distance-based observation of each subsequence for checking the presence of highly correlated aftershock sequences (hot-spots) in it. The existence of aftershocks is determined based on the coefficient of variation (COV). A sliding temporal window on t with earthquake’s magnitude M is applied on the selective subsequence to filter out the presence of time-correlated events and make the meaningful time stationary Poissonian subsequences. This proposed approach is applied to the regional Sumatra-Andaman (2000–2021) and worldwide ISC-GEM (2000–2016) earthquake catalog. Simulation results indicate that meaningful subsequences (background events) can be modeled by a homogeneous Poisson process after achieving a linear cumulative rate and time-independent λ in the exponential distribution of t . The relations COV_a(T)>COV_o(T)> (COV_b(T)≈ 1) and COV_a(d)>COV_o(d)>COV_b(d) are achieved for both studied catalogs. Comparative analysis justifies the competitive performance of the proposed approach to the state-of-art approaches and recently introduced methods.
Interval-valued data have been commonly encountered in practice, and Symbolic Data Analysis provides a solution to the statistical treatment of these data. Regression analysis for interval-valued symbolic data is a topic that has been widely investigated in the literature of symbolic data analysis, and several models from different paradigms have been proposed. There are basic regression assumptions, and it is essential to validate them. This paper introduces an approach to check interval regression model adequacy based on residual analysis. Concepts of ordinary and standardized interval residual are presented, and graphical analysis of these residuals is also proposed. To show the usefulness of the proposed approach, an application for estimating school dropout in the scenario of Brazilian municipalities is performed. We observed some outliers from the interval residuals analysis, and interval robust regression models are more suitable for estimating school dropout.
The random subspace method (RSM) is an ensemble procedure in which each constituent learner is constructed using a randomly chosen subset of the data features. Regression trees are ideal candidate learners in RSM ensembles. By constructing trees upon different feature subsets, RSM reduces correlation between trees resulting in a stronger ensemble. Furthermore, it lessens computational burden by only considering a subset of the features when building each tree. Despite its apparent advantages, RSM has a notable drawback. In some instances a randomly chosen subspace may lack informative features. This is especially true in situations in which the number of truly informative variables is small relative to the total number of variables. Trees that are constructed using feature subsets lacking informative features can be damaging to the ensemble. Here we present grafted random subspace (GRS) and vanishing random subspace (VRS), two novel ensemble procedures designed to remedy the aforementioned drawback by reusing information across trees. Both techniques borrow from RSM by growing individual trees on randomly selected feature subsets. For each tree in a GRS ensemble, the most important variable is identified and guaranteed inclusion into the next q feature subsets. This allows GRS to recycle a promising feature from one tree across several successive trees, effectively grafting the variable into the next q active subsets. In the VRS procedure the least important feature is guaranteed exclusion from the next q feature subsets. This creates a more enriched pool of candidate variables from which the successive feature subsets are drawn.
Zero-shot learning (ZSL) is a transfer learning paradigm that aims to recognize unseen categories just by having a high-level description of them. While deep learning has greatly pushed the limits of ZSL for object classification, ZSL for gesture recognition (ZSGL) remains largely unexplored. Previous attempts to address ZSGL were focused on the creation of gesture attributes and algorithmic improvements, and there is little or no research concerned with feature selection for ZSGL. It is indisputable that deep learning has obviated the need for feature engineering for problems with large datasets. However, when the data are scarce, it is critical to leverage the domain information to create discriminative input features. The main goal of this work is to study the effect of three different feature extraction techniques (velocity, heuristical and latent features) on the performance of ZSGL. In addition, we propose a bilinear auto-encoder approach, referred to as Joint Semantic Encoder (JSE), for ZSGL that jointly minimizes the reconstruction, semantic and classification losses. We conducted extensive experiments to compare and contrast the feature extraction techniques and to evaluate the performance of JSE with respect to existing ZSL methods. For attribute-based classification scenario, irrespective of the feature type, results showed that JSE outperforms other approaches by 5% (p<0.01). When JSE is trained with heuristical features in across-category condition, we showed that JSE significantly outperforms other methods by 5% (p<0.01)).
The world's population is aging, and eldercare services that use smart facilities such as smart homes are widely common in societies now. With the aid of smart facilities, the present study aimed at understanding an elder's moods based on the person’s activities of daily living (ADLs). With this end in view, an explainable probabilistic graphical modeling approach, applying the Bayesian network (BN), was proposed. The proposed BN-based model was capable of defining the relationship between the elder's ADLs and moods in three different levels: Activity-based Feature (AbF), Category of Activity (CoA), and the mood state. The model also allowed us to explain the transformations among the different levels/nodes on the defined BNs. A framework featured with smart facilities, including a smart home, a smartphone, and a wristband, was utilized to assess the model. The smart home was an elderly woman's house, equipped with a set of binary-based sensors. For about five months, the ADLs' data have been recorded through daily behavioral-based information, registered by experts using a defined questionnaire. The obtained results proved that the proposed BN-based model of the current study could promisingly estimate the elder's moods and CoA states. Moreover, in contrast to the machine learning techniques that behave like a black box, the effect of each feature from the lower levels to the higher levels of information of the BNs can be traced. Implications of the findings for future diagnosis and treatment of the elderly are considered.
This paper presents a novel approach for iris recognition through symbolic modeling and symbolic similarity analysis of iris features, which are transformed using Savitzky-Golay filter. The proposed approach is based on iris Savitzky-Golay filter energy features. Prior to symbolic modeling of iris feature data, Canny edge detector and Hough Transform technique is used to segment iris region from the eye image. In the normalization stage, Daugman's rubber sheet model is used to obtain a fixed size rectangular block from segmented iris region to account for imaging inconsistencies. Using Savitzky-Golay filter, iris features are obtained by computing smoothing coefficients from normalized iris region. Further, the extracted features are represented as symbolic object by employing symbolic data modeling approach. The symbolic similarity analysis technique is employed for computing similarity between probe iris image and symbolic objects. The symbolic object which represents maximum similarity is considered as a resultant object. The SGGSIE&T and CASIA-Iris 4.0 Interval databases were selected to evaluate the performance of proposed model and achieved 99.33% and 95% recognition rates respectively.
With the rapid development of computer technology, data collection becomes easier, and data object presents more complex. Data analysis method based on machine learning is an important, active, and multi-disciplinarily research field. Support vector machine (SVM) is one of the most powerful and fast classification models. The main challenges SVM faces are the selection of feature subset and the setting of kernel parameters. To improve the performance of SVM, a metaheuristic algorithm is used to optimize them simultaneously. This paper first proposes a novel classification model called IBMO-SVM, which hybridizes an improved barnacle mating optimizer (IBMO) with SVM. Three strategies, including Gaussian mutation, logistic model, and refraction-learning, are used to improve the performance of BMO from different perspectives. Through 23 classical benchmark functions, the impact of control parameters and the effectiveness of introduced strategies are analyzed. The convergence accuracy and stability are the main gains, and exploration and exploitation phases are more properly balanced. We apply IBMO-SVM to 20 real-world datasets, including 4 extremely high-dimensional datasets. Experimental results are compared with 6 state-of-the-art methods in the literature. The final statistical results show that the proposed IBMO-SVM achieves a better performance than the standard BMO-SVM and other compared methods, especially on high-dimensional datasets. In addition, the proposed model also shows significant superiority compared with 4 other classifiers.
[This corrects the article DOI: 10.1007/s10044-020-00950-0.].
Coronavirus (COVID-19) is one of the most serious problems that has caused stopping the wheel of life all over the world. It is widely spread to the extent that hospital places are not available for all patients. Therefore, most hospitals accept patients whose recovery rate is high. Machine learning techniques and artificial intelligence have been deployed for computing infection risks, performing survival analysis and classification. Survival analysis (time-to-event analysis) is widely used in many areas such as engineering and medicine. This paper presents two systems, Cox_COVID_19 and Deep_ Cox_COVID_19 that are based on Cox regression to study the survival analysis for COVID-19 and help hospitals to choose patients with better chances of survival and predict the most important symptoms (features) affecting survival probability. Cox_COVID_19 is based on Cox regression and Deep_Cox_COVID_19 is a combination of autoencoder deep neural network and Cox regression to enhance prediction accuracy. A clinical dataset for COVID-19 patients is used. This dataset consists of 1085 patients. The results show that applying an autoencoder on the data to reconstruct features, before applying Cox regression algorithm, would improve the results by increasing concordance, accuracy and precision. For Deep_ Cox_COVID_19 system, it has a concordance of 0.983 for training and 0.999 for testing, but for Cox_COVID_19 system, it has a concordance of 0.923 for training and 0.896 for testing. The most important features affecting mortality are, age, muscle pain, pneumonia and throat pain. Both Cox_COVID_19 and Deep_ Cox_COVID_19 prediction systems can predict the survival probability and present significant symptoms (features) that differentiate severe cases and death cases. But the accuracy of Deep_Cox_Covid_19 outperforms that of Cox_Covid_19. Both systems can provide definite information for doctors about detection and intervention to be taken, which can reduce mortality.
Since the emergence of COVID-19, thousands of people undergo chest X-ray and computed tomography scan for its screening on everyday basis. This has increased the workload on radiologists, and a number of cases are in backlog. This is not only the case for COVID-19, but for the other abnormalities needing radiological diagnosis as well. In this work, we present an automated technique for rapid diagnosis of COVID-19 on computed tomography images. The proposed technique consists of four primary steps: (1) data collection and normalization, (2) extraction of the relevant features, (3) selection of the most optimal features and (4) feature classification. In the data collection step, we collect data for several patients from a public domain website, and perform preprocessing, which includes image resizing. In the successive step, we apply discrete wavelet transform and extended segmentation-based fractal texture analysis methods for extracting the relevant features. This is followed by application of an entropy controlled genetic algorithm for selection of the best features from each feature type, which are combined using a serial approach. In the final phase, the best features are subjected to various classifiers for the diagnosis. The proposed framework, when augmented with the Naive Bayes classifier, yields the best accuracy of 92.6%. The simulation results are supported by a detailed statistical analysis as a proof of concept.
In manifold learning, the intrinsic geometry of the manifold is explored and preserved by identifying the optimal local neighborhood around each observation. It is well known that when a Riemannian manifold is unfolded correctly, the observations lying spatially near to the manifold, should remain near on the lower dimension as well. Due to the nonlinear properties of manifold around each observation, finding such optimal neighborhood on the manifold is a challenge. Thus, a sub-optimal neighborhood may lead to erroneous representation and incorrect inferences. In this paper, we propose a rotation-based affinity metric for accurate graph Laplacian approximation. It exploits the property of aligned tangent spaces of observations in an optimal neighborhood to approximate correct affinity between them. Extensive experiments on both synthetic and real world datasets have been performed. It is observed that proposed method outperforms existing nonlinear dimensionality reduction techniques in low-dimensional representation for synthetic datasets. The results on real world datasets like COVID-19 prove that our approach increases the accuracy of classification by enhancing Laplacian regularization.
The 2019 novel coronavirus disease (COVID-19), with a starting point in China, has spread rapidly among people living in other countries and is approaching approximately 101,917,147 cases worldwide according to the statistics of World Health Organization. There are a limited number of COVID-19 test kits available in hospitals due to the increasing cases daily. Therefore, it is necessary to implement an automatic detection system as a quick alternative diagnosis option to prevent COVID-19 spreading among people. In this study, five pre-trained convolutional neural network-based models (ResNet50, ResNet101, ResNet152, InceptionV3 and Inception-ResNetV2) have been proposed for the detection of coronavirus pneumonia-infected patient using chest X-ray radiographs. We have implemented three different binary classifications with four classes (COVID-19, normal (healthy), viral pneumonia and bacterial pneumonia) by using five-fold cross-validation. Considering the performance results obtained, it has been seen that the pre-trained ResNet50 model provides the highest classification performance (96.1% accuracy for Dataset-1, 99.5% accuracy for Dataset-2 and 99.7% accuracy for Dataset-3) among other four used models.
The aim of this research is to develop a fusion concept to component-based face recognition algorithms for features analysis of binary facial components (BFCs), which are invariant to illumination, expression, pose variations and partial occlusion. To analyze the features, using statistical pattern matching concepts, which are the combination of Chi-square (CSQ), Hu moment invariants (HuMIs), absolute difference probability of white pixels (AbsDifPWPs) and geometric distance values (GDVs) have been proposed for face recognition. The individual grayscale face image is cropped by applying the Viola–Jones face detection algorithm from a face database having variations in illumination, appearance, pose and partial occlusion with complex backgrounds. Doing illumination correction through histogram linearization technique, the grayscale face components such as eyes, nose and mouth regions are extracted using the 2D geometric positions. The binary face image is created by applying cumulative probability distribution function with Otsu adaptive thresholding method and then extracted BFCs such as eyes, nose and mouth regions. Five statistical pattern matching tools such as the standard deviation of CSQ values with probability of white pixels (PWPs), standard deviation of HuMIs with Hu’s seven moment invariants, AbsDifPWPs and GDVs are developed for recognition purpose. GDVs are determined between two similar facial corner points (FCPs) and nine FCPs are extracted from binary whole face and BFCs. Pixel Intensity Values (PIVs) which are determined using L2 norms from grayscale values of the whole face and grayscale values of the face components. Experiment is performed using BioID Face Database on the basis of these pattern matching tools and appropriate threshold values with logical and conditional operators and gives the best expected results from true positive rate perspective.
Physical traits such as the shape of the hand and face can be used for human recognition and identification in video surveillance systems and in biometric authentication smart card systems, as well as in personal health care. However, the accuracy of such systems suffers from illumination changes, unpredictability, and variability in appearance (e.g. occluded faces or hands, cluttered backgrounds, etc.). This work evaluates different statistical and chrominance models in different environments with increasingly cluttered backgrounds where changes in lighting are common and with no occlusions applied, in order to get a reliable neural network reconstruction of faces and hands, without taking into account the structural and temporal kinematics of the hands. First a statistical model is used for skin colour segmentation to roughly locate hands and faces. Then a neural network is used to reconstruct in 3D the hands and faces. For the filtering and the reconstruction we have used the growing neural gas algorithm which can preserve the topology of an object without restarting the learning process. Experiments conducted on our own database but also on four benchmark databases (Stirling’s, Alicante, Essex, and Stegmann’s) and on deaf individuals from normal 2D videos are freely available on the BSL signbank dataset. Results demonstrate the validity of our system to solve problems of face and hand segmentation and reconstruction under different environmental conditions.
The automated identification of brain structure in Magnetic Resonance Imaging is very important both in neuroscience research and as a possible clinical diagnostic tool. In this study, a novel strategy for fully automated hippocampal segmentation in MRI is presented. It is based on a supervised algorithm, called RUSBoost, which combines data random undersampling with a boosting algorithm. RUSBoost is an algorithm specifically designed for imbalanced classification, suitable for large data sets because it uses random undersampling of the majority class. The RUSBoost performances were compared with those of ADABoost, Random Forest and the publicly available brain segmentation package, FreeSurfer. This study was conducted on a data set of 50 T1-weighted structural brain images. The RUSBoost-based segmentation tool achieved the best results with a Dice’s index of \(0.88 \pm 0.01\) (\(0.87 \pm 0.01\)) for the left (right) brain hemisphere. An independent data set of 50 T1-weighted structural brain scans was used for an independent validation of the fully trained strategies. Again the RUSBoost segmentations compared favorably with manual segmentations with the highest performances among the four tools. Moreover, the Pearson correlation coefficient between hippocampal volumes computed by manual and RUSBoost segmentations was 0.83 (0.82) for left (right) side, statistically significant, and higher than those computed by Adaboost, Random Forest and FreeSurfer. The proposed method may be suitable for accurate, robust and statistically significant segmentations of hippocampi.
Considering the main disadvantage of the existing gaze point estimation methods which restrict user's head movement and have potential injury on eyes, we propose a gaze point estimation method based on facial normal and binocular vision. Firstly, we calibrate stereo cameras to determine the extrinsic and intrinsic parameters of the cameras; Secondly, face is quickly detected by Viola-Jones framework and the center position of the two irises can be located based on integro-differential operators; The two nostrils and mouth are detected based on the saturation difference and their 2D coordinates can be calculated; Thirdly, the 3D coordinates of these five points are obtained by stereo matching and 3D reconstruction; After that, a plane fitting algorithm based on least squares is adopted to get the approximate facial plane, then, the normal via the midpoint of the two pupils can be figured out; Finally, the point-of-gaze can be obtained by getting the intersection point of the facial normal and the computer screen. Experimental results confirm the accuracy and robustness of the proposed method.
A large variety of well-known scale-invariant texture recognition methods is tested with respect to their scale invariance. The scale invariance of these methods is estimated by comparing the results of two test setups. In the first test setup, the images of the training and evaluation set are acquired under same scale conditions and in the second test setup, the images in the evaluation set are gathered under different scale conditions than those of the training set. For the first test setup, scale invariance is not needed, whereas for the second test setup, scale invariance is obviously crucial. The difference between the results of these two test setups indicates the scale invariance of a method (the higher the scale invariance the lower the difference). The scale invariance of the methods is additionally estimated by analyzing the similarity of the feature vectors of images and their scaled versions. Additionally to the scale invariance, we also test eventual viewpoint and illumination invariance of the methods. As texture databases for our tests we use the KTH-TIPS database and the CUReT database. Results imply that many of the considered methods are not as scale-invariant as expected.
