Search

Monitoring physiological changes [e.g., heart rate (HR), respiration, and HR variability (HRV)] is important for measuring human emotions. Physiological responses are more reliable and harder to alter compared to explicit behaviors (such as facial expressions and speech), but they require special contact sensors to obtain. Research in the last decade has shown that photoplethysmography (PPG) signals can be remotely measured (rPPG) from facial videos under ambient light, from which physiological changes can be extracted. This promising finding has attracted much interest from researchers, and the field of rPPG measurement has been growing fast. In this article, we review current progress on intelligent signal processing approaches for rPPG measurement, including earlier works on unsupervised approaches and recently proposed supervised models, benchmark data sets, and performance evaluation. We also review studies on rPPG-based affective applications and compare them with other affective computing modalities. We conclude this article by emphasizing the current main challenges and highlighting future directions.

A consistent implementation of Lees-Edwards boundary conditions is proposed for the Multi-Sphere Discrete Element Method, which can mitigate various unphysical effects at the bulk and micro-structural levels. These effects include non-linear velocity profiles and inhomogeneous particle distributions, which result in significant errors with respect to pressure and granular temperature. In order to allow for a fair assessment of different implementations, a novel compound sphere particle shape is devised for comparison to reliable benchmark data generated from systems of spherical particles. The Multi-Sphere Discrete Element Method is utilised to examine two implementations of these conditions. The commonly used Naive approach results in the aforementioned unphysical effects, which are numerical artefacts causing deviations from the benchmark results of up to one order of magnitude. Meanwhile, the proposed consistent implementation fulfils the fundamental requirements of Lees-Edwards boundary conditions and produces data which are in excellent agreement with the benchmark results, as well as the available literature. Comparing the aforementioned implementations, general principles are developed for implementing Lees-Edwards boundary conditions for the Multi-Sphere Discrete Element Method.

The fuzzy cognitive map (FCM) is an effective tool for modeling and simulating complex dynamic systems. The research on the problem of learning FCM from the available time series is outstanding. Many batch FCM learning methods have been proposed to address this issue and the performance of these methods is satisfactory. However, these batch-learning methods are difficult to cope with large-scale data sets (for example, the memory in computers is not enough to store all instances) and real-time streaming data, leading to the failure of real-time and online analysis of complex systems. In this article, unlike the existing batch learning methods, such as evolutionary and regression-based methods, we first extend the FCM learning to an online setting, and then develop an effective algorithm based on a follow-the-regularized-leader (FTRL)-proximal style learning algorithm to address the online FCM learning problem, termed as OFCM. The performance of OFCM is validated on constructed benchmark data sets, including synthetic data sets, and gene regulatory network reconstruction data sets. The experimental results demonstrate the merits of OFCM, which can effectively solve online FCM learning problems.

This article reflects research in the field of artificial intelligence and demonstrates a higher efficiency achievement of conventional clustering methods in combination with unconventional methods. It concerns a new hybrid approach based on the SOM (Self-Organizing Maps) method. We focused on the possibility of combining SOM with other clustering methods—CLARA, CURE a K-means. Method SOM is primarily useful in the first phases of the process, where knowledge of the data is too vague. It is thus followed by the use of a selected clustering algorithm. It then works with preprocessed data. Its performance, compared with its outputs on unprocessed data, is more efficient, which is proved by the performed experimental study on the benchmark data set Fundamental Clustering Problems Suite (FCPS). Part of the experimental verification was also a comparison of the achieved outputs with other approaches using this dataset based on a standard metrics—Rand index.

Pattern classification is a fundamental problem in many data-driven application domains. New-labeled data refers to the data with the labels that are new and different from source labels. How to learn the new-labeled data is a crucial research in the data classification. In this paper, an evolved fuzzy min-max neural network for new-labeled data classification (FMM-NLA) is proposed. In FMM-NLA, the network can be self-evolved. Unlike the traditional FMM methods, the trained network of FMM-NLA can be expanded when new-labeled data added. FMM-NLA is a continuing-learning method, which can realize the continuing training process without retraining all the data. In order to verify the superiority of the proposed method, benchmark data sets are used. The experimental results show that FMM-NLA is effective in handling new-labeled data. Moreover, the application result on the pipeline defect recognition in depth shows that FMM-NLA is effective in solving the new-labeled defect recognition problem.

Pansharpening aims to sharpen a lowspatial-resolution (LR) multispectral (MS) image using a high-spatial-resolution (HR) panchromatic (Pan) image to obtain the HR MS image. It has been a fundamental and active research topic in remote sensing, and pansharpening methods have been developed for nearly 40 years. While the performance evaluation of pansharpening methods is still based on a small number of individual images, datadriven pansharpening approaches are attracting increasing attention. However, few publicly available benchmark data sets for pansharpening are available, especially large-scale ones. This has been a serious limitation for the future development of pansharpening methods.

Estimating the optimal number of clusters (NC) is pivotal in cluster analysis. From the viewpoint of sample geometry, a novel internal clustering validity index, which is termed the between-within cluster (BWC) index, is designed in this paper. Moreover, a method is proposed to estimate the optimal NC. The BWC index improves the well-known Silhouette index. BWC validates the clustering results from a certain clustering algorithm (e.g., affinity propagation or hierarchical) and estimates the optimal NC for many kinds of data sets, including synthetic data sets, benchmark data sets, UCI data sets, gene expression data sets, and images. Theoretical analysis and experimental studies demonstrate the effectiveness and high efficiency of the new index and method.

Purpose The aim of this study is to propose an efficient rule extraction and integration approach for identifying phishing websites. The proposed approach can elucidate patterns of phishing websites and identify them accurately. Design/methodology/approach Hyperlink indicators along with URL-based features are used to build the identification model. In the proposed approach, very simple rules are first extracted based on individual features to provide meaningful and easy-to-understand rules. Then, the F-measure score is used to select high-quality rules for identifying phishing websites. To construct a reliable and promising phishing website identification model, the selected rules are integrated using a simple neural network model. Findings Experiments conducted using self-collected and benchmark data sets show that the proposed approach outperforms 16 commonly used classifiers (including seven non–rule-based and four rule-based classifiers as well as five deep learning models) in terms of interpretability and identification performance. Originality/value Investigating patterns of phishing websites based on hyperlink indicators using the efficient rule-based approach is innovative. It is not only helpful for identifying phishing websites, but also beneficial for extracting simple and understandable rules.

Human beings may make random guesses in decision-making. Occasionally, their guesses may generate consistency with the real situation. This kind of consistency is termed random consistency. In the area of machine leaning, the randomness is unavoidable and ubiquitous in learning algorithms. However, the accuracy (A), which is a fundamental performance measure for machine learning, does not recognize the random consistency. This causes that the classifiers learnt by A contain the random consistency. The random consistency may cause an unreliable evaluation and harm the generalization performance. To solve this problem, the pure accuracy (PA) is defined to eliminate the random consistency from the A. In this paper, we mainly study the necessity, learning consistency and leaning method of the PA. We show that the PA is insensitive to the class distribution of classifier and is more fair to the majority and the minority than A. Subsequently, some novel generalization bounds on the PA and A are given. Furthermore, we show that the PA is Bayes-risk consistent in finite and infinite hypothesis space. We design a plug-in rule that maximizes the PA, and the experiments on twenty benchmark data sets demonstrate that the proposed method performs statistically better than the kernel logistic regression in terms of PA and comparable performance in terms of A. Compared with the other plug-in rules, the proposed method obtains much better performance.

For incremental machine-learning applications it is often important to robustly estimate the system accuracy during training, especially if humans perform the supervised teaching. Cross-validation and interleaved test/train error are here the standard supervised approaches. We propose a novel semi-supervised accuracy estimation approach that clearly outperforms these two methods. We introduce the Configram Estimation (CGEM) approach to predict the accuracy of any classifier that delivers confidences. By calculating classification confidences for unseen samples, it is possible to train an offline regression model, capable of predicting the classifier&rsquo, s accuracy on novel data in a semi-supervised fashion. We evaluate our method with several diverse classifiers and on analytical and real-world benchmark data sets for both incremental and active learning. The results show that our novel method improves accuracy estimation over standard methods and requires less supervised training data after deployment of the model. We demonstrate the application of our approach to a challenging robot object recognition task, where the human teacher can use our method to judge sufficient training.

As a single hidden layer feed-forward neural network, the extreme learning machine (ELM) has been extensively studied for its short training time and good generalization ability. Recently, with the deep learning algorithm becoming a research hotspot, some deep extreme learning machine algorithms such as multi-layer extreme learning machine (ML-ELM) and hierarchical extreme learning machine (H-ELM) have also been proposed. However, the deep ELM algorithm also has many shortcomings: (1) when the number of model layers is shallow, the random feature mapping makes the sample features cannot be fully learned and utilized; (2) when the number of model layers is deep, the validity of the sample features will decrease after continuous abstraction and generalization. In order to solve the above problems, this paper proposes a densely connected deep ELM algorithm: dense-HELM (D-HELM). Benchmark data sets of different sizes have been employed for the property of the D-HELM algorithm. Compared with the H-ELM algorithm on the benchmark dataset, the average test accuracy is increased by 5.34% and the average training time is decreased by 21.15%. On the NORB dataset, the proposed D-HELM algorithm still maintains the best classification results and the fastest training speed. The D-HELM algorithm can make full use of the features of hidden layer learning by using the densely connected network structure and effectively reduce the number of parameters. Compared with the H-ELM algorithm, the D-HELM algorithm significantly improves the recognition accuracy and accelerates the training speed of the algorithm.

Extreme learning machine (ELM) is one of the most remarkable machine learning algorithm in consequence of superior properties particularly its speed. ELM algorithm tends to have some drawbacks like instability and poor generalization performance in the presence of perturbation and multicollinearity. This paper introduces a novel algorithm based on Liu regression estimator (L-ELM) to handle these drawbacks. Different selection approaches have been used to determine the appropriate Liu biasing parameter. The new algorithm is tested against the basic ELM, RR-ELM, AUR-ELM and OP-ELM on nine well-known benchmark data sets. Statistical significance tests have been carried out. Experimental results show that L-ELM for at least one Liu biasing parameter generally outperforms basic ELM, RR-ELM, AUR-ELM and OP-ELM in terms of stability and generalization performance with a little lost of speed. Conversely, the training time of L-ELM is generally much slower than RR-ELM, AUR-ELM and OP-ELM. Consequently, the proposed algorithm can be considered a powerful alternative to avoid the loss of performance in regression studies

The demand to increase port throughput has driven container ships to travel relatively fast in shallow water whilst avoiding grounding and hence, there is need for more accurate high-speed squat predictions. A study has been undertaken to determine the most suitable method to predict container ship squat when travelling at relatively high speeds (Frh ≥ 0.5) in finite water depth (1.1 ≤ h/T ≤ 1.3). The accuracy of two novel self-propelled URANS CFD squat model are compared with that of readily available empirical squat prediction formulae. Comparison of the CFD and empirical predictions with benchmark data demonstrates that for very low water depth (h/T < 1.14) and when Frh < 0.46; Barass II (1979), ICORELS (1980), and Millward’s (1992) formulae have the best correlation with benchmark data for all cases investigated. However, at relatively high speeds (Frh ≥ 0.5) which is achievable in deeper waters (h/T ≥ 1.14), most of the empirical formulae severely underestimated squat (7-49%) whereas the quasi-static CFD model presented has the best correlation. The changes in wave patterns and effective wake fraction with respect to h/T are also presented.

For the split delivery vehicle routing problem in the distribution link of logistics where the optimization objective is to minimize the total distance of the transportation path, a hybrid discrete differential evolution (HDDE) algorithm is adopted in this study. The algorithm applies the encoding and decoding method based on the remaining vehicle capacity to split the demand of customer point and adjust the splitting node to the optimal position of its path. In order to avoid the prematurity of the algorithm, the perturbation mechanism is added, i.e. , the three-point insertion method is used in the crossover stage. To ensure the higher quality of the solution, the idea of tabu search is introduced in three cases: initial population generation, selection operation and local optimal situation. Finally, through the numerical experiments based on benchmark data, the HDDE algorithm is compared with other competitive algorithms to demonstrate its domination.

Support vector regression (SVR) is one of the most powerful and widely used machine learning algorithms regarding prediction. The kernel type, penalty factor and other parameters influence the efficiency and performance of SVR deeply. The optimization of these parameters is held a hot issue. In this work, we propose a SVR based prediction approach using henry gas solubility optimization algorithm (HGSO), which is a recent meta-heuristic algorithm inspired by Henry's law. First, SVR parameters are randomly generated in some certain ranges to form parameter population. Second, the prediction accuracies (PAs) are obtained using the population and SVR. Thirdly, the population and optimal SVR parameters are updated via PAs and HGSO. We repeat the second and third steps until the cut-off conditions are met. Ten low- and high-dimensional benchmark data sets are utilized to assess the prediction accuracy, convergence performance and computational complexity of the presented approach and other well-known algorithms. The experimental results reveal that our approach has the optimum comprehensive performance.

Next basket recommendation is a challenging problem, mainly due to the relationships among the items in a basket almost not being considered in current research. In this article, we address next basket recommendation with a novel deep learning architecture. In particular, we consider both the short-term user interests and the long-term user preferences, and we design a new attention that considers the relationships among the items in a basket. We extensively evaluated the proposed model on two benchmark data sets, the Ta-Feng and JingDong datasets. The experimental results show that the proposed model outperforms several state-of-the-art next basket recommendation models. In the experiments of the modules' effects, we also verify the effectiveness of each module. The model significantly improves the NDCG by 25.5 percentage points and 5 percentage points when compared with the uncompleted network model on the JingDong and Ta-Feng datasets, respectively; while in terms of the F1, the performance is improved by 14.6 and 23.4 percentage points, respectively.

In this paper we propose a novel method for learning how algorithms perform. Classically, algorithms are compared on a finite number of existing (or newly simulated) benchmark datasets based on some fixed metrics. The algorithm(s) with the smallest value of this metric are chosen to be the `best performing'. We offer a new approach to flip this paradigm. We instead aim to gain a richer picture of the performance of an algorithm by generating artificial data through genetic evolution, the purpose of which is to create populations of datasets for which a particular algorithm performs well on a given metric. These datasets can be studied so as to learn what attributes lead to a particular progression of a given algorithm. Following a detailed description of the algorithm as well as a brief description of an open source implementation, a case study in clustering is presented. This case study demonstrates the performance and nuances of the method which we call Evolutionary Dataset Optimisation. In this study, a number of known properties about preferable datasets for the clustering algorithms known as (k)-means and DBSCAN are realised in the generated datasets., 33 pages, 15 figures

Learning from the lack of labeled data is a challenging task which often limits the performance of the classifier. Since the unlabeled data is easy to obtain, using both of the labeled and unlabeled data in the training process provide a way to solve this problem. In this paper, a semi-supervised classification method based on fuzzy min–max neural network (SS-FMM) is proposed. In SS-FMM, the network has been modified for handling both of the labeled and unlabeled data. In addition, the staged feedback process is designed to modify the network structure of the traditional fuzzy min–max neural network. A staged-threshold function designed in SS-FMM, the hyperbox pruning process and the hyperbox relabeling process can be started dynamically. Moreover, the hyperboxes relabeling process and the hyperbox pruning process are designed to maximize using the unlabeled data and control the amount of the hyperboxes. In order to testify the effectiveness of SS-FMM, various experiments are carried out with several benchmark data sets. In addition, SS-FMM has been applied on the internal inspection data of our system. The results show that SS-FMMM has got good performance.

Protein complexes are the fundamental units for many cellular processes. Identifying protein complexes accurately is critical for understanding the functions and organizations of cells. With the increment of genome-scale protein–protein interaction (PPI) data for different species, various computational methods focus on identifying protein complexes from PPI networks. In this article, we give a comprehensive and updated review on the state-of-the-art computational methods in the field of protein complex identification, especially focusing on the newly developed approaches. The computational methods are organized into three categories, including cluster-quality-based methods, node-affinity-based methods and ensemble clustering methods. Furthermore, the advantages and disadvantages of different methods are discussed, and then, the performance of 17 state-of-the-art methods is evaluated on two widely used benchmark data sets. Finally, the bottleneck problems and their potential solutions in this important field are discussed.

Support vector machine (SVM) is a powerful technique in pattern classification, but its performance is highly dependent on its parameters. In this paper, a new SVM optimized by a novel differential evolution (DE) with a hybrid parameter setting strategy and a population size adaptation method is proposed and simplified as FDE-PS-SVM. In the hybrid parameter setting strategy, the SVM parameter offspring are generated by DE operators with evolutionary parameters that are fixed or with the ones generated by fuzzy logic inference (FLI) according to a given probability. In the population size adaptation method, the population size is shrunk gradually during the search, which tries to balance the diversity and concentration ability of the algorithm to find better SVM parameters. Some benchmark data sets are used to evaluate the proposed algorithm. Experimental results show that the two proposed strategies are effective to search for better SVM parameters while the proposed FDE-PS-SVM algorithm outperforms other algorithms published in other literature.

Correlation filter (CF) theory has gained a sustainable attractiveness on the tracking field by virtue of its efficiency in training and decision stage. The seminal idea of cyclic shift operations on image patch makes the CF-based trackers accomplish the dense sampling scheme in an efficient manner. Due to all shifted samples merely provide the single appearance of the target in the previous frame, the filters cannot acquire sufficient information of the target appearance, which leads to the tracking failed. To tackle this problem, we designed a novel tracker using collaborative correlation filters, which collaboratively exploit the historical appearances of the target and the surrounding contexts of the previous frame for the filters learning. Therefore, the collaborative correlation filter simultaneously acquires the variations in target appearance and promotes discriminability when the target suffers from occlusion or motion blur. In addition, a unified optimization procedure is proposed to enable abundant information to be embedded in filter learning, which obtains a closed-form solution. Since the historical appearances of the target are collected from previous tracking results and the target appearance may change drastically or suffer from occlusion, using all of these templates for training is inappropriate. To eliminate the useless samples from the template set, we provide a template updating strategy based on collaborative representation. The experimental results on the standard $$\hbox {OTB}100$$ benchmark data sets illustrate the robustness of the proposed tracker.

Purpose In recent years, the application of metaheuristics in training neural network models has gained significance due to the drawbacks of deterministic algorithms. This paper aims to propose the use of a recently developed “memory based hybrid dragonfly algorithm” (MHDA) for training multi-layer perceptron (MLP) model by finding the optimal set of weight and biases. Design/methodology/approach The efficiency of MHDA in training MLPs is evaluated by applying it to classification and approximation benchmark data sets. Performance comparison between MHDA and other training algorithms is carried out and the significance of results is proved by statistical methods. The computational complexity of MHDA trained MLP is estimated. Findings Simulation result shows that MHDA can effectively find the near optimum set of weight and biases at a higher convergence rate when compared to other training algorithms. Originality/value This paper presents MHDA as an alternative optimization algorithm for training MLP. MHDA can effectively optimize set of weight and biases and can be a potential trainer for MLPs.

As a high-efficiency training method for single layer feedforward neural networks, extreme learning machine (ELM) has drawn much interest recently, but its robustness is not good due to the adoption of the square loss function. Hence, the convex loss function in ELM is replaced with a nonconvex loss function, i.e., the C-loss function, so a novel algorithm, called C-loss based ELM (CELM), is proposed in this paper. According to the experimental results on a toy example and two benchmark data sets, CELM performs better than the other algorithms with respect to the generalization performance. To be more important, when CELM is used to estimate power of small-scale turbojet engine, it also dominates the other algorithms, which makes the development of the potential control structure, viz., the direct power control, for the unmanned aerial vehicles more feasible.

This paper introduces a newly developed segregated algorithm (named SIMPLERR algorithm) to overcome two major drawbacks in the SIMPLE-family algorithm. The full coupling between velocity and pressure is ensured in an easy segregated manner. Comprehensive comparisons among the SIMPLER algorithm, the IDEAL algorithm and the SIMPLERR algorithm are carried out via four benchmark incompressible fluid flow and heat transfer problems analyzed with unstructured grids. The results demonstrate that the IDEAL algorithm and the SIMPLERR algorithm have substantial advantages over the SIMPLER algorithm in both convergence rate and robustness, moreover, compared with the IDEAL algorithm, the SIMPLERR algorithm can offer a higher computational efficiency for its simplicity, especially for high-Re/Ra and fine-mesh cases; the computational results given by the present algorithm agree very well with benchmark data.

An alternative manner is provided for solving the classical linearised problem of the radiation and diffraction of regular water waves caused by oscillation of a floating body in deep water. It is shown that the singular wave integrals of the three-dimensional free surface Green function G and its gradient ∇ G can be regarded as regular wave integrals and are integrated directly. The method is validated by comparing with benchmark data for a floating or submerged body undergoing oscillatory wave motions. The comparison shows that the evaluation is sufficiently accurate for practical purposes. As the significance of the method, the numerical approximation stability for the gradient ∇ G is shown to be the same with that for G .

In this paper, we address the problem of compliant geo-distributed query processing. In particular, we focus on dataflow policies that impose restrictions on movement of data across geographical or institutional borders. Traditional ways to distributed query processing do not consider such restrictions and therefore in geo-distributed environments may lead to non-compliant query execution plans. For example, an execution plan for a query over data sources from Europe, North America, and Asia, which may otherwise be optimal, may not comply with dataflow policies as a result of shipping some restricted (intermediate) data. We pose this problem of compliance in the setting of geo-distributed query processing. We propose a compliance-based query optimizer that takes into account dataflow policies, which are declaratively specified using our policy expressions, to generate compliant geo-distributed execution plans. Our experimental study using a geo-distributed adaptation of the TPC-H benchmark data indicates that our optimization techniques are effective in generating efficient compliant plans and incur low overhead on top of traditional query optimizers.

Multi-dimensional classification (MDC) deals with the problem where each instance is associated with multiple class variables, each of which corresponds to a specific class space. One of the mainstream solutions for MDC is to adapt traditional machine learning techniques to deal with MDC data. In this paper, a first attempt towards adapting instance-based techniques for MDC is investigated, and a new approach named Md-knn is proposed. Specifically, Md-knn identifies unseen instance's $k$ nearest neighbors and obtains its corresponding $k\text{NN}$ counting statistics for each class space, based on which maximum a posteriori (MAP) inference is made for each pair of class spaces. After that, the class label w.r.t. each class space is determined by synergizing predictions from the learned classifiers via consulting empirical $k\text{NN}$ accuracy. Comparative studies over ten benchmark data sets clearly validate Md-knn's effectiveness.

Research upon artificial neural networks has yielded regularization techniques in order to reduce overfitting, reduce weight increase rate, and in consequence improve model performance. One of the commonly used regularization technique is dropout. In order to make the technique more resistant to emerging interdependencies while learning, the dropout has been generalized to its improved version called Soft Dropout. In this paper we have proposed an extended version of Soft Dropout called Contextual Soft Dropout. This technique is dedicated to be used with Contextual Neural Networks and utilizing their multi-step aggregation functions. Experimental results have shown improvement in overfitting decrease and have achieved lower logloss values for used real-life and benchmark data sets.

Depression is a severe psychological disorder that is experienced by a significant number of individuals across the globe. It greatly changes the way one thinks, triggering a constant decline in mood. Studies have shown that gender can act as a good indicator of depression. In this paper, we analyse the effects of gender information in the estimation of depression. We have carried out different experiments on the benchmark data set named Distress Analysis Interview Corpus - a Wizard of Oz (DAIC-WOZ). Concretely, we discovered that a) gender information substantially improves the performance of depression severity estimation, and b) adversarially learning to predict the depression score distributed by gender improves the performance of depression severity estimation.

In many real-world machine learning problems, the features are changing along the time, with some old features vanishing and some other new features augmented, while the remaining features survived. In this paper, we propose the cross-feature attention network to handle the incremental and decremental features. This network is composed of multiple cross-feature attention encoding-decoding layers. In each layer, the data samples are firstly encoded by the combination of other samples with vanished/augmented features and weighted by the attention weights calculated by the survived features. Then, the samples are encoded by the combination of samples with the survived features weighted by the attention weights calculated from the encoded vanished/augmented feature data. The encoded vanished/augmented/survived features are then decoded and fed to the next cross-feature attention layer. In this way, the incremental and decremental features are bridged by paying attention to each other, and the gap between data samples with a different set of features are filled by the attention mechanism. The outputs of the cross-feature attention network are further concatenated and fed to the class-specific attention and global attention network for the purpose of classification. We evaluate the proposed network with benchmark data sets of computer vision, IoT, and bio-informatics, with incremental and decremental features. Encouraging experimental results show the effectiveness of our algorithm.

The results from resistance measurements in calm water and load-varying self-propulsion tests in calm water and regular head and following waves are presented. The experimental campaign is conducted in the large towing tank at SINTEF Ocean (formerly MARINTEK). The openly accessible hull of the single screw Duisburg Test Case is selected as the test case. The wave added resistance, ship motions RAOs, axial wake fraction, thrust deduction fraction, relative rotative efficiency, hull efficiency, propeller open-water efficiency, propeller efficiency behind ship, and propulsive efficiency are determined. Regarding the calculation of the thrust deduction fraction, the results of the experiments show the effect of utilizing the bare hull resistance instead of the linearly extrapolated ship resistance at zero propeller thrust. If the former is applied, the thrust deduction fraction will be dependent on the load of the propeller. If the latter is utilized, the thrust deduction fraction will be independent of the propeller loading. As expected, the wave added resistance is lower in following waves than in head waves. The heave and pitch motions are larger in head waves, whereas the surge motion is higher in following waves. The effective wake fraction is affected by both the propeller loading and the ship motions. In the case of the former, the higher the propeller loading, the lower the effective wake fraction. For the latter, a general decrease in effective wake fraction is noticed in head waves compared to calm water. On the contrary, the effective wake fraction is higher in following waves in comparison to calm water. The thrust deduction fraction computed with the extrapolated ship resistance appears to be slightly affected by the ship motions. However, a final conclusion was not drawn because of the large uncertainties in the measurements of this propulsive coefficient. The variation in propeller open-water efficiency is mainly related to the change in propeller loading. The relative rotative efficiency is barely affected by both the propeller loading and the motions of the ship. Except in the case of very large wave amplitudes, the hull efficiency is hardly influenced by the ship motions. The propulsive efficiency is primarily affected by the change in propeller open-water efficiency. Based on the results of the experimental campaign, overload tests in calm water provide a good estimation of the propulsion efficiency in waves for the selected case vessel.

Skin covers the entire body and is the largest organ. Skin cancer is one of the most dreadful cancers that is primarily triggered by sensitivity to ultraviolet rays from the sun. However, the riskiest is melanoma, although it starts in a few different ways. The patient is extremely unaware of recognizing skin malignant growth at the initial stage. Literature is evident that various handcrafted and automatic deep learning features are employed to diagnose skin cancer using the traditional machine and deep learning techniques. The current research presents a comparison of skin cancer diagnosis techniques using handcrafted and non-handcrafted features. Additionally, clinical features such as Menzies method, seven-point detection, asymmetry, border color and diameter, visual textures (GRC), local binary patterns, Gabor filters, random fields of Markov, fractal dimension, and an oriental histography are also explored in the process of skin cancer detection. Several parameters, such as jacquard index, accuracy, dice efficiency, preciseness, sensitivity, and specificity, are compared on benchmark data sets to assess reported techniques. Finally, publicly available skin cancer data sets are described and the remaining issues are highlighted.

Timely and accurate Land Cover (LC) information is required for various applications, such as climate change analysis and sustainable development. Although machine learning algorithms are most likely successful in LC mapping tasks, the class imbalance problem is known as a common challenge in this regard. This problem occurs during the training phase and reduces classification accuracy for infrequent and rare LC classes. To address this issue, this study proposes a new method by integrating random under-sampling of majority classes and an ensemble of Support Vector Machines, namely Random Under-sampling Ensemble of Support Vector Machines (RUESVMs). The performance of RUESVMs for LC classification was evaluated in Google Earth Engine (GEE) over two different case studies using Sentinel-2 time-series data and five well-known spectral indices, including the Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), Soil-Adjusted Vegetation Index (SAVI), Normalized Difference Built-up Index (NDBI), and Normalized Difference Water Index (NDWI). The performance of RUESVMs was also compared with the traditional SVM and combination of SVM with three benchmark data balancing techniques namely the Random Over-Sampling (ROS), Random Under-Sampling (RUS), and Synthetic Minority Over-sampling Technique (SMOTE). It was observed that the proposed method considerably improved the accuracy of LC classification, especially for the minority classes. After adopting RUESVMs, the overall accuracy of the generated LC map increased by approximately 4.95 percentage points, and this amount for the geometric mean of producer’s accuracies was almost 3.75 percentage points, in comparison to the most accurate data balancing method (i.e., SVM-SMOTE). Regarding the geometric mean of users’ accuracies, RUESVMs also outperformed the SVM-SMOTE method with an average increase of 6.45 percentage points.

With the rapid development of machine learning (ML) and deep learning (DL) in computer vision, adopting these learning tools in vision-based structural health monitoring (SHM) and rapid dam...

Background Federated learning (FL) is a newly proposed machine-learning method that uses a decentralized dataset. Since data transfer is not necessary for the learning process in FL, there is a significant advantage in protecting personal privacy. Therefore, many studies are being actively conducted in the applications of FL for diverse areas. Objective The aim of this study was to evaluate the reliability and performance of FL using three benchmark datasets, including a clinical benchmark dataset. Methods To evaluate FL in a realistic setting, we implemented FL using a client-server architecture with Python. The implemented client-server version of the FL software was deployed to Amazon Web Services. Modified National Institute of Standards and Technology (MNIST), Medical Information Mart for Intensive Care-III (MIMIC-III), and electrocardiogram (ECG) datasets were used to evaluate the performance of FL. To test FL in a realistic setting, the MNIST dataset was split into 10 different clients, with one digit for each client. In addition, we conducted four different experiments according to basic, imbalanced, skewed, and a combination of imbalanced and skewed data distributions. We also compared the performance of FL to that of the state-of-the-art method with respect to in-hospital mortality using the MIMIC-III dataset. Likewise, we conducted experiments comparing basic and imbalanced data distributions using MIMIC-III and ECG data. Results FL on the basic MNIST dataset with 10 clients achieved an area under the receiver operating characteristic curve (AUROC) of 0.997 and an F1-score of 0.946. The experiment with the imbalanced MNIST dataset achieved an AUROC of 0.995 and an F1-score of 0.921. The experiment with the skewed MNIST dataset achieved an AUROC of 0.992 and an F1-score of 0.905. Finally, the combined imbalanced and skewed experiment achieved an AUROC of 0.990 and an F1-score of 0.891. The basic FL on in-hospital mortality using MIMIC-III data achieved an AUROC of 0.850 and an F1-score of 0.944, while the experiment with the imbalanced MIMIC-III dataset achieved an AUROC of 0.850 and an F1-score of 0.943. For ECG classification, the basic FL achieved an AUROC of 0.938 and an F1-score of 0.807, and the imbalanced ECG dataset achieved an AUROC of 0.943 and an F1-score of 0.807. Conclusions FL demonstrated comparative performance on different benchmark datasets. In addition, FL demonstrated reliable performance in cases where the distribution was imbalanced, skewed, and extreme, reflecting the real-life scenario in which data distributions from various hospitals are different. FL can achieve high performance while maintaining privacy protection because there is no requirement to centralize the data.

This work employed recent model outputs from coupled model intercomparison project phase six to simulate surface mean temperature during the June&ndash, July&ndash, August (JJA) and December&ndash, January&ndash, February (DJF) seasons for 1970&ndash, 2014 over Pakistan. The climatic research unit (CRU TS4.03) dataset was utilized as benchmark data to analyze models&rsquo, performance. The JJA season exhibited the highest mean temperature, whilst DJF displayed the lowest mean temperature in the whole study period. The JJA monthly empirical cumulative distribution frequency (ECDF) range (26 to 28 &deg, C) was less than that of DJF (7 to 10 &deg, C) since JJA matched closely to CRU. The JJA and DJF seasons are warming, with higher warming trends in winters than in summers. On temporal scale, models performed better in JJA with overall low bias, low RMSE (root mean square error), and higher positive CC (correlation coefficient) values. DJF performance was undermined with higher bias and RMSE with weak positive correlation estimates. Overall, CanESM5, CESM2, CESM2-WACCM, GFDL-CM4, HadGEM-GC31-LL, MPI-ESM1-2-LR, MPI-ESM1-2-HR, and MRI-ESM-0 performed better for JJA and DJF.

This paper explores Data Mining (DM) techniques such as Decision Tree, Random Forest Decision Trees, and Artificial Neural Networks, employing Principal Component and Cluster Analysis for solving the problems of direct marketing. In our experiments the prediction accuracy 91% has been achieved with the DM models built for solving the problem. The experiments were run in RStudio which is a Popular DM platform, allowing analysts to efficiently carry out experiments and obtain the results. In our experiments we analysed patterns existing in the marketing benchmark data, which can affect the prediction accuracy. Boosting explored in our experiments has demonstrated an efficient reduction of the prediction error. Based on our results we conclude that the DT and ANN models provide analysts with understandable decision making concepts which can meet the realistic expectations of prediction accuracy.

The container relocation problem is one of important issues in seaport terminals which could bring a significant saving on the operating cost even with a slight improvement due to the huge number of containers processed across the world each year. Given a specific layout and container retrieval priorities, the container relocation problem aims to find the optimal movement sequence to minimize the total number of container relocation operations. In this paper, we propose a new upper bound method called MLUB that incorporates branch pruners. These pruners are derived from some machine learning techniques through using the optimal solution values of many small-scale instances. The tightened upper bounds generated by MLUB are used subsequently in the exact branchand-bound algorithm called IB&B. Moreover, we also provide a tighter lower bound for the problem by additionally considering the interaction between consecutive target containers. Based on the benchmark data published recently in the literature, extensive experiments are conducted to test the performance of the proposed algorithms.

Facing with the absence of supervised information to guide the search of relevant features and the grid-search of model/hyper-parameters, it is more preferred to develop parameter-free methods and avoid additional hyper-parameters tuning. In this paper, we propose a new simple and effective parameter-free unsupervised feature selection algorithm by minimizing the linear reconstruction weight between the nearest neighbor graphs constructed from all candidate features and each single feature. The obtained global optimal reconstruction weights actually select those features with highest relevance and lowest redundancy simultaneously. The experimental results on many benchmark data sets demonstrate that the proposed method outperforms many of the state-of-the-art unsupervised feature selection methods.

Spatio-temporal fusion (STF) aims at fusing (temporally dense) coarse resolution images and (temporally sparse) fine resolution images to generate image series with adequate temporal and spatial resolution. In the last decade, STF has drawn a lot of attention and many STF methods have been developed. However, to datethe STF domain still lacks benchmark datasets, which is a pressing issue that needs to be addressed in order to foster the development of this field. In this review, we provide (for the first time in the literature) a robust benchmark STF dataset that includes three important characteristics: (1) diversity of regions, (2) long timespan, and (3) challenging scenarios. We also provide a survey of highly representative STF techniques, along with a detailed quantitative and qualitative comparison of their performance with our newly presented benchmark dataset. The proposed dataset is public and available online.

In this study, an advanced radioactive particle tracking (RPT) technique was used to investigate for the first time the details of the cells’ movements (trajectory) and multiphase flow hydrodynamics during microalgae culturing in a cylindrical split airlift photobioreactor. The cells’ trajectory, liquid velocity field, distributions of shear stresses, and the turbulent kinetic energy field were studied under superficial gas velocity of 1 and 3 cm/s. The structures of the flow in the whole reactor, the riser, the downcomer, as well as the structure above and below the split plate were characterized. The effects of the cells’ concentration and different aeration rate at different axial levels on the studied parameters were discussed. It has been found that the cells’ fluctuations reduced and its movement frequency between the light (wall) and dark zone decreased during the culturing particularly when the cells concentrations becomes large after 30 days of culturing. Distinguishing behaviors were observed for all the parameters, with a higher magnitude at the superficial gas velocity 3 cm/sec than at 1 cm/sec. This effect positively enhanced the liquid circulation and the movement between the reactor sides, the riser, and the downcomer. This circulation and good mixing phenomena had a large positive impact on the culture’s continuity. The obtained results are reliable as benchmark data to validate computational fluid dynamics (CFD) simulation and other models that can be later used to be integrated with dynamic growth and light intensity models for optimized.

This paper primarily evaluates the efficacy of shallow and deep networks to statically detect malicious windows portable executable (PE) files. This uses recently released, labeled and benchmark data set, EMBER malware benchmark data set. As deep networks are parameterized, the parameters are chosen based on comparing the performance of various network parameters and network topologies over various trials of experiments. The experiments of such chosen efficient configurations of deep models are run up to 1000 epochs with varying learning rates between 0.01 and 0.5. The observed results of deep networks are high compared to the shallow networks. Keywords: Static analysis, Malicious and benign binaries and deep networks

Scene classification is usually based on supervised learning, but collecting instances is expensive and time-consuming. TrAdaBoost has achieved great success in transferring source instances to target images, but it has problems, such as the excessive focus on instances harder to classify, the rapid convergence speed of the source instances, and the weight mismatch caused by the big gap between the number of source and target instances, leading to decreased classification accuracy. In this paper, in order to address these problems, classical particle swarm optimization (PSO) is modified to select the optimal feature subspace for classifying the “harder” and “easier” instances by reducing unimportant dimensions. A modified correction factor is proposed by considering the classification accuracy of the instances from both domains, to decrease the convergence speed. Iterative selective TrAdaBoost is also proposed to reduce the weight mismatch by removing the indiscriminate source instances. The experimental results obtained with three benchmark data sets confirm that the proposed method outperforms most of the previous methods of scene classification with limited target samples. It is also proved that modified PSO for optimal feature subspace selection, the modified correction factor, and iterative selective TrAdaBoost are all effective in improving the classification accuracy, giving improvements of 3.6%, 4.3%, and 2.7%, and these three contributions together increase the classification accuracy by about 8% in total.

Gap-filling streamflow data is a critical step for most hydrological studies, such as streamflow trend, flood, and drought analysis and hydrological response variable estimates and predictions. However, there is a lack of quantitative evaluation of the gap-filled data accuracy in most hydrological studies. Here we show that when the missing data rate is less than 10 %, the gap-filled streamflow data obtained using calibrated hydrological models perform almost the same as the benchmark data (less than 1 % missing) when estimating annual trends for 217 unregulated catchments widely spread across Australia. Furthermore, the relative streamflow trend bias caused by the gap filling is not very large in very dry catchments where the hydrological model calibration is normally poor. Our results clearly demonstrate that the gap filling using hydrological modelling has little impact on the estimation of annual streamflow and its trends.

Employing the merged quasi-geoid, we analyses the causes of systematic errors in modelling of quasi-geoid of neighbouring areas in the paper, and the efficient method is introduced to improve the accuracy of quasi-geoid. First, the systematic error is weakened with the moving window method applied to established quasi-geoids in two adjacent regions, and the accuracy of the merged quasi-geoid in the stitching region is checked using the measured GPS benchmark data; Second, the whole quasi-geoid is recomputed with data obtained from two adjacent regions if the accuracy of the quasi-geoid obtained from the first step in the stitching region is low; Finally, the quasi-geoids in two adjacent regions are recomputed respectively using GPS benchmark data of own region and adjacent region with the same solution if the accuracy of whole quasi-geoid obtained from the second step also is low. Actual data sets from Sichuan Province and Chongqing City are employed to test the moving window method. It is shown that the quasi-geoid models with high resolution and accuracy were obtained. Keywords: Merged quasi-geoid, Moving window, Systematic error

In this paper, a robust extreme learning machine is proposed. In comparison with the original extreme learning machine and the regularized extreme learning machine, this robust algorithm minimizes both the mean and variance of modeling errors in the objective function to overcome the bias-variance dilemma. As a result, its generalization performance and robustness are enhanced, and these merits are further proved theoretically. In addition, this proposed algorithm can keep the same computational efficiency as the original extreme learning machine and the regularized extreme learning machine. Then, several benchmark data sets are used to test the effectiveness and soundness of the proposed algorithm. Finally, it is employed to model a real small-scale turbojet engine. This engine is fit well. Especially, on the idle phase, where the signal-to-noise ratio is low and it is very hard to model, the proposed algorithm performs well and its robustness is sufficiently showcased. All in all, the proposed algorithm provides a candidate technique for modeling real systems.

The concept of `diversity' has been one of the main open issues in the field of multiple classifier systems. In this paper we address a facet of diversity related to its effectiveness for ensemble construction, namely, explicitly using diversity measures for ensemble construction techniques based on the kind of overproduce and choose strategy known as ensemble pruning. Such a strategy consists of selecting the (hopefully) more accurate subset of classifiers out of an original, larger ensemble. Whereas several existing pruning methods use some combination of individual classifiers' accuracy and diversity, it is still unclear whether such an evaluation function is better than the bare estimate of ensemble accuracy. We empirically investigate this issue by comparing two evaluation functions in the context of ensemble pruning: the estimate of ensemble accuracy, and its linear combination with several well-known diversity measures. This can also be viewed as using diversity as a regularizer, as suggested by some authors. To this aim we use a pruning method based on forward selection, since it allows a direct comparison between different evaluation functions. Experiments on thirty-seven benchmark data sets, four diversity measures and three base classifiers provide evidence that using diversity measures for ensemble pruning can be advantageous over using only ensemble accuracy, and that diversity measures can act as regularizers in this context.

This paper concerns one of intelligent computational techniques which is case based reasoning (CBR) more particularly the case base maintenance (CBM). It aims to ensure the CBR systems quality. Throughout this paper, we were faced to a problematic question: how to shrink the size of the case base while preserving as much as possible the performance and the competence of the CBR system in soft context. To answer this question, we have first analyzed and revised the theoretical foundations of the existing CBM methods. Then, we have proposed a novel soft case base maintenance (SCBM) method based on a soft competence model (SCM) and a fuzzy clustering technique. Our method has the objective to guarantee the CBR systems efficiency in terms of improving the competence, and reducing both the storage requirements and search time. We support our approach with empirical evaluation using different benchmark data sets to show the effectiveness of our method in terms of improving the competence and the performance of the system.

In this paper, we propose a method for solving the twin bounded support vector machine (TBSVM) for the binary classification. To do so, we use the augmented Lagrangian (AL) optimization method and smoothing technique, to obtain new unconstrained smooth minimization problems for TBSVM classifiers. At first, the augmented Lagrangian method is recruited to convert TBSVM into unconstrained minimization programming problems called as AL-TBSVM. We attempt to solve the primal programming problems of AL-TBSVM by converting them into smooth unconstrained minimization problems. Then, the smooth reformulations of AL-TBSVM, which we called AL-STBSVM, are solved by the well-known Newton's algorithm. Finally, experimental results on artificial and several University of California Irvine (UCI) benchmark data sets are provided along with the statistical analysis to show the superior performance of our method in terms of classification accuracy and learning speed.

Natural image quality assessment (NIQA) wins increasing attention, while NIQA models are rarely used in the medical community. A couple of studies employ the NIQA methodologies for medical image quality assessment (MIQA), but building the benchmark data sets necessitates considerable time and professional skills. In particular, the characteristics of synthesized distortions are different from those of clinical distortions, which make the results not so convincing. In clinic, signal-to-noise ratio (SNR) is widely used, which is defined as the quotient of the mean signal intensity measured in a tissue region of interest (ROI) and the standard deviation of the signal intensity in an air region outside the imaged object, and both regions are outlined by specialists. We take advantage of the knowledge that SNR is routinely used and concern whether SNR measure can perform as a baseline metric for the development of MIQA algorithms. To address the issue, the inter-observer reliability of SNR measure is investigated regarding to different tissue ROIs [white matter (WM); cerebral spinal fluid (CSF)] in magnetic resonance (MR) images. A total of 192 ${\text{T}}_{2}^{*}$ , 88 ${\text{T}}_{1}$ , 76 ${\text{T}}_{2}$ and 55 contrast-enhanced ${\text{T}}_{1}$ ( ${\text{T}}_{1}\text{C}$ ) weighted images are analyzed. Statistical analysis indicates that SNR values show consistency between different observers to the same ROI in each modality (Wilcoxon rank sum test, $p_{w}\ge 0.11$ ; and paired sample $t$ -test, $p_{p}\ge 0.28$ ). Moreover, whether off-the-shelf NIQA models can predict MR image quality is considered by using SNR values as the reference scores. Four NIQA models (BIQI, BLIINDS-II, BRISQUE, and NIQE) are evaluated, and the correlation between SNR values and NIQA results is evaluated. Pearson correlation coefficient ( $r_{p}$ ) shows that WM-based SNR values correlates well with BIQI, BLIINDS-II and BRISQUE in ${\text{T}}_{2}^{*}$ images ( $r_{p}\geq 0.77$ ), BRISQUE and NIQE in ${\text{T}}_{1}$ images ( $r_{p}\geq 0.75$ ), BLIINDS-II in ${\text{T}}_{2}$ images ( $r_{p}\geq 0.67$ ), and BRISQUE and NIQE in ${\text{T}}_{1}\text{C}$ images ( $r_{p}\geq 0.58$ ), while CSF-based SNR values correlates well with BLIINDS-II in ${\text{T}}_{2}^{*}$ images ( $r_{p}\geq 0.64$ ) and ${\text{T}}_{2}$ images ( $r_{p}\geq 0.60$ ), and all $p_{p} . The prediction performance analysis further proves the result from the correlation analysis. Conclusively, SNR measure is reliable to different observations and can perform as a baseline indicator for the development of MIQA algorithms. In general, BRISQUE and BLIINDS-II are full of potential to be conditionally used as objective MIQA models toward human brain MR images. This paper presents the first attempt of using SNR measure to bridge the gap between NIQA and MIQA, and large-scale experiments should be further conducted to confirm the conclusion in this paper.