Search

In this paper, we propose a novel paper currency defect detection algorithm using quaternion uniform strength. We first build paper currency image preprocessing integration framework which includes intensity balancing, paper currency location, and geometric correction. We then propose a global–local paper currency image registration algorithm by moving key areas within certain range which can eliminate the false difference effectively. Finally, the quaternion uniform strength is calculated by using quaternion convolution edge detector. The defect degree of paper currency is determined by using the quaternion uniform color difference. The proposed algorithm is tested using different datasets from five countries: CNY, USD, EUR, VND, and RUB. The experimental results demonstrate that the proposed algorithm yields better results than the existing state-of-the-art paper currency defect detection techniques. The demo of the proposed paper currency defect detection algorithm will be publicly available. [ABSTRACT FROM AUTHOR]

Manual term extraction is similar to literal meaning: A translator browses text, classifies words, and prepares for translation. Terminology, as a centralized carrier of expertise, creation, popularization, and disappearance, dynamically reflects the development and evolution of an industry. The automatic extraction of terminology is a key technology for creating a professional terminology database, and it is also a key topic in the field of natural language processing. The purpose of this paper is to study how to analyse a term extraction algorithm based on machine learning and a comprehensive feature strategy. Focusing on the problems of poor generality and single statistical features of current term extraction algorithms, this paper proposes an improved domain ontology term extraction algorithm based on a comprehensive feature strategy. Moreover, automatic term extraction experiments based on a word-based maximum entropy model and a conditional random field model based on machine learning are conducted in this paper. Its word-based conditional random field model outperforms the maximum entropy model. The experimental results show that the algorithm based on the comprehensive feature strategy improves the accuracy by 8.6% compared with the TF-IDF algorithm and the C-value term extraction algorithm. This algorithm can be used to effectively extract the terms in a text and has good generality. [ABSTRACT FROM AUTHOR]

One of the most popular techniques for changing the purpose of an image or resizing a digital image with content awareness is the seam-carving method. The performance of image resizing algorithms based on seam machining shows that these algorithms are highly dependent on the extraction of importance map techniques and the detection of salient objects. So far, various algorithms have been proposed to extract the importance map. In this paper, a new method based on Rényi entropy is proposed to extract the importance map. Also, a deep learning network has been used to detect salient objects. The simulator results showed that combining Rényi's importance map with a deep network of salient object detection performed better than classical seam-carving and other extended seam-carving algorithms based on deep learning. [ABSTRACT FROM AUTHOR]

This letter identifies a fundamental mistake in the published paper titled "A projective general linear group-based algorithm for the construction of substitution box for block ciphers." The projective general linear group-based algorithm is designed to provide high nonlinearity, low differential uniformity, and good resistance against linear and differential attacks. The projective general linear group-based algorithm uses the linear fractional transformation defined over finite field GF 2 8 . The authors of the paper claimed that the irreducible polynomial of degree 8 is used for the construction of a finite field GF 2 8 . We have observed that the projective general linear group-based algorithm uses a reducible polynomial of degree 8. Thus, the inverse of each element does not exist in the finite field GF 2 8 . We have also shown that the values of the S-Box given in the paper are also incorrect. [ABSTRACT FROM AUTHOR]

Semi-supervised clustering has gained significant attention from researchers due to its advantages over unsupervised clustering. However, existing studies have predominantly focused on discrete data. This paper pioneers the application of semi-supervised clustering to probability density functions. The proposed algorithm encompasses detailed implementation steps, a convergence proof, and the ability to address computational challenges. The algorithm has been effectively implemented on image data, resulting in the transformation of each image into a probability density function that is representative. In comparison to existing unsupervised algorithms, the efficacy of the proposed algorithm in partitioning and reducing computational costs is demonstrated through numerical examples and applications. [ABSTRACT FROM AUTHOR]

In the field of machine learning, clustering has become an increasingly popular research topic due to its critical importance. Many clustering algorithms have been proposed utilizing a variety of approaches. This study focuses on clustering of high-dimensional data using the maximum margin clustering approach. In this paper, two methods are introduced: The first method employs the classical maximum margin clustering approach, which separates data into two clusters with the greatest margin between them. The second method takes cluster compactness into account and searches for two parallel hyperplanes that best fit to the cluster samples while also being as far apart from each other as possible. Additionally, robust variants of these clustering methods are introduced to handle outliers and noise within the data samples. The stochastic gradient algorithm is used to solve the resulting optimization problems, enabling all proposed clustering methods to scale well with large-scale data. Experimental results demonstrate that the proposed methods are more effective than existing maximum margin clustering methods, particularly in high-dimensional clustering problems, highlighting the efficacy of the proposed methods. [ABSTRACT FROM AUTHOR]

The chimp optimization algorithm (ChOA) is a nature-inspired algorithm that imitates chimpanzees' individual intelligence and hunting behaviors. In this algorithm, the hunting process consists of four steps: driving, blocking, chasing, and attacking. Because of the novelty of ChOA, the steps of the hunting process have been modeled in the simplest possible way, leading to slow and premature convergence similar to other iterative algorithms. This paper proposes six spiral functions and introduces two novel hybrid spiral functions (SEB-ChOA) to rectify the abovementioned deficiencies. The SEB-ChOAs' performance is evaluated on 23 standard benchmarks, 20 benchmarks of IEEE CEC-2005, 10 cases of IEEE CEC06-2019 test-suite, and 12 constrained real-world engineering problems of IEEE CEC-2020. The SEB-ChOAs are compared with three groups of optimization algorithms, including Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) as the most well-known optimization algorithms, Slime Mould Algorithm (SMA), Marine Predators Algorithm (MPA), Ant Lion Optimization (ALO), Henry Gas Solubility Optimization (HGSO), as almost novel optimization algorithms, and jDE100 and DISHchain1e+12, as winners of IEEE CEC06-2019 competition, and also EBOwithCMAR and CIPDE as superior secondary optimization algorithms. The SEB-ChOAs reached the first rank among almost all benchmarks and demonstrated very competitive results compared to jDE100 and DISHchain1e+12 as the best-performing optimizers. Statistical evidence shows that the SEB-ChOA outperforms the PSO, GA, SMA, MPA, ALO, and HGSO optimizers while producing results comparable to those of the jDE100 and DISHchain1e+12 algorithms. [ABSTRACT FROM AUTHOR]

Arithmetic Optimization Algorithm (AOA) is a heuristic method developed in recent years. The original version was developed for continuous optimization problems. Its success in binary optimization problems has not yet been sufficiently tested. In this paper, the binary form of AOA (BinAOA) has been proposed. In addition, the candidate solution production scene of BinAOA is developed with the xor logic gate and the BinAOAX method was proposed. Both methods have been tested for success on well-known uncapacitated facility location problems (UFLPs) in the literature. The UFL problem is a binary optimization problem whose optimum results are known. In this study, the success of BinAOA and BinAOAX on UFLP was demonstrated for the first time. The results of BinAOA and BinAOAX methods were compared and discussed according to best, worst, mean, standard deviation, and gap values. The results of BinAOA and BinAOAX on UFLP are compared with binary heuristic methods used in the literature (TSA, JayaX, ISS, BinSSA, etc.). As a second application, the performances of BinAOA and BinAOAX algorithms are also tested on classical benchmark functions. The binary forms of AOA, AOAX, Jaya, Tree Seed Algorithm (TSA), and Gray Wolf Optimization (GWO) algorithms were compared in different candidate generation scenarios. The results showed that the binary form of AOA is successful and can be preferred as an alternative binary heuristic method. [ABSTRACT FROM AUTHOR]

Online streaming feature selection (OSFS) methods are used to dynamically update the feature space as well as remove irrelevant and redundant features from the data. Since most Big Data in real-world applications are generated in the form of data streams, effective methods should be developed in this area. Further, methods with low computational complexity are required to make online decisions. In this paper, the OSFS process is modeled as a multi-objective optimization problem. To the best of our knowledge, this is the first time that the concept of Pareto dominance has been applied to find the optimal subset of features in OSFS. When a new feature arrives, it is evaluated in the multi-objective space. The non-dominated features are the optimal subset for each timestamp. We proposed an efficient and effective method which enhances the classification accuracy in OSFS by minimizing the number of features within a short time. In addition, the proposed method is insensitive to the feature streams. Experiments are conducted using two classifiers and seven OSFS methods, including OSFSMI, K-OFSD, OFS-A3M, OFS-Density, Alpha-Investing, SAOLA, and OFSS-FI. [ABSTRACT FROM AUTHOR]

In practical applications, the optimal solutions of multi-objective optimization are not unique. Some problems exist different Pareto Sets (PSs) in the decision space mapped to the same Pareto Front (PF) in the objective space, which are called multimodal multi-objective problems (MMOPs). To tackle this issue, this paper proposes a multimodal multi-objective optimization based on a local optimal neighborhood crowding distance differential evolution algorithm. First, an adaptive partitioning strategy in the initialization phase is proposed by using the characteristics of the heuristic stochastic search. That ensures the local optimal solution is quickly found among multiple PSs. Second, opposition-based learning is combined with differential mutation to generate vectors, which accelerate the convergence of the population to the optimal solution. Finally, a method for neighborhood crowding distances on different Pareto ranks is designed. The distance is computed by a weighted sum of Euclidean distances for the nearest neighbors. While reducing computational complexity, this strategy reflects realistic crowding degree. With these methods, balances the diversity performance of the decision and the objective space, while improving the search capability. Multiple PSs reveal the problem's potential characteristics and meet the needs of the decision-maker. The practical significance is verified by the application of actual distance minimization problem. According to experimental results, the proposed method can achieve a high level of comprehensive performance. [ABSTRACT FROM AUTHOR]

Twin support vector regression (TSVR) has been widely applied in regression problems. TSVR seeks a pair of ε -insensitive proximal planes by solving two support vector machine type problems. TSVR assumes that the matrices appearing in the dual formulation are positive definite. However, in real-world scenarios, such an assumption may not be fulfilled and, hence, leads to suboptimal performance. ε -Twin support vector regression (ε -TSVR) improved the TSVR by introducing the regularisation term to avoid the singularity issues. Most of the twin support vector machine models involve the computation of matrix inverses. Also, TSVR implements the empirical risk minimization principle. In this paper, we propose an improved twin support vector regression (ITSVR) for brain age estimation by introducing different Lagrangian functions for the primal problems of the TSVR. The proposed ITSVR implements the structural risk minimization principle and avoids the computation of the matrix inverses. To solve the optimization problem more efficiently, we used successive overrelaxation (SOR) technique. We evaluated the proposed ITSVR on cognitively healthy subjects, mild cognitive impairment subjects and Alzheimer's disease subjects. The experimental results demonstrate that the proposed ITSVR has superior performance compared to the baseline models for brain age estimation. [ABSTRACT FROM AUTHOR]

After 70 years of intricate development, machine learning, represented by deep learning, is based on the multilevel structure of the human brain and the layer-by-layer analysis and processing mechanism of neuron connection and interaction information. The powerful parallel information processing ability of self-adaptation and self-learning has allowed for breakthroughs in many fields, among which the most representative is image recognition. Therefore, this paper proposed optimizing the RBF algorithm with machine learning (ML) to improve the recognition rate of spectrum sensing. The results showed that the average detection success rates of the RBF algorithm were 93.62%, 95.07%, 96.91%, 98.78% and 99.37% when the SNRs were − 8 dB, − 4 dB, 0 dB, 4 dB and 8 dB, respectively, and the other conditions were kept the same. The average detection success rates of the SVM/RBF algorithm were 97.65%, 99.63%, 99.76%, 99.91% and 99.88%, respectively. The average detection success rate of the SVM/RBF algorithm was significantly higher than that of the RBF algorithm. This indicates that analyzing the RBF neural network algorithm through ML can improve the success rate of spectrum sensing, which highlights a new direction for the application of ML and neural networks. [ABSTRACT FROM AUTHOR]

Cluster analysis is one of the most fundamental methods in data mining, and it has been widely used in economics, social sciences and computer science. However, with the rapid development of Internet technology, the volume of data required for various web applications has grown rapidly, making the traditional clustering analysis methods face technical challenges. How to obtain useful information in a large amount of data quickly and efficiently is an urgent problem in many industrial fields. With the continuous development of cloud computing technology, large amounts of data can be performed quickly and efficiently. Hadoop is an open source distributed cloud computing platform with HDFS (Digital File System) and MapReduce as its core. HDFS provides massive data storage, while MapReduce uses the MapReduce programming model to achieve parallel processing. Compared with the traditional parallel programming model, it contains basic functions such as data partitioning, task scheduling, and parallel processing, making it possible for users to develop distributed applications on their own without understanding the basics of distributed basics, thus facilitating the design of parallel programs. K-means algorithm is a typical clustering analysis method, which is widely used in industry, but the number of iterations will increase significantly due to the growth of data volume, thus reducing the efficiency of computation. In order to better apply to the cluster analysis of large-scale data, this paper firstly implements a parallelization algorithm based on MapReduce on Hadoop platform using the basic idea of MapReduce and improves the K-means algorithm for the problems of blindness and easy to fall into local optimum when selecting randomly in clusters. [ABSTRACT FROM AUTHOR]

The transition towards full network virtualization will see services for smart ecosystems including smart metering, healthcare and transportation among others, being deployed as Service Function Chains (SFCs) comprised of an ordered set of virtual network functions. However, since such services are usually deployed in remote cloud networks, the SFCs may transcend multiple domains belonging to different Infrastructure Providers (InPs), possibly with differing policies regarding billing and Quality-of-service (QoS) guarantees. Therefore, efficiently allocating the exhaustible network resources to the different SFCs while meeting the stringent requirements of the services such as delay and QoS among others, remains a complex challenge, especially under limited information disclosure by the InPs. In this work, we formulate the SFC deployment problem across multiple domains focusing on delay constraints, and propose a framework for SFC orchestration which adheres to the privacy requirements of the InPs. Then, we propose a reinforcement learning (RL)-based algorithm for partitioning the SFC request across the different InPs while considering service reliability across the participating InPs. Such RL-based algorithms have the intelligence to infer undisclosed InP information from historical data obtained from past experiences. Simulation results, considering both online and offline scenarios, reveal that the proposed algorithm results in up to 10% improvement in terms of acceptance ratio and provisioning cost compared to the benchmark algorithms, with up to more than 90% saving in execution time for large networks. In addition, the paper proposes an enhancement to a state-of-the-art algorithm which results in up to 5% improvement in terms of provisioning cost. [ABSTRACT FROM AUTHOR]

In this paper, the impact of rated power and the total capacity of all photovoltaic units on the energy loss reduction of radial distribution networks is investigated. An IEEE 69-node system is supposed to be a radial distribution network in the North and South of Vietnam to take the real mean solar radiations of 1 year from the global solar map for simulation. Five study cases are implemented corresponding to the difference of rated power of five installed photovoltaic units, including optimal size and four fixed sizes of 250, 500, 750 and 1000 kW. The proposed modified slime mould algorithm (MSMA) and three other algorithms including the original Coot optimization algorithm (COA), original Transient search optimization algorithm (TSOA) and original slime mould algorithm (SMA) are used to find optimal solutions of the five study cases. As a result, MSMA is the best method for reaching smaller 1-year energy loss and finding a higher number of more effective solutions than others. In addition, the study also suggests the placement of photovoltaic units should use optimal rated power for each unit and optimal total capacity for all units rather than using the same rated power for each unit and a predetermined total capacity for all units. [ABSTRACT FROM AUTHOR]

Nonnegative matrix factorization (NMF) is a crucial method for image clustering. However, NMF may obtain low accurate clustering results because the factorization results contain no data structure information. In this paper, we propose an algorithm of nonnegative matrix factorization under structure constraints (SNMF). The factorization results of SNMF could maintain data global and local structure information simultaneously. In SNMF, the global structure information is captured by the cosine measure under the ℓ 2 norm constraints. Meanwhile, ℓ 2 norm constraints are utilized to get more discriminant data representations. A graph regularization term is employed to maintain the local structure. Effective updating rules are given in this paper. Moreover, the effects of different normalizations on similarities are investigated through experiments. On real datasets, the numerical results confirm the effectiveness of the SNMF. [ABSTRACT FROM AUTHOR]

Differential evolutionary (DE) algorithm is one of the most frequently used evolutionary computation method for the solution of non-differentiable, complex and discontinuous real value numerical problems. The analytical structure of the mutation and crossover operators used by DE and the initial values of the parameters of the relevant operators affect the problem-solving ability of DE. Unfortunately, there is no analytical method for selecting and initializing the best artificial genetic operators that DE can use to solve a problem. Therefore, there is a need to develop new evolutionary search methods that are parameter-free and insensitive to the artificial genetic operators they use. In this paper, the Bernstein–Levy differential evolution (BDE) algorithm, which has a unique elitist-mutation operator and a Bernstein polynomials-based stochastic parameter-free crossover operator, is introduced. The numerical problem-solving success of BDE is statistically evaluated by using 30 benchmark problems of CEC2014 in the numerical experiments presented. BDE's success in solving the related benchmark problems is statistically compared with six state-of-the-art comparison algorithms. In this paper, three real-world optimization problems are also solved by using the proposed algorithm, BDE. According to statistics generated from the experimental results, BDE is statistically better than comparison methods in solving the related real-world problems. [ABSTRACT FROM AUTHOR]

Previous major earthquake events have revealed that soils susceptible to liquefaction are one of the factors causing significant damages to the structures. Therefore, accurate prediction of the liquefaction phenomenon is an important task in earthquake engineering. Over the past decade, several researchers have been extensively applied machine learning (ML) methods to predict soil liquefaction. This paper presents the prediction of soil liquefaction from the SPT dataset by using relatively new and robust tree-based ensemble algorithms, namely Adaptive Boosting, Gradient Boosting Machine, and eXtreme Gradient Boosting (XGBoost). The innovation points introduced in this paper are presented briefly as follows. Firstly, Stratified Random Sampling was utilized to ensure equalized sampling between each class selection. Secondly, feature selection methods such as Recursive Feature Elimination, Boruta, and Stepwise Regression were applied to develop models with a high degree of accuracy and minimal complexity by selecting the variables with significant predictive features. Thirdly, the performance of ML algorithms with feature selection methods was compared in terms of four performance metrics, Overall Accuracy, Precision, Recall, and F-measure to select the best model. Lastly, the best predictive model was determined using a statistical significance test called Wilcoxon's sign rank test. Furthermore, computational cost analyses of the tree-based ensemble algorithms were performed based on parallel and non-parallel processing. The results of the study suggest that all developed tree-based ensemble models could reliably estimate soil liquefaction. In conclusion, according to both validation and statistical results, the XGBoost with the Boruta model achieved the most stable and better prediction performance than the other models in all considered cases. [ABSTRACT FROM AUTHOR]

Well-honed CNN architectures trained with massive labeled images datasets are the state-of-the-art solution in many fields. In this paper, the weights of five commonly used pre-trained models are carefully analyzed for extracting their numerical characteristics and spatial distribution law. The general characteristics are: (1) the weights of a single convolutional layer conform to the distribution of symmetric power law. (2) the power exponent at the center of its convolutional kernel is relatively large, and the power exponent decreases radially from the center. (3) the value range of power exponents between layers is continuous from - 0.5 to - 3.5 . Based on these founding, a weight initialization method is proposed in order to speed up the convergence and improve the performance of CNN models. The proposed weight initialization method is compared with several commonly used methods. Extensive experiments show that it can improve the convergence speed of the CNN models, and the model accuracy is improved by 1–3%. [ABSTRACT FROM AUTHOR]

The integration of demand response programs (DRPs) into the energy management (EM) system of microgrids (MGs) helps in improving the load characteristics by allowing consumers to interoperate for achieving techno-economic advantages. In this paper, an improved algorithm is called LINFO is proposed for modifying search ability of the original weIghted meaN oF vectOrs (INFO) algorithm as well as avoiding its weaknesses like trapping in a local optima. The improved algorithm's efficiency is confirmed by comparing its results with those obtained by the original INFO and other optimization techniques using different standard benchmark test functions. Moreover, this improved algorithm and the original version are applied for solving the EM problem with the aim of optimizing the operation cost of the MGs in the presence DRPs. They are used to solve day-ahead EM problem for optimal operation of renewable energy resources, the optimal generation from a conventional diesel engines (DEs); taking into account the participation of customers in DRP for minimizing MG operating cost, which includes the cost of DEs fuel and the power transactions cost with the main grid. To demonstrate the efficacy of the proposed LINFO, simulation results are compared with the results of well-known and newly developed optimization techniques. [ABSTRACT FROM AUTHOR]

Graph neural networks (GNNs) had shown excellent performance in complex graph data modelings such as node classification, link prediction and graph classification. However, GNNs are vulnerable to adversarial attacks resulting in severe performance degradation, which brings many security and privacy issues. Such vulnerability of GNNs limits its application in safety–critical fields such as finance and transportation. Studying the principles and implementation of graph adversarial attacks and their countermeasures can provide insight into the reason behind the vulnerability of GNNs and consequently improve the robustness and generalization of the model. This paper introduces the related concepts of existing graph adversarial attack and defense algorithms and analyzes the basic idea and implementation of each algorithm. Moreover, we compare the strategies, target tasks, advantages and disadvantages of typical algorithms. Through the summary of the state of the art, the limitations and possible development directions of graph adversarial attacks and defenses are analyzed, which provides a useful reference for the further developing relative researches. [ABSTRACT FROM AUTHOR]

Intrusion Detection (ID) is an essential task in the cyberattacks domain built to secure Internet applications and networks from malicious actors. The main shortcoming of the current Intrusion Detection Systems (IDSs) is their attack detection performance because the existing solutions fail to detect the threats efficiently. Therefore, improving the prediction model of the IDS can improve its performance. Several studies suggest improving the Machine Learning (ML)-based IDS prediction models by selecting the most informative features from the security data before utilizing it in the ML-IDS model. Thus, meta-heuristic algorithms such as White Shark Optimizer (WSO) are adapted to deal with such feature selection problems. For the IDS prediction model, in this paper, the WSO algorithm is improved to cope with the binary domain of the feature selection task as follows: Firstly, two transfer functions are used to map the continuous domain into binary. Secondly, the modified K-means algorithm is proposed to assess building the initial population with a high level of diversity. Finally, several crossover operators are utilized to improve the evolution process of the binary WSO. The three improved versions are BIWSO1 which is WSO with transfer functions, BIWSO2, which is BIWSO1 with a modified k-means algorithm, and BIWSO3, which is BIWSO2 with crossover operators. The proposed versions of BIWSO are tested using twelve public real-world IDS and IoT datasets. Comparative evaluations against well-established meta-heuristic algorithms are conducted where the BIWSO3 proves its efficiency in terms of classification accuracy, precision, recall, and F1 measurements. For further validation, statistical evidence using Friedman's, Wilcoxon's, and Mann–Whitney U tests has been conducted where BIWSO3 proves its performance at a remarkably significant level. [ABSTRACT FROM AUTHOR]

Detection of small objects is one of the main challenges to be improved in deep learning, mainly due to the small number of pixels and scene's context, leading to a loss in performance. In this paper, we present an optimized approach based on deep object detection models that allow the detection of a higher number of elements and improve the score obtained for their class inference. The main advantage of the presented methodology is that it is not necessary to modify the internal structure of the selected convolutional neural network model or re-training for a specific scene. Our proposal is based on detecting initial regions to generate several sub-images using super-resolution (SR) techniques, increasing the number of pixels of the elements, and re-infer over these areas using the same pre-trained model. A reduced set of windows is calculated in the super-resolved image by analyzing a computed graph that describes the distances among the preliminary object detections. This analysis is done by finding maximal cliques on it. This way, the number of windows to be examined is diminished, significantly speeding up the detection process. This framework has been successfully tested on real traffic sequences obtained from the U.S. Department of Transportation. An increase of up to 44.6% is achieved, going from an average detection rate for the EfficientDet D4 model of 14.5% compared to 59.1% using the methodology presented for the first sequence. Qualitative experiments have also been performed over the Cityscapes and VisDrone datasets. [ABSTRACT FROM AUTHOR]

Scheduling complex applications as task graphs on finite computational resources assuring task interdependencies is a well-known NP-complete optimization problem. This problem is well-addressed for microprocessor systems but for Dynamically Reconfigurable Hardware (DRHW) systems in which, in addition to tasks, the reconfiguration time and complexity also have to be scheduled; this problem is more complicated. DRHW reconfiguration overhead is considerable and can be crucial for real-world applications. To deal with this overhead, in this paper, a meta-heuristic method named Graph-Oriented Simulated Annealing (GOSA) is proposed. By introducing an innovative graph and solution structures called schedule graphs, and also some controlling functions which are inherited from the nature of the problem, the proposed method adapts itself to the characteristics of the problem. This helps the algorithm to adjust its exploration and exploitation speed and accuracy according to the requirements of the given problem and consequently find high-quality solutions quickly. To demonstrate the performance of the proposed method, it was tested on several synthetic and real-world benchmark task graphs, and the results were compared with a selection of classic and state-of-the-art algorithms. The method is comprehensively evaluated by performing numerous experiments in terms of execution time, makespan, scalability, and reliability. The results of the experiments on benchmarks show that in terms of the quality of the solutions, GOSA outperforms BGA, HPSO-GA, and FATS by 17%, 13%, and 5%, respectively, and its execution time is considerably less than all competing algorithms. Moreover, according to the experiments done on synthetic graphs, the makespan of the solutions generated by GOSA, Genetic Algorithm (GA), and the Gxhaustive Search over the List Scheduler are improved on average by 7.2%, 8.1%, and 19.1%, respectively. The most significant achievement of the proposed method is its execution time which is 31 times faster than GA. Finally, the results confirm that the proposed method is scalable for large task graphs, and its reliability is superior. [ABSTRACT FROM AUTHOR]

Due to the wide zero-voltage-switching range and low power losses, triple-phase-shift (TPS) modulation is commonly utilized in dual-active-bridge (DAB) converters. However, it is difficult to model it and design its controller for the reasons of model uncertainties and nonlinearity. In this paper, a deep reinforcement learning (DRL)-supervised proportional–integral (PI) control algorithm is proposed. The PI controller is used as a base controller to stabilize the output voltage of the DAB converter. In order to improve the control accuracy and the dynamic performance, the PI parameters are tuned by DRL. Besides, all operation modes of the TPS are learned during the training process. Thus, the operation mode with maximum power efficiency can be selected under a wide operation range. The simulation comparison results demonstrate the efficacy and superiorities of the proposed method. [ABSTRACT FROM AUTHOR]

Agriculture faces a significant challenge in predicting crop yields, a critical aspect of decision-making at international, regional, and local levels. Crop yield prediction utilizes soil, climatic, environmental, and crop traits extracted via decision support algorithms. This paper presents a novel approach, the Crop Yield Prediction Algorithm (CYPA), utilizing IoT techniques in precision agriculture. Crop yield simulations simplify the comprehension of cumulative impacts of field variables such as water and nutrient deficits, pests, and illnesses during the growing season. Big data databases accommodate multiple characteristics indefinitely in time and space and can aid in the analysis of meteorology, technology, soils, and plant species characterization. The proposed CYPA incorporates climate, weather, agricultural yield, and chemical data to facilitate the anticipation of annual crop yields by policymakers and farmers in their country. The study trains and verifies five models using optimal hyper-parameter settings for each machine learning technique. The DecisionTreeRegressor achieved a score of 0.9814, RandomForestRegressor scored 0.9903, and ExtraTreeRegressor scored 0.9933. Additionally, we introduce a new algorithm based on active learning, which can enhance CYPA's performance by reducing the number of labeled data needed for training. Incorporating active learning into CYPA can improve the efficiency and accuracy of crop yield prediction, thereby enhancing decision-making at international, regional, and local levels. [ABSTRACT FROM AUTHOR]

Deep Neural Networks (DNNs) are widely regarded as the most effective learning tool for dealing with large datasets, and they have been successfully used in thousands of applications in a variety of fields. Based on these large datasets, they are trained to learn the relationships between various variables. The adaptive moment estimation (Adam) algorithm, a highly efficient adaptive optimization algorithm, is widely used as a learning algorithm in various fields for training DNN models. However, it needs to improve its generalization performance, especially when training with large-scale datasets. Therefore, in this paper, we propose HN Adam, a modified version of the Adam Algorithm, to improve its accuracy and convergence speed. The HN_Adam algorithm is modified by automatically adjusting the step size of the parameter updates over the training epochs. This automatic adjustment is based on the norm value of the parameter update formula according to the gradient values obtained during the training epochs. Furthermore, a hybrid mechanism was created by combining the standard Adam algorithm and the AMSGrad algorithm. As a result of these changes, the HN_Adam algorithm, like the stochastic gradient descent (SGD) algorithm, has good generalization performance and achieves fast convergence like other adaptive algorithms. To test the proposed HN_Adam algorithm performance, it is evaluated to train a deep convolutional neural network (CNN) model that classifies images using two different standard datasets: MNIST and CIFAR-10. The algorithm results are compared to the basic Adam algorithm and the SGD algorithm, in addition to other five recent SGD adaptive algorithms. In most comparisons, the HN Adam algorithm outperforms the compared algorithms in terms of accuracy and convergence speed. AdaBelief is the most competitive of the compared algorithms. In terms of testing accuracy and convergence speed (represented by the consumed training time), the HN-Adam algorithm outperforms the AdaBelief algorithm by an improvement of 1.0% and 0.29% for the MNIST dataset, and 0.93% and 1.68% for the CIFAR-10 dataset, respectively. [ABSTRACT FROM AUTHOR]

This paper proposes an e-diagnosis system based on machine learning (ML) algorithms to be implemented on the Internet of Medical Things (IoMT) environment, particularly for diagnosing diabetes mellitus (type 2 diabetes). However, the ML applications tend to be mistrusted because of their inability to show the internal decision-making process, resulting in slow uptake by end-users within certain healthcare sectors. This research delineates the use of three interpretable supervised ML models: Naïve Bayes classifier, random forest classifier, and J48 decision tree models to be trained and tested using the Pima Indians diabetes dataset in R programming language. The performance of each algorithm is analyzed to determine the one with the best accuracy, precision, sensitivity, and specificity. An assessment of the decision process is also made to improve the model. It can be concluded that a Naïve Bayes model works well with a more fine-tuned selection of features for binary classification, while random forest works better with more features. [ABSTRACT FROM AUTHOR]

To improve the effect of commercial insurance risk decision, this paper applies neural network algorithms to commercial insurance risk decision under the guidance of machine learning ideas, and selects the neural network algorithm based on the actual situation. Moreover, this paper analyzes the nature of risks of commercial insurance, analyzes the types of risks and risk relevance, constructs a commercial insurance risk decision model based on neural network algorithms, and determines the system process. In addition, this paper uses a combination method of qualitative and quantitative to identify the influencing factors of risk estimation to obtain relevant influencing factors, and verify the model proposed in this paper in combination with experimental research. From the experimental research results, it can be seen that the commercial insurance risk decision system based on neural network algorithm is very good in terms of decision effect. [ABSTRACT FROM AUTHOR]

Recently, substantial progress has been made in developing efficient algorithms for solving combinatorial optimization problems. In this paper, following this direction, the problem of designing supply chains which is an important combinatorial optimization problem is considered. The structural properties of supply chain models are investigated to transform such models into a generalized network optimization model. The transformation to a generalized network optimization model reduces the algorithms' solution time. Moreover, this paper proposes a new efficient hybrid algorithm based on tabu search and generalized network simplex algorithm (GN-TSA) for designing multi-objectives supply chain models. The developed algorithm's parameters are tuned properly, validated, and evaluated. The algorithm's performance is then compared to an exact algorithm embedded in the General Algebraic Modeling System (GAMS) and two metaheuristic algorithms, namely a linear programming simplex algorithm integrated with the tabu search approach (LP-TSA) and simulated annealing. The findings indicated that the GN-TSA obtains solutions very close to the exact algorithm with less computation time. In addition, the proposed algorithm outperforms the LP-TSA and simulated annealing in terms of computation time, while the quality of the solutions is the same as solutions obtained by LP-TSA and better than simulated annealing. On average, the results revealed that the reduction in computational time is more than 26.30% using GN-TSA compared to LP-TSA and simulated annealing. [ABSTRACT FROM AUTHOR]

The main objective of this paper is to present an improved neural network algorithm (INNA) for solving the reliability-redundancy allocation problem (RRAP) with nonlinear resource constraints. In this RRAP, both the component reliability and the redundancy allocation are to be considered simultaneously. Neural network algorithm (NNA) is one of the newest and efficient swarm optimization algorithms having a strong global search ability that is very adequate in solving different kinds of complex optimization problems. Despite its efficiency, NNA experiences poor exploitation, which causes slow convergence and also restricts its practical application of solving optimization problems. Considering this deficiency and to obtain a better balance between exploration and exploitation, searching procedure for NNA is reconstructed by implementing a new logarithmic spiral search operator and the searching strategy of the learner phase of teaching–learning-based optimization (TLBO) and an improved NNA has been developed in this paper. To demonstrate the performance of INNA, it is evaluated against seven well-known reliability optimization problems and finally compared with other existing meta-heuristics algorithms. Additionally, the INNA results are statistically investigated with the Wilcoxon sign-rank test and Multiple comparison test to show the significance of the results. Experimental results reveal that the proposed algorithm is highly competitive and performs better than previously developed algorithms in the literature. [ABSTRACT FROM AUTHOR]

Algorithm recommendation based on meta-learning was studied previously. The research on the meta-features extraction, which is a key for the success of recommendation, is lacking for multi-objective optimization problems (MOPs). This paper proposes four sets of meta-features to characterize MOPs. In addition, the algorithm recommendation model based on meta-learning is extended to the field of multi-objective optimization. To evaluate the efficiency and effectiveness of the extracted meta-features, 29 MOPs benchmark functions with different dimensions and two real-world MOPs are employed for comprehensive comparison. Experimental results show that the proposed meta-features in this paper can fully characterize MOPs and are empirically efficient for algorithm recommendation. [ABSTRACT FROM AUTHOR]

Total variation denoising (TVD) algorithm is suitable for the restoration of piecewise constant signals (PCS). The amplitude accuracy of discontinuities affects the overall denoising effect. Improving the amplitude accuracy of discontinuities is the difficulty of PCS denoising. Recent papers have presented TVD-based algorithms to improve the denoising precision. After applying the forward–backward splitting (FBS) method, the iteration of the denoising is comprised by the traditional TVD. This paper proposes a new upper bound function to enhance the accuracy of the traditional TVD method. Then, the enhanced TVD is used to update the TVD-based algorithm to improve the denoising efficiency. The experimental results demonstrate that the proposed method has superior performance compared to other methods in PCS denoising. [ABSTRACT FROM AUTHOR]

In order to improve the effectiveness of signal processing, according to the actual needs of signal processing and the current problems in signal processing, this paper introduces an improved LSTM recurrent neural network algorithm to construct the signal processing algorithm. Moreover, this paper sets up the functional structure of this paper based on the neural network model structure, and builds an intelligent analysis system for signal processing tasks based on the LSTM recurrent neural network algorithm. In addition, this paper analyzes the system algorithm flow in detail, and combines experimental research to carry out the effectiveness of the system constructed in this paper, and conducts quantitative analysis from two aspects of signal threshold prediction and signal processing effect. Finally, this paper conducts experimental results research with the support of mathematical statistics methods. From the research point of view, it can be known that the system constructed in this paper has good signal processing functions. [ABSTRACT FROM AUTHOR]

Many works have been done in stereo vision due to its critical importance to the different uses and various areas of computer vision. Recently, the popularity of research on the local block matching algorithm is difficult to compete with the convolutional neural network (CNN) approach, whereby the latter method can achieve higher accuracy at the expense of time consumption. However, the simplicity of the local block matching algorithm is easier to be implemented in real applications at a lower cost. Therefore, this paper presents a new study that improved computation and cost aggregation using a progressive approach. A new algorithm to aid in searching locally similar blocks called the ranking-gradient-similarity multi-block strategy (RGSMB) has been presented, improving processing time and accuracy. A new cost computation has been introduced in the proposed algorithm, which fully utilises the information from the limited local window region by combining the cost from three different constraints, including a Ranking constraint, a Gradient constraint, and a Similarity constraint. The proposed algorithm can accomplish remarkable accuracy by combining with multi-block matching (MBM), which is developed to diminish the conventional cost aggregation's constrained shape limitation. We evaluate the RGSMB algorithm's performance and other CNN methods for comparison. KITTI 2012 dataset of 194 and 195 image pairs and the KITTI 2015 dataset of 200 image pairs were used for training and testing. The results indicate that the proposed algorithm is better than the existing approaches, with a higher matching rate using the KITTI benchmark dataset. Compared to existing local block matching algorithms, the average errors of the RGS are less than the CT approach by nearly 9%, which is the 2nd ranked in this analysis. Also, the RGSMB has average errors of 21% less than the S 7 + correlation when comparing with existing CNN methods. It was also observed that the RGSMB could estimate the disparity value more accurately in the depth discontinuities and homogenous regions. [ABSTRACT FROM AUTHOR]

In sports, advance sensing technologies generate massive amount of unstructured telemedicine data that need to be refined for accurate diagnosis of underlying diseases. For accurate prediction of diseases and classification of athletes' data, deep learning algorithms are frequently used at the cloud. However, the transmission of raw data of athletes to the cloud faces numerous challenges. Among them, security and privacy are a major challenge in view of the sensitive and personal information present within the unstructured data. In this paper, first we present a data block scrambling algorithm (without key management) for secured transmission and storage of ECG (electrocardiogram) data of table tennis players at the cloud. A small piece of original data stored at the cloud is used for scrambling the massive amount of remaining ECG data. The secured telemedicine data is then imported into Hadoop Distributed File System for data management, which is read by Spark framework to form Resilient Distributed Datasets. Finally, a deep learning approach is used that extracts useful features, learns the related information, and weights and sums the feature vectors at different layers for classification. Theoretical analysis proves that our proposed approach is highly robust and resilient to brute force attacks and at the same time has a much better accuracy, sensitivity, and specificity as compared to the existing approaches. [ABSTRACT FROM AUTHOR]

A hybrid algorithm combines two or more individual algorithms in order to integrate the advantages of these individual algorithms. In recent years, a huge number of hybrid algorithms have been proposed, and their efficiencies have been demonstrated by experimental studies. However, few theoretical results that support the experimental conclusions are available. In this paper, we focus on theoretical investigation about the efficiency of hybrid algorithms for combinatorial optimization problems. We first analyze the runtime of three specific hybrid algorithms and their component algorithms on two instances. One is a well-known pseudo-Boolean function, and the other is an instance of a classical combinatorial optimization problem. We then theoretically investigate the relationship between the upper bounds of the expected runtime of three kinds of general hybrid algorithms and their component algorithms. [ABSTRACT FROM AUTHOR]

In this paper, a novel water cycle algorithm is presented by dynamically assigning sea and rivers and devising an enhanced position updating strategy. To effectively maintain the diversity of solutions and ensure the convergence of algorithm, an adaptive distance-based assignment mechanism is first developed to set sea, rivers and their corresponding streams. In this mechanism, the fitness values and position information of solutions are simultaneously considered, and the total number of sea and rivers is nonlinearly reduced during the search process. Meanwhile, an enhanced position updating strategy is designed to update the solutions by incorporating both the gravitational search and greedy strategy. Moreover, a modified evaporation operation is further proposed to dynamically refresh the search capability of algorithm by properly making full use of the promising information of solutions. Differing from the existing WCA variants, the proposed algorithm dynamically assigns sea, rivers and their streams, additionally incorporates the gravitational search and greedy strategy, and fully exploits the obtained promising information in the raining process. Then it could availably strengthen the search effectiveness and balance the exploration and exploitation. Finally, the performance of the proposed algorithm is evaluated by comparing with 12 typical algorithms on 30 CEC2017 benchmark functions. Numerical results show that the proposed algorithm has better performance. [ABSTRACT FROM AUTHOR]

The knapsack problem (KP) is a discrete combinatorial optimization problem that has different utilities in many fields. It is described as a non-polynomial time (NP) problem and has several applications in many fields. The differential evolution (DE) algorithm has been successful in solving continuous optimization problems, but it needs further work to solve discrete and binary optimization problems and avoid local optima. According to the literature, no DE search operator or algorithm is optimal for all optimization tasks. As a result, using more than one search operator in a single algorithm architecture, called multi-operator-based algorithms, is a solution to address this problem. These methods outperformed single-based methods for continuous optimization problems. Thus, in this paper, a binary multi-operator differential evolution (BMODE) approach is presented to tackle the 0–1 KP. The presented methodology utilizes multiple differential evolution (DE) mutation strategies with complementary characteristics, with the best mutation operator being asserted utilizing the produced solutions' quality and the population's diversity. In BMODE, two types of transfer functions (TFs) (S-shaped and V-shaped) are used to transfer the continuous solutions to binary ones to be able to calculate the fitness function value. To handle the capacity constraints, a feasibility rule is utilized and some of the infeasible solutions are repaired. The performance of BMODE is tested by solving 40 instances with multiple dimensions, i.e., low, medium, and high. Experimental results of the proposed BMODE are compared with well-known state-of-the-art 0–1 knapsack algorithms. Based on Wilcoxon's nonparametric statistical test ( α = 0.05 ), the proposed BMODE can obtain the best results against the rival algorithms in most cases, and can work well on stability and computational accuracy. [ABSTRACT FROM AUTHOR]

Computations that mimic nature are known as nature-inspired computing. Nature presents a wealthy source of thoughts and ideas for computing. The use of natural galvanized techniques has been found to provide machine solutions to complex problems. One of the challenging issues among researchers is high-dimensional data which contains a large number of unwanted, redundant, and irrelevant features. These redundant or unwanted features reduce the accuracy of machine learning models. Therefore, to solve this problem nowadays metaheuristic techniques are being used. The paper presents both surveys as well as comparison of five metaheuristic algorithms for feature selection. A wrapper-based feature selection approach using five nature-inspired techniques for feature selection has been applied. The binary version of the five swarm-based nature-inspired algorithms (NIAs), namely particle swarm optimization, whale optimization algorithm (WOA), grey wolf optimization (GWO), firefly algorithm, and bat algorithm. WOA and GWO are recent algorithms used for finding optimal feature subsets when there is no empirical information. The S-shape transfer function has been used to convert the continuous value to binary form and K-nearest neighbor is used to calculate the classification accuracy of selected feature subsets. To validate the results of the selected NIAs eleven benchmark datasets from the UCI repository are used. The strength of each NIA has been verified using a nonparametric test called the Friedman rank and Holm test. p value obtained shows that WOA is statistically significant and performs better than other models. [ABSTRACT FROM AUTHOR]

The discounted {0-1} knapsack problem (DKP) is an NP-hard problem that is more challenging than the classical knapsack problem. In this paper, an enhanced version of the sine-cosine algorithm (SCA) called TMHSCA is proposed to solve the DKP. The SCA is a novel metaheuristic algorithm based on sine-cosine theory. Three effective improvements are proposed for the SCA to enhance the convergence speed and exploitation capability of the algorithm. First, by dividing the initial population into three subpopulations using random, greedy and opposition-based learning (OBL) strategies, the diversity of the population can be effectively enhanced. Furthermore, a forward-inverse sine-cosine search that expands the search space by utilizing the worst solution as another target is proposed. The proposed forward-inverse sine-cosine search strategy facilitates the algorithm to explore the easily ignored solution space. The last improvement includes the use of an OBL mutation and the mutation operator of differential evolution at a two-stage mutation, which increases the algorithm's convergence speed and fitness. In our proposed two-stage mutation, the number of mutations can be adaptively adjusted based on the optimization capabilities of the optimal individual. Additionally, a reshuffling repair strategy for repairing infeasible solutions by sorting items into groups to obtain the weight-to-value ratio is proposed. Extensive experiments are conducted on 40 publicly available datasets and are compared to the greedy algorithm, 3 SCAs and 8 competitive baseline methods. The experimental results demonstrate that our method achieves a state-of-the-art performance. [ABSTRACT FROM AUTHOR]

The evolutionary game theory aims to simulate different decision strategies in populations of individuals and to determine how the population evolves. Compared to strategies between two agents, such as cooperation or noncooperation, strategies on multiple agents are rather challenging and difficult to be simulated via traditional methods. Particularly, in a knowledge spillover problem (KSP), cooperation strategies among more than hundreds of individuals need to be simulated. At the same time, the brain storm optimization (BSO) algorithm, which is a data-driven and model-driven hybrid paradigm, has the potential to simulate the complex behaviors in a group of simple individuals. In this paper, a modified BSO algorithm has been used to solve KSP from the perspective of evolutionary game theory. Knowledge spillover (KS) is the sharing or exchanging of knowledge resources among individuals. Firstly, the KS and evolutionary game theory were introduced. Then, the KS model and KS optimization problems were built from the evolutionary game perspective. Lastly, the modified BSO algorithms were utilized to solve KS optimization problems. Based on the applications of BSO algorithms for KSP, the properties of different swarm optimization algorithms can be understood better. More efficient algorithms could be designed to solve different real-world evolutionary game problems. [ABSTRACT FROM AUTHOR]

The optimal design of Power System Stabilizer (PSS) parameters is a significant optimization problem in power systems. Metaheuristic search (MHS) algorithms are among the most commonly used methods for optimizing PSS parameters. The preferred MHS algorithm must have strong exploration capability and an effective exploitation-exploration balance. The Fitness-Distance Balance (FDB) is a novel method for MHS algorithms, and it is featured by balanced search and effective diversity capabilities. In this paper, the exploration and balanced search capabilities of the Social Network Search (SNS) have been developed by using the FDB method. The FDB method guides the search process in the SNS and enables a more efficient selection of solution candidates in the search space. The performance of FDB-based SNS (FDBSNS) has been tested in the CEC-2014 and CEC-2017 benchmark test suits, and its superiority against SNS has been verified. Moreover, the effect of the FDBSNS was investigated in WSCC 3-machine 9-bus and 10-machine 39-bus New England test systems in the optimal PSS design problem. The obtained results and statistical analyses demonstrated that FDBSNS is a robust MHS algorithm compared to the algorithms in the literature for solving the CEC benchmark suits and optimal PSS design problem. [ABSTRACT FROM AUTHOR]

The heat exchanger has been widely used in the energy and chemical industry and plays an irreplaceable role in the featured applications. The design of heat exchanger is a mixed integer complex optimization problem, where the efficient design significantly improves the efficiency and reduces the cost. Many intelligent methods have been developed for heat exchanger optimal design. In this paper, a novel variant of sine and cosine algorithm named EDOLSCA is proposed, enhanced by dynamic opposite learning algorithm and the elite strategy. The proposed method is tested in CEC2014 benchmark and proved to be of significant advantages over the original algorithm. The new algorithm is then validated in the plate-fin heat exchanger (PFHE) optimal design problem. The comparison results of the proposed algorithm and other algorithms prove that EDOLSCA also has demonstrated superiority in heat exchanger optimal design. [ABSTRACT FROM AUTHOR]

This paper proposes an Advanced Actor-Critic algorithm, which is improved based on the conventional Actor-Critic algorithm, to train the agent to play the complex strategy game StarCraft II. A series of advanced features have been incorporated, including the distributional advantage estimation, information entropy-based uncertainty estimation, self-confidence-based exploration, and normal constraint-based update strategy. A case study including seven StarCraft II mini-games is investigated to identify the effectiveness of the proposed approach, where the famous A3C algorithm is adopted as the comparative baseline. The results verify the superiority of the improved algorithm in accuracy and training efficacy, in complex environment with high-dimensional and hybrid state and action space. [ABSTRACT FROM AUTHOR]

This paper proposes the multi-objective variant of the recently-introduced fitness dependent optimizer (FDO). The algorithm is called a multi-objective fitness dependent optimizer (MOFDO) and is equipped with all five types of knowledge (situational, normative, topographical, domain, and historical knowledge) as in FDO. MOFDO is tested on two standard benchmarks for the performance-proof purpose: classical ZDT test functions, which is a widespread test suite that takes its name from its authors Zitzler, Deb, and Thiele, and on IEEE Congress of Evolutionary Computation benchmark (CEC-2019) multi-modal multi-objective functions. MOFDO results are compared to the latest variant of multi-objective particle swarm optimization, non-dominated sorting genetic algorithm third improvement (NSGA-III), and multi-objective dragonfly algorithm. The comparative study shows the superiority of MOFDO in most cases and comparative results in other cases. Moreover, MOFDO is used for optimizing real-world engineering problems (e.g., welded beam design problems). It is observed that the proposed algorithm successfully provides a wide variety of well-distributed feasible solutions, which enable the decision-makers to have more applicable-comfort choices to consider. [ABSTRACT FROM AUTHOR]

Metaheuristics are algorithms that are used to solve difficult optimization problems. They are typically stochastic approaches; hence, proper statistical tests are needed to compare them. However, choosing an appropriate statistical test is not trivial given that each test requires some assumptions to be true before the test can be used. Moreover, the p-values associated with a statistical test is usually difficult to interpret. In this paper, we propose the use of Laplace's rule of succession to compare different metaheuristic approaches. The rule is simple, intuitive and easy to compute. It can be used alone or to complement a statistical test. The process of using the rule for comparison purposes is clearly explained and applied to a typical scenario encountered in the field of metaheuristics. In this scenario, an improved variant of an existing metaheuristic algorithm is proposed. To evaluate the performance of the two algorithms, Laplace's rule and a traditional statistical test are used. Analysis of the results and how to interpret them are provided. The results show that Laplace's rule is consistent with the used statistical test. Furthermore, the rule is easier to compute and interpret. [ABSTRACT FROM AUTHOR]

Rib segmentation based on chest X-ray images is essential in the computer-aided diagnosis systems of lung cancer, which serves as an important step in the quantitative analysis of various types of lung diseases. However, the traditional methods are unable to segment ribs effectively due to the unclear edges and overlapping regions in X-ray images. A novel rib segmentation framework based on Unpaired Sample Augmentation and Multi-Scale Network is presented in this paper, aiming to improve the accuracy of ribs segmentation with limited labeled samples. First, the algorithm learns pneumonia-related texture changes via unpaired chest x-ray images and generates various augmented samples. Then, a multi-scale network attempts to learn hierarchical features using global supervision. Finally, the refined segmentation result of each organ is achieved by using a deep separation module and a comprehensive loss function. Specifically, the hierarchical features can greatly improve the robustness of multi-organ segmentation networks. The complex multi-organ segmentation task with limited labeled data is simplified with the designed deep separation module. We justify the proposed framework through extensive experiments. It achieves good performance with DSC, Precision, Recall, and Jaccard of 88.03, 88.25, 88.36, and 79.02%, respectively. The DSC value increases nearly by 3% compared to other popular methods. The experimental results show that our algorithm presents better segmentation performance for the overlapping region and fuzzy region of multiple organs, which holds research value and prospects for application. [ABSTRACT FROM AUTHOR]

With graph structure data becoming more common in practical problems, graph neural networks have shown their potential for processing graph structure data. Now, a heterogeneous graph containing different types of nodes and edges as a more complex type of graph structure is the hot research area. However, existing studies only focus on supervised methods/semi-supervised methods for heterogeneous graph, which require a lot of labeled data, demanding lots of money and time. In this paper, we propose a novel framework, i.e., the self-supervised contrastive learning for heterogeneous graph based on multi-pretext tasks (HGMT). In our proposed HGMT, considering the essential and inherent properties of heterogeneous graph, we utilize both metapath-based encoder and network schema-based encoder to get two different node representations to construct contrastive samples. And then, we design three complementary pretext tasks, i.e., local–local self-supervised contrastive learning task, local–context self-supervised contrastive learning task and local–global self-supervised contrastive learning task, which can make model have a deeper understanding of the interaction relationships on local, context and global structure information of heterogeneous graph. At last, we view it as multitask learning to train our proposed HGMT model by balancing these three pretext tasks. Empirical results validate the performance of our proposed HGMT compared with the existing state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

Discovering a hearing disorder at an earlier intervention is critical for reducing the effects of hearing loss and the approaches to increase the remaining hearing ability can be implemented to achieve the successful development of human communication. Recently, the explosive dataset features have increased the complexity for audiologists to decide the proper treatment for the patient. In most cases, data with irrelevant features and improper classifier parameters causes a crucial influence on the audiometry system in terms of accuracy. This is due to the dependent processes of these two, where the classification accuracy performance could be worsened if both processes are conducted independently. Although the filter algorithm is capable of eliminating irrelevant features, it still lacks the ability to consider feature reliance and results in a poor selection of significant features. Improper kernel parameter settings may also contribute to poor accuracy performance. In this paper, an ensemble filters feature selection based on Information Gain (IG), Gain Ratio (GR), Chi-squared (CS), and Relief-F (RF) with harmonize optimization of Particle Swarm Optimization (PSO) and Support Vector Machine (SVM) is presented to mitigate these problems. Ensemble filters are utilized so that the initial top dominant features relevant for classification can be considered. Then, PSO and SVM are optimized simultaneously to achieve the optimal solution. The results on a standard Audiology dataset show that the proposed method produces 96.50% accuracy with optimal solution compared to classical SVM, which signifies the proposed method is effective in handling high dimensional data for hearing disorder prediction. [ABSTRACT FROM AUTHOR]