Showing papers in "Knowledge Based Systems in 2022"

Snake Optimizer: A novel meta-heuristic optimization algorithm

Abstract: In recent years, several metaheuristic algorithms have been introduced in engineering and scientific fields to address real-life optimization problems. In this study, a novel nature-inspired metaheuristics algorithm named as Snake Optimizer (SO) is proposed to tackle a various set of optimization tasks which imitates the special mating behavior of snakes. Each snake (male/female) fights to have the best partner if the existed quantity of food is enough and the temperature is low. This study mathematically mimics and models such foraging and reproduction behaviors and patterns to present a simple and efficient optimization algorithm. To verify the validity and superiority of the proposed method, SO is tested on 29 unconstrained Congress on Evolutionary Computation (CEC) 2017 benchmark functions and four constrained real-world engineering problems. SO is compared with other 9 well-known and newly developed algorithms such as Linear population size reduction-Success-History Adaptation for Differential Evolution (L-SHADE), Ensemble Sinusoidal incorporated with L-SHADE (LSHADE-EpSin), Covariance matrix adaptation evolution strategy (CMAES), Coyote Optimization Algorithm (COA), Moth-flame Optimization, Harris Hawks Optimizer, Thermal Exchange optimization, Grasshopper Optimization Algorithm, and Whale Optimization Algorithm. Experimental results and statistical comparisons prove the effectiveness and efficiency of SO on different landscapes with respect to exploration–exploitation balance and convergence curve speed. The source code is currently available for public from: https://se.mathworks.com/matlabcentral/fileexchange/106465-snake-optimizer

Aspect-based sentiment analysis via affective knowledge enhanced graph convolutional networks

Abstract: Aspect-based sentiment analysis is a fine-grained sentiment analysis task, which needs to detection the sentiment polarity towards a given aspect. Recently, graph neural models over the dependency tree are widely applied for aspect-based sentiment analysis. Most existing works, however, they generally focus on learning the dependency information from contextual words to aspect words based on the dependency tree of the sentence, which lacks the exploitation of contextual affective knowledge with regard to the specific aspect. In this paper, we propose a graph convolutional network based on SenticNet to leverage the affective dependencies of the sentence according to the specific aspect, called Sentic GCN . To be specific, we explore a novel solution to construct the graph neural networks via integrating the affective knowledge from SenticNet to enhance the dependency graphs of sentences. Based on it, both the dependencies of contextual words and aspect words and the affective information between opinion words and the aspect are considered by the novel affective enhanced graph model. Experimental results on multiple public benchmark datasets illustrate that our proposed model can beat state-of-the-art methods.

Aspect-based sentiment analysis via affective knowledge enhanced graph convolutional networks

Abstract: Aspect-based sentiment analysis is a fine-grained sentiment analysis task, which needs to detection the sentiment polarity towards a given aspect. Recently, graph neural models over the dependency tree are widely applied for aspect-based sentiment analysis. Most existing works, however, they generally focus on learning the dependency information from contextual words to aspect words based on the dependency tree of the sentence, which lacks the exploitation of contextual affective knowledge with regard to the specific aspect. In this paper, we propose a graph convolutional network based on SenticNet to leverage the affective dependencies of the sentence according to the specific aspect, called Sentic GCN. To be specific, we explore a novel solution to construct the graph neural networks via integrating the affective knowledge from SenticNet to enhance the dependency graphs of sentences. Based on it, both the dependencies of contextual words and aspect words and the affective information between opinion words and the aspect are considered by the novel affective enhanced graph model. Experimental results on multiple public benchmark datasets illustrate that our proposed model can beat state-of-the-art methods.

Bi-CLKT: Bi-Graph Contrastive Learning based Knowledge Tracing

Abstract: The goal of Knowledge Tracing (KT) is to estimate how well students have mastered a concept based on their historical learning of related exercises. The benefit of knowledge tracing is that students’ learning plans can be better organised and adjusted, and interventions can be made when necessary. With the recent rise of deep learning, Deep Knowledge Tracing (DKT) has utilised Recurrent Neural Networks (RNNs) to accomplish this task with some success. Other works have attempted to introduce Graph Neural Networks (GNNs) and redefine the task accordingly to achieve significant improvements. However, these efforts suffer from at least one of the following drawbacks: (1) they pay too much attention to details of the nodes rather than to high-level semantic information; (2) they struggle to effectively establish spatial associations and complex structures of the nodes; and (3) they represent either concepts or exercises only, without integrating them. Inspired by recent advances in self-supervised learning, we propose a Bi-Graph Contrastive Learning based Knowledge Tracing (Bi-CLKT) to address these limitations. Specifically, we design a two-layer comparative learning scheme based on an “exercise-to-exercise” (E2E) relational subgraph. It involves node-level contrastive learning of subgraphs to obtain discriminative representations of exercises, and graph-level contrastive learning to obtain discriminative representations of concepts. Moreover, we designed a joint contrastive loss to obtain better representations and hence better prediction performance. Also, we explored two different variants, using RNN and memory-augmented neural networks as the prediction layer for comparison to obtain better representations of exercises and concepts respectively. Extensive experiments on four real-world datasets show that the proposed Bi-CLKT and its variants outperform other baseline models.

Dispersed foraging slime mould algorithm: Continuous and binary variants for global optimization and wrapper-based feature selection

Abstract: The slime mould algorithm (SMA) is a logical swarm-based stochastic optimizer that is easy to understand and has a strong optimization capability. However, the SMA is not suitable for solving multimodal and hybrid functions. Therefore, in the present study, to enhance the SMA and maintain population diversity, a dispersed foraging SMA (DFSMA) with a dispersed foraging strategy is proposed. We conducted extensive experiments based on several functions in IEEE CEC2017. The DFSMA were compared with 11 other meta-heuristic algorithms (MAs), 10 advanced algorithms, and 3 recently proposed algorithms. Moreover, to conduct more systematic data analyses, the experimental results were further evaluated using the Wilcoxon signed-rank test. The DFSMA was shown to outperform other optimizers in terms of convergence speed and accuracy. In addition, the binary DFSMA (BDFSMA) was obtained using the transform function. The performance of the BDFSMA was evaluated on 12 datasets in the UCI repository. The experimental results reveal that the BDFSMA performs better than the original SMA, and that, compared with other optimization algorithms, it improves classification accuracy and reduces the number of selected features, demonstrating its practical engineering value in spatial search and feature selection. • An improved slime mould algorithm (DFSMA) is proposed for feature selection. • The performance of DFSMA is verified by comparing with several famous algorithms. • DFSMA has faster convergence speed and accuracy compared with others. • DFSMA has achieved higher classification accuracy and smaller number of features.

Beluga whale optimization: A novel nature-inspired metaheuristic algorithm

Abstract: In this paper, a novel swarm-based metaheuristic algorithm inspired from the behaviors of beluga whales, called beluga whale optimization (BWO), is presented to solve optimization problem. Three phases of exploration, exploitation and whale fall are established in BWO, corresponding to the behaviors of pair swim, prey, and whale fall, respectively. The balance factor and probability of whale fall in BWO are self-adaptive which play significant roles to control the ability of exploration and exploitation. Besides, the Levy flight is introduced to enhance the global convergence in the exploitation phase. The effectiveness of the proposed BWO is tested using 30 benchmark functions, with qualitative, quantitative and scalability analysis, and the statistical results are compared with 15 other metaheuristic algorithms. According to the results and discussion, BWO is a competitive algorithm in solving unimodal and multimodal optimization problems, and the overall rank of BWO is the first in the scalability analysis of benchmark functions among compared metaheuristic algorithms through the Friedman ranking test. Finally, four engineering problems demonstrate the merits and potential of BWO in solving complex real-world optimization problems. The source code of BWO is currently available to public: https://ww2.mathworks.cn/matlabcentral/fileexchange/112830-beluga-whale-optimization-bwo/ . • A novel metaheuristic algorithm named as Beluga Whale Optimization (BWO) is proposed. • The behaviors of swim, prey and whale fall are designed on BWO algorithm. • The BWO is tested on 30 well-known benchmark functions and 4 engineering problems. • The BWO is compared with 15 well-known metaheuristic algorithms. • The BWO outperforms comparing algorithms in benchmark functions, especially for scalability of dimension.

An enhanced black widow optimization algorithm for feature selection

Abstract: Feature selection is an important data processing method to reduce dimension of the raw datasets while preserving the information as much as possible. In this paper, an enhanced version of Black Widow Optimization Algorithm called SDABWO is proposed to solve the feature selection problem. The Black Widow Optimization Algorithm (BWO) is a new population-based meta-heuristic algorithm inspired by the evolution process of spider population. Three main improvements were included into the BWO to overcome the shortcoming of low accuracy, slow convergence speed and being easy to fall into local optima. Firstly, a novel strategy for selecting spouses by calculating the weight of female spiders and the distance between spiders is proposed. By applying the strategy to the original algorithm, it has faster convergence speed and higher accuracy. The second improvement includes the use of mutation operator of differential evolution at mutation phase of BWO which helps the algorithm escape from the local optima. And then, three key parameters are set to adjust adaptively with the increase of iteration times. To confirm and validate the performance of the improved BWO, other 10 algorithms are used to compared with the SDABWO on 25 benchmark functions. The results show that the proposed algorithm enhances the exploitation ability, improves the convergence speed and is more stable when solving optimization problems. Furthermore, the proposed SDABWO algorithm is employed for feature selection. Twelve standard datasets from UCI repository prove that SDABWO-based method has stronger search ability in the search space of feature selection than the other five popular feature selection methods. These results confirm the capability of the proposed method simultaneously improve the classification accuracy while reducing the dimensions of the original datasets. Therefore, SDABWO-based method was found to be one of the most promising for feature selection problem over other approaches that are currently used in the literature.

An enhanced black widow optimization algorithm for feature selection

Abstract: Feature selection is an important data processing method to reduce dimension of the raw datasets while preserving the information as much as possible. In this paper, an enhanced version of Black Widow Optimization Algorithm called SDABWO is proposed to solve the feature selection problem. The Black Widow Optimization Algorithm (BWO) is a new population-based meta-heuristic algorithm inspired by the evolution process of spider population. Three main improvements were included into the BWO to overcome the shortcoming of low accuracy, slow convergence speed and being easy to fall into local optima. Firstly, a novel strategy for selecting spouses by calculating the weight of female spiders and the distance between spiders is proposed. By applying the strategy to the original algorithm, it has faster convergence speed and higher accuracy. The second improvement includes the use of mutation operator of differential evolution at mutation phase of BWO which helps the algorithm escape from the local optima. And then, three key parameters are set to adjust adaptively with the increase of iteration times. To confirm and validate the performance of the improved BWO, other 10 algorithms are used to compared with the SDABWO on 25 benchmark functions. The results show that the proposed algorithm enhances the exploitation ability, improves the convergence speed and is more stable when solving optimization problems. Furthermore, the proposed SDABWO algorithm is employed for feature selection. Twelve standard datasets from UCI repository prove that SDABWO-based method has stronger search ability in the search space of feature selection than the other five popular feature selection methods. These results confirm the capability of the proposed method simultaneously improve the classification accuracy while reducing the dimensions of the original datasets. Therefore, SDABWO-based method was found to be one of the most promising for feature selection problem over other approaches that are currently used in the literature.

A game-based deep reinforcement learning approach for energy-efficient computation in MEC systems

Abstract: Many previous energy-efficient computation optimization works for mobile edge computing (MEC) systems have been based on the assumption of synchronous offloading, where all mobile devices have the same data arrival time or calculation deadline in orthogonal frequency division multiple access (OFDMA) or time division multiple access (TDMA) systems. However, the actual offloading situations are more complex than synchronous offloading following the first-come, first-served rule. In this paper, we study a polling callback energy-saving offloading strategy, that is, the arrival time of data transmission and task processing time are asynchronous. Under the constraints of task processing time, the time-sharing MEC data transmission problem is modeled as the total energy consumption minimization model. Using the semi-closed form optimization technology, energy consumption optimization is transformed into two subproblems: computation (data partition) and transmission (time division). To reduce the computational complexity of offloading computation under time-varying channel conditions, we propose a game-learning algorithm, that combines DDQN and distributed LMST with intermediate state transition (named DDQNL-IST). DDQNL-IST combines distributed LSTM and double-Q learning as part of the approximator to improve the ability of processing and predicting time intervals and delays in time series. The proposed DDQNL-IST algorithm ensures rationality and convergence. Finally, the simulation results show that our proposed algorithm performs better than the DDQN, DQN and BCD-based optimal methods.

Renewable quantile regression for streaming datasets

Abstract: Streaming data analysis has drawn much attention, where large amounts of data arrive in streams. Because limited memory can only store a small batch of data, fast analysis without access to the historical data is necessary. Quantile regression has been widely used in many fields because of its robustness and comprehensiveness. However, in the streaming data environment, it is challenging to implement quantile regression by the conventional methods, because they are all based on the assumption that the memory can fit all the data. To fix this issue, this paper proposes a novel online renewable quantile regression strategy, in which the resulting estimator is renewed with current data and summary statistics of historical data. Thus, it is computationally efficient, and not storage-intensive. What is more, the theoretical results also confirm that the proposed estimator is asymptotically equivalent with the oracle estimator calculated using the entire data together. Numerical experiments on both synthetic and real data verify the theoretical results and illustrate the good performance of the new method.

Dynamic Levy Flight Chimp Optimization

Abstract: Background: The Chimp Optimization Algorithm (ChOA) is a hunting-based model and can be utilized as a set of optimization rules to tackle optimization problems. Due to agents’ insufficient diversity in some complex problems, this algorithm is sometimes exposed to local optima stagnation. Objective: This paper introduces a Dynamic Levy Flight (DLF) technique to smoothly and gradually transit the search agents from the exploration phase to the exploitation phase. Methods: To investigate the efficiency of the DLFChOA, this paper evaluates the performance of DLFChOA on twenty-three standard benchmark functions, twenty challenging functions of CEC-2005, ten suit tests of IEEE CEC06-2019, and twelve real-world optimization problems. The results are compared to benchmark optimization algorithms, including CMA-ES, SHADE, ChOA, HGSO, LGWO and ALEP (as the best benchmark Levy-based algorithms), and eighteen state-of-the-art algorithms (as the winners of the CEC2019, the GECCO2019, and the SEMCCO2019). Result and conclusion: Among forty-three numerical test functions, DLFChOA and CMA-ES gain the first and second rank with thirty and eleven best results. In the 100-digit challenge, jDE100 with a score of 100 provides the best results, followed by DISHchain1e+12, and DLFChOA with a score of 85.68 is ranked fifth among eighteen state-of-the-art algorithms achieved the best score in seven out of ten problems. Finally, DLFChOA and CMA-ES respectively gain the best results in five and four real-world engineering problems.

An enhanced manta ray foraging optimization algorithm for shape optimization of complex CCG-Ball curves

Abstract: The shape optimization of complex curves is a crucial and intractable technique in computer aided geometric design and widely used in many product design and manufacturing fields involving complex curve modeling. In this paper, an enhanced manta ray foraging optimization (MRFO) algorithm is used to optimize the shape of complex composite cubic generalized Ball (CCG-Ball, for short) curves. Firstly, to solve the problems of shape optimization for Ball curves, we construct a class of new cubic generalized Ball basis, and then present the CCG-Ball curves with multiple shape parameters based on the constructed basis functions. The shapes of the curves can be modified and optimized easily by using the shape parameters. Secondly, the shape optimization of CCG-Ball curves is mathematically an optimization problem that can be efficiently dealt with by swarm intelligence algorithm. In this regard, an enhanced MRFO called WMQIMRFO algorithm, combined with control parameter adjustment, wavelet mutation and quadratic interpolation strategy, is developed to enhance its capability of jumping out of the local minima and improve the calculation accuracy of the native algorithm. Furthermore, the superiority of the WMQIMRFO algorithm is verified by comparing with standard MRFO, other improved MRFO and popular nature-inspired optimization algorithms on the well-known CEC’14 and CEC’17 test suite as well as four engineering optimization problems, respectively. Finally, by minimizing the bending energy of the CCG-Ball curves as the evaluation standard, the shape optimization models of the curves with 1th-order and 2th-order geometric continuity are established, respectively. The WMQIMRFO algorithm is utilized to solve the established models, and the CCG-Ball curves with minimum energy are obtained. Some representative numerical examples illustrate the ability of the proposed WMQIMRFO algorithm in effectively solving the shape optimization problems of complex CCG-Ball curves in terms of precision, robustness, and convergence characteristics.

Fast forest fire smoke detection using MVMNet

Abstract: Forest fires are a huge ecological hazard, and smoke is an early characteristic of forest fires. Smoke is present only in a tiny region in images that are captured in the early stages of smoke occurrence or when the smoke is far from the camera. Furthermore, smoke dispersal is uneven, and the background environment is complicated and changing, thereby leading to inconspicuous pixel-based features that complicate smoke detection. In this paper, we propose a detection method called multioriented detection based on a value conversion-attention mechanism module and Mixed-NMS (MVMNet). First, a multioriented detection method is proposed. In contrast to traditional detection techniques, this method includes an angle parameter in the data loading process and calculates the target’s rotation angle using the classification prediction method, which has reference significance for determining the direction of the fire source. Then, to address the issue of inconsistent image input size while preserving more feature information, Softpool-spatial pyramid pooling (Soft-SPP) is proposed. Next, we construct a value conversion-attention mechanism module (VAM) based on the joint weighting strategy in the horizontal and vertical directions, which can specifically extract the colour and texture of the smoke. Ultimately, the DIoU-NMS and Skew-NMS hybrid nonmaximum suppression methods are employed to address the issues of smoke false detection and missed detection. Experiments are conducted using the homemade forest fire multioriented detection dataset, and the results demonstrate that compared to the traditional detection method, our model’s mAP reaches 78.92%, mAP 50 reaches 88.05%, and FPS reaches 122.

Unified medical image segmentation by learning from uncertainty in an end-to-end manner

Abstract: Automatic segmentation is a fundamental task in computer-assisted medical image analysis. Convolutional neural networks (CNNs) have been widely used for medical image segmentation tasks. Currently, most deep learning-based methods output a probability map and use a hand-crafted threshold to generate the final segmentation result, while how confident the network is of the probability map remains unclear. The segmentation result can be quite unreliable even though the probability is much higher than the threshold since the uncertainty of the probability can also be high. Moreover, boundary information loss caused by consecutive pooling layers and convolution strides makes the object’s boundary in segmentation even more unreliable. In this paper, we propose an uncertainty guided network referred to as UG-Net for automatic medical image segmentation. Different from previous methods, our UG-Net can learn from and contend with uncertainty by itself in an end-to-end manner. Specifically, UG-Net consists of three parts: a coarse segmentation module (CSM) to obtain the coarse segmentation and the uncertainty map, an uncertainty guided module (UGM) to leverage the obtained uncertainty map in an end-to-end manner, and a feature refinement module (FRM) embedded with several dual attention (DAT) blocks to generate the final segmentations. In addition, to formulate a unified segmentation network and extract richer context information, a multi-scale feature extractor (MFE) is inserted between the encoder and decoder of the CSM. Experimental results show that the proposed UG-Net outperforms the state-of-the-art methods on nasopharyngeal carcinoma (NPC) segmentation, lung segmentation, optic disc segmentation and retinal vessel detection. • A deep learning-based network named UG-Net is proposed for medical image segmentation. • An uncertainty guided module is proposed to learn from uncertainty in an end-to-end manner. • A feature refinement module with dual attention mechanism is designed for further performance promotion. • A multi-scale feature extractor is devised to fit into different segmentation tasks.

Designing ECG monitoring healthcare system with federated transfer learning and explainable AI

Abstract: Deep learning plays a vital role in classifying different arrhythmias using electrocardiography (ECG) data. Nevertheless, training deep learning models normally requires a large amount of data and can lead to privacy concerns. Unfortunately, a large amount of healthcare data cannot be easily collected from a single silo. Additionally, deep learning models are like black-box, with no explainability of the predicted results, which is often required in clinical healthcare. This limits the application of deep learning in real-world health systems. In this paper, to address the above-mentioned challenges, we design a novel end-to-end framework in a federated setting for ECG-based healthcare using explainable artificial intelligence (XAI) and deep convolutional neural networks (CNN). The federated setting is used to solve challenges such as data availability and privacy concerns. Furthermore, the proposed framework effectively classifies different arrhythmias using an autoencoder and a classifier, both based on a CNN. Additionally, we propose an XAI-based module on top of the proposed classifier for interpretability of the classification results, which helps clinical practitioners to interpret the predictions of the classifier and to make quick and reliable decisions. The proposed framework was trained and tested using the baseline Massachusetts Institute of Technology - Boston’s Beth Israel Hospital (MIT-BIH) Arrhythmia database. The trained classifier outperformed existing work by achieving accuracy up to 94.5% and 98.9% for arrhythmia detection using noisy and clean data, respectively, with five-fold cross-validation. We also propose a new communication cost reduction method to reduce the communication costs and to enhance the privacy of users’ data in the federated setting. While the proposed framework was tested and validated for ECG classification, it is general enough to be extended to many other healthcare applications.

Meta-learning as a promising approach for few-shot cross-domain fault diagnosis: Algorithms, applications, and prospects

Abstract: The advances of intelligent fault diagnosis in recent years show that deep learning has strong capability of automatic feature extraction and accurate identification for fault signals. Nevertheless, data scarcity and varying working conditions can degrade the performance of the model. More recently, a tool has been proposed to address the above challenges simultaneously. Meta-learning, also known as learning to learn, uses a small sample to quickly adapt to a new task. It has great application potential in few-shot and cross-domain fault diagnosis, and thus has become a promising tool. However, there is a lack of a survey to conclude existing work and look into the future. This paper comprehensively investigates deep meta-learning in fault diagnosis from three views: (i) what to use, (ii) how to use, and (iii) how to develop, i.e. algorithms, applications, and prospects. Algorithms are illustrated by optimization-, metric-, and model-based methods, the applications are concluded in few-shot cross-domain fault diagnosis, and open challenges, as well as prospects, are given to motivate the future work. Additionally, we demonstrate the performance of three approaches on two few-shot cross-domain tasks. Typical meta-learning methods are implemented and available at https://github.com/fyancy/MetaFD .

A multi-source information transfer learning method with subdomain adaptation for cross-domain fault diagnosis

Abstract: In modern industrial equipment maintenance, transfer learning is a promising tool that has been widely utilized to solve the problem of the insufficient generalization ability of diagnostic models, caused by changes in working conditions. However, owing to the single knowledge transfer source and fuzzy marginal distribution matching, the ability of traditional transfer learning methods for cross-domain fault diagnosis is not ideal. In practice, collecting multi-source data from different scenarios can provide richer generalization knowledge, and fine-grained information matching of relevant subdomains can achieve more accurate knowledge transfer, which is conducive to the improvement of the cross-domain fault diagnosis performance. To this end, a multi-source subdomain adaptation transfer learning method is proposed to transfer diagnostic knowledge from multiple sources for cross-domain fault diagnosis. This approach exploits a multi-branch network structure to match the feature spatial distributions of each source and target domain separately, where the local maximum mean discrepancy is used for fine-grained local alignment of subdomain distributions within the same category of different domains. Moreover, the weighted score of a source-specific is obtained according to its distribution distance, and multiple source classifiers are combined with the corresponding weighted scores for the joint diagnosis of the device status. Extensive experiments are conducted on three rotating machinery datasets to verify the effectiveness of the proposed model for cross-domain fault diagnosis.

Dual Regularized Unsupervised Feature Selection Based on Matrix Factorization and Minimum Redundancy with application in gene selection

Abstract: Gene expression data have become increasingly important in machine learning and computational biology over the past few years. In the field of gene expression analysis, several matrix factorization-based dimensionality reduction methods have been developed. However, such methods can still be improved in terms of efficiency and reliability. In this paper, an innovative approach to feature selection, called Dual Regularized Unsupervised Feature Selection Based on Matrix Factorization and Minimum Redundancy (DR-FS-MFMR), is introduced. The major focus of DR-FS-MFMR is to discard redundant features from the set of original features. In order to reach this target, the primary feature selection problem is defined in terms of two aspects: (1) the matrix factorization of data matrix in terms of the feature weight matrix and the representation matrix, and (2) the correlation information related to the selected features set. Then, the objective function is enriched by employing two data representation characteristics along with an inner product regularization criterion to perform both the redundancy minimization process and the sparsity task more precisely. To demonstrate the proficiency of the DR-FS-MFMR method, a large number of experimental studies are conducted on nine gene expression datasets. The obtained computational results indicate the efficiency and productivity of DR-FS-MFMR for the gene selection task.

Coronavirus herd immunity optimizer with greedy crossover for feature selection in medical diagnosis

Abstract: The importance of medical data and the crucial nature of the decisions that are based on such data, as well as the large increase in its volume, has encouraged researchers to develop feature selection (FS)-based approaches to identify the most relevant data for specific medical problems In this paper, two intelligent wrapper FS approaches based on a new metaheuristic algorithm named the coronavirus herd immunity optimizer (CHIO) were applied with and without the incorporation of a greedy crossover (GC) operator strategy to enhance exploration of the search space by CHIO. The two proposed approaches, CHIO and CHIO-GC, were evaluated using 23 medical benchmark datasets and a real-world COVID-19 dataset. The experimental results indicated that CHIO-GC outperformed CHIO in terms of search capability, as reflected in classification accuracy, selection size, F-measure, standard deviation and convergence speed. The GC operator was able to enhance the balance between exploration and exploitation of the CHIO in the search and correct suboptimal solutions for faster convergence. The proposed CHIO-GC was also compared with two previous wrapper FS approaches, namely, binary moth flame optimization with Lévy flight (LBMFO_V3) and the hyper learning binary dragonfly algorithm (HLBDA), as well as four filter methods namely, Chi-square, Relief, correlation-based feature selection and information gain. CHIO-GC surpassed LBMFO_V3 and the four filter methods with an accuracy rate of 0.79 on 23 medical benchmark datasets. CHIO-GC also surpassed HLBDA with an accuracy rate of 0.93 when applied to the COVID-19 dataset. These encouraging results were obtained by striking a sufficient balance between the two search phases of CHIO-GC during the hunt for correct solutions, which also increased the convergence rate. This was accomplished by integrating a greedy crossover technique into the CHIO algorithm to remedy the inferior solutions found during premature convergence and while locked into a local optimum search space.

Coronavirus herd immunity optimizer with greedy crossover for feature selection in medical diagnosis

Abstract: The importance of medical data and the crucial nature of the decisions that are based on such data, as well as the large increase in its volume, has encouraged researchers to develop feature selection (FS)-based approaches to identify the most relevant data for specific medical problems In this paper, two intelligent wrapper FS approaches based on a new metaheuristic algorithm named the coronavirus herd immunity optimizer (CHIO) were applied with and without the incorporation of a greedy crossover (GC) operator strategy to enhance exploration of the search space by CHIO. The two proposed approaches, CHIO and CHIO-GC, were evaluated using 23 medical benchmark datasets and a real-world COVID-19 dataset. The experimental results indicated that CHIO-GC outperformed CHIO in terms of search capability, as reflected in classification accuracy, selection size, F-measure, standard deviation and convergence speed. The GC operator was able to enhance the balance between exploration and exploitation of the CHIO in the search and correct suboptimal solutions for faster convergence. The proposed CHIO-GC was also compared with two previous wrapper FS approaches, namely, binary moth flame optimization with Levy flight (LBMFO_V3) and the hyper learning binary dragonfly algorithm (HLBDA), as well as four filter methods namely, Chi-square, Relief, correlation-based feature selection and information gain. CHIO-GC surpassed LBMFO_V3 and the four filter methods with an accuracy rate of 0.79 on 23 medical benchmark datasets. CHIO-GC also surpassed HLBDA with an accuracy rate of 0.93 when applied to the COVID-19 dataset. These encouraging results were obtained by striking a sufficient balance between the two search phases of CHIO-GC during the hunt for correct solutions, which also increased the convergence rate. This was accomplished by integrating a greedy crossover technique into the CHIO algorithm to remedy the inferior solutions found during premature convergence and while locked into a local optimum search space.

ConvUNeXt: An efficient convolution neural network for medical image segmentation

Abstract: Recently, ConvNeXts constructing from standard ConvNet modules has produced competitive performance in various image applications. In this paper, an efficient model based on the classical UNet, which can achieve promising results with a low number of parameters, is proposed for medical image segmentation. Inspired by ConvNeXt, the designed model is called ConvUNeXt and towards reduction in the amount of parameters while retaining outstanding segmentation superiority. Specifically, we firstly improved the convolution block of UNet by using large convolution kernels and depth-wise separable convolution to considerably decrease the number of parameters; then residual connections in both encoder and decoder are added and pooling is abandoned via adopting convolution for down-sampling; during skip connection, a lightweight attention mechanism is designed to filter out noise in low-level semantic information and suppress irrelevant features, so that the network can pay more attention to the target area. Compared to the standard UNet, our model has 20% fewer parameters, meanwhile, experimental results on different datasets show that it exhibits superior segmentation performance when the amount of data is scarce or sufficient. Code will be available at https://github.com/1914669687/ConvUNeXt.

From Concept Drift to Model Degradation: An Overview on Performance-Aware Drift Detectors

Abstract: The dynamicity of real-world systems poses a significant challenge to deployed predictive machine learning (ML) models. Changes in the system on which the ML model has been trained may lead to performance degradation during the system's life cycle. Recent advances that study non-stationary environments have mainly focused on identifying and addressing such changes caused by a phenomenon called concept drift. Different terms have been used in the literature to refer to the same type of concept drift and the same term for various types. This lack of unified terminology is set out to create confusion on distinguishing between different concept drift variants. In this paper, we start by grouping concept drift types by their mathematical definitions and survey the different terms used in the literature to build a consolidated taxonomy of the field. We also review and classify performance-based concept drift detection methods proposed in the last decade. These methods utilize the predictive model's performance degradation to signal substantial changes in the systems. The classification is outlined in a hierarchical diagram to provide an orderly navigation between the methods. We present a comprehensive analysis of the main attributes and strategies for tracking and evaluating the model's performance in the predictive system. The paper concludes by discussing open research challenges and possible research directions.

Meta-features for meta-learning

Abstract: Meta-learning is increasingly used to support the recommendation of machine learning algorithms and their configurations. These recommendations are made based on meta-data, consisting of performance evaluations of algorithms and characterizations on prior datasets. These characterizations, also called meta-features, describe properties of the data which are predictive for the performance of machine learning algorithms trained on them. Unfortunately, despite being used in many studies, meta-features are not uniformly described, organized and computed, making many empirical studies irreproducible and hard to compare. This paper aims to deal with this by systematizing and standardizing data characterization measures for classification datasets used in meta-learning. Moreover, it presents an extensive list of meta-features and characterization tools, which can be used as a guide for new practitioners. By identifying particularities and subtle issues related to the characterization measures, this survey points out possible future directions that the development of meta-features for meta-learning can assume.

Optimized radio-frequency identification system for different warehouse shapes

Abstract: Advanced technologies are receiving increased attention in every sector of industry. One of such technologies is radio-frequency identification. In some business models, a manufacturer’s responsibility is to correctly manage inventories for supply chain while maximizing profit for every member. On the other hand, the issue of supply chain member unreliability persists in the real situation. The manufacturer uses radio-frequency identification system in order to better control inventory, which then may manage inventory pooling and labor investment. An efficient number of a radio-frequency identification readers are required for detecting radio-frequency identification tags. However, radio-frequency identification system is different for different cases, especially when there are different layouts even in the same system. If radio-frequency identification technology is not deployed correctly according to the warehouse specifics, the issue of optimized benefits may arise. This study explores how radio-frequency identification technology can be used to increase profit of supply chain members in an uncertain situation like unreliability within the supply chain management. In terms of deployment, the optimal gap between the readers while having optimal number of readers for most common warehouse shapes is investigated in this study. Because the life of product sometimes is unpredictable, it is possible that no specified form of distribution function will be followed. To identify the global optimal solution, the Kuhn–Tucker approach is utilized. The numerical study shows that the manufacturer may gain more profit by implementing revenue sharing and the optimal spacing between readers while optimizing radio-frequency identification system costs for different layouts.

An enhanced guided LDA model augmented with BERT based semantic strength for aspect term extraction in sentiment analysis

Abstract: Aspect level sentiment analysis is a fine-grained task in sentiment analysis. It extracts aspects and their corresponding sentiment polarity from opinionated text. The first subtask of identifying the opinionated aspects is called aspect extraction, which is the focus of the work. Social media platforms are an enormous resource of unlabeled data. However, data annotation for fine-grained tasks is quite expensive and laborious. Hence unsupervised models would be highly appreciated. The proposed model is an unsupervised approach for aspect term extraction, a guided Latent Dirichlet Allocation (LDA) model that uses minimal aspect seed words from each aspect category to guide the model in identifying the hidden topics of interest to the user. The guided LDA model is enhanced by guiding inputs using regular expressions based on linguistic rules. The model is further enhanced by multiple pruning strategies, including a BERT based semantic filter, which incorporates semantics to strengthen situations where co-occurrence statistics might fail to serve as a differentiator. The thresholds for these semantic filters have been estimated using Particle Swarm Optimization strategy. The proposed model is expected to overcome the disadvantage of basic LDA models that fail to differentiate the overlapping topics that represent each aspect category. The work has been evaluated on the restaurant domain of SemEval 2014, 2015 and 2016 datasets and has reported an F-measure of 0.81, 0.74 and 0.75 respectively, which is competitive in comparison to the state of art unsupervised baselines and appreciable even with respect to the supervised baselines.

Deep embedded hybrid CNN-LSTM network for lane detection on NVIDIA Jetson Xavier NX

Abstract: In recent years, lane detection has become one of the most important factors in the progress of intelligent vehicles. To deal with the challenging problem of low detection precision and real-time performance of most traditional systems, we proposed a real-time deep lane detection system based on CNN Encoder–Decoder and Long Short-Term Memory (LSTM) networks for dynamic environments and complex road conditions. The CNN Encoder network is used to extract deep features from a dataset and to reduce their dimensionality. A corresponding decoder network is used to map the low resolution encoder feature maps to dense feature maps that correspond to road lane. The LSTM network processes historical data to improve the detection rate through the removal of the influence of false alarm patches on detection results. We propose three network architectures to predict the road lane: CNN Encoder–Decoder network, CNN Encoder–Decoder network with the application of Dropout layers and CNN Encoder–LSTM-Decoder network that are trained and tested on a public dataset comprising 12764 road images under different conditions. Experimental results show that the proposed hybrid CNN Encoder–LSTM-Decoder network that we have integrated into a Lane-Departure-Warning-System (LDWS) achieves high prediction performance namely an average accuracy of 96.36%, a Recall of 97.54%, and a F1-score of 97.42%. A NVIDIA Jetson Xavier NX supercomputer has been used, for its performance and efficiency, to realize an Embedded Deep LDWS.

Graph transformer network with temporal kernel attention for skeleton-based action recognition

Abstract: Skeleton-based human action recognition has caused wide concern, as skeleton data can robustly adapt to dynamic circumstances such as camera view changes and background interference thus allowing recognition methods to focus on robust features. In recent studies, the human body is modeled as a topological graph, and the graph convolution network (GCN) is used to extract features of actions. Although GCN has a strong ability to learn spatial modes, it ignores the varying degrees of higher-order dependencies that are captured by message passing. Moreover, the joints represented by vertices are interdependent, and hence incorporating an attention mechanism to weigh dependencies is beneficial. In this work, we propose a kernel attention adaptive graph transformer network (KA-AGTN), which models the higher-order spatial dependencies between joints by the graph transformer operator based on multihead self-attention. In addition, the Temporal Kernel Attention (TKA) block in KA-AGTN generates a channel-level attention score using temporal features, which can enhance temporal motion correlation. After combining the two-stream framework and adaptive graph strategy, KA-AGTN outperforms the baseline 2s-AGCN by 1.9% and by 1% under X-Sub and X-View on the NTU-RGBD 60 dataset, by 3.2% and 3.1% under X-Sub and X-Set on the NTU-RGBD 120 dataset, and by 2% and 2.3% under Top-1 and Top-5 and achieves the state-of-the-art performance on the Kinetics-Skeleton 400 dataset.

Multi-objective scheduling of IoT-enabled smart homes for energy management based on Arithmetic Optimization Algorithm: A Node-RED and NodeMCU module-based technique

Abstract: The home energy management system (HEMS) based on advanced internet of things (IoT) technology has attracted the special attention of engineers in the field of smart grid (SG), which has the task of the demand side management (DSM) and helps to control the equality between demand and electricity supply. The main performance of HEMS is based on the optimal scheduling of home appliances because it manages power consumption by automatically controlling the loads and transferring them from peak hours to off-peak hours. This paper presents a multi-objective version of a newly introduced metaheuristic, called Arithmetic Optimization Algorithm (AOA) to discover optimal scheduling of the home appliances, which is called Multi-Objective Arithmetic Optimization Algorithm (MOAOA). Furthermore, the HEMS architecture has been programmed based on the Raspberry Pi minicomputer with Node-RED and NodeMCU modules. HEMS uses the MOAOA algorithm to find the optimal schedule pattern to reduce daily electricity costs, reduce the peak to average ratio (PAR), and increase user comfort (UC). Real-time pricing (RTP) and critical peak pricing (CPP) signals are presumed as energy tariffs. Simulations are performed in two different scenarios: (I) appliance scheduling scheme and (II) appliance scheduling scheme with the integration of renewable energy sources (RES). The results of MOAOA are compared with Multi-Objective Particle Swarm Optimization (MOPSO), Multi-Objective Gray Wolf Optimizer (MOGWO), and Multi-Objective Antlion optimization (MOALO) algorithms. The results demonstrate that the use of the presented scheme remarkably reduces the cost of electricity consumption as well as PAR, in addition to the integration of MOAOA with RES, which greatly increases user comfort.