Search

Abstract: Two players of Rock-Paper-Scissors are modeled as adaptive agents which use a reinforcement learning algorithm and exhibit chaotic behavior in terms of trajectories of probability in mixed strategies space. This paper demonstrates that an external super-agent can exploit the behavior of the other players to predict favorable moments to play against one of the other players the symbol suggested by a sub-optimal strategy. This third agent does not affect the learning process of the other two players, whose only goal is to beat each other. The choice of the best moment to play is based on a threshold associated with the Local Lyapunov Exponent or the Entropy, each computed by using the time series of symbols played by one of the other players. A method for automatically adapting such a threshold is presented and evaluated. The results show that these techniques can be used effectively by a super-agent in a game involving adaptive agents that exhibit collective chaotic behavior. [Copyright &y& Elsevier]

The quest for higher software quality remains a paramount concern in software testing, prompting a shift towards leveraging machine learning techniques for enhanced testing efficacy. The objective of this paper is to identify , categorize , and systematically compare the present studies on software testing utilizing machine learning methods. This study conducts a systematic literature review (SLR) of 40 pertinent studies spanning from 2018 to March 2024 to comprehensively analyze and classify machine learning methods in software testing. The review encompasses supervised learning, unsupervised learning, reinforcement learning, and hybrid learning approaches. The strengths and weaknesses of each reviewed paper are dissected in this study. This paper also provides an in-depth analysis of the merits of machine learning methods in the context of software testing and addresses current unresolved issues. Potential areas for future research have been discussed, and statistics of each review paper have been collected. By addressing these aspects, this study contributes to advancing the discourse on machine learning's role in software testing and paves the way for substantial improvements in testing efficacy and software quality. • A comprehensive systematic review on machine learning methods in software testing is provided. • The main ideas, methods, tools, merits, demerits, evaluation metrics, and evaluation methods are discussed. • A scientific taxonomy of machine learning methods in software testing is presented. • A detailed list of challenges, open issues, and future research directions is outlined. [ABSTRACT FROM AUTHOR]

The paper describes three lossy data compression techniques based on the principal component analysis (PCA), which are compared using the image compression task. The presented approach uses both classical PCA method based on the eigen-decomposition of the image data covariance matrix and two different neural network structures. The first neural structure is a two-layer feed-forward network with supervised learning acting as the so-called autoencoder, and the second one is a single-layered network with an unsupervised learning method commonly known as the generalized Hebbian algorithm. For each compression method mentioned above, the influence of the number of image segments (frames) and the number of eigenvalues/eigenvectors on the compression ratio and the image quality are examined using three different gray-scale test images. The paper also addresses a vital implementation issue regarding selecting appropriate data types to represent the compressed data. The paper's main conclusion is that the classical PCA method outperforms its neural counterparts, both in terms of the decompressed image quality and the time required to perform the compression procedure. The positive aspect of using neural networks as a tool for PCA-based lossy data compression is that they do not require calculating the correlation matrix explicitly and thus can be used in online data acquisition schemes. [Display omitted] • Classical PCA compression outweighs neural nets in image quality and operation time. • Appropriate learning algorithm is essential for the neural PCA compression. • Neural PCA compression can have an advantage in online data acquisition schemes. • Appropriate data types are important for compression ratio and image quality. [ABSTRACT FROM AUTHOR]

This paper proposes a novel algorithm for incremental learning over streaming data in a non-stationary environment. The idea refers to the applicability of Probabilistic Neural Networks (PNNs), commonly used as a fast and robust method for solving classification problems in a stationary scenario. It is well known that PNNs have solid mathematical foundations but they fail in the non-stationary scenario and are not able to deal with concept drift. In this paper, the method based on the orthogonal series expansions of unknown probability densities is proposed. This approach significantly extends the applicability of the PNNs. The nonparametric procedures called Incremental Probabilistic Neural Networks (IPNNs) for tracking drifting probability densities or conditional class densities, in the case of classification, are presented. We prove their convergence as the stream data size is growing to infinity. More precisely, we show convergence in probability and with probability one as the sample size tends to infinity. Compared to the existing literature, almost exclusively based on various heuristics, the proposed method is mathematically justified and has solid theoretical foundations. Simulations performed on synthetic and air pollution data confirm the effectiveness of the algorithm. • The incremental version of the Probabilistic Neural Network (IPNN) based on the orthogonal series, able to work in non-stationary environments. • Mathematical proofs of the convergence of the proposed estimators for different types of concept drift. • Extension to the classification task, using estimated density functions as discriminant functions. • Application of the proposed IPNN to air pollution monitoring. [ABSTRACT FROM AUTHOR]

As an important performance indicator of battery management systems, lithium-ion battery state of health (SOH) information is crucial to ensure battery safety and extend battery lifetime. Aiming at the problems of feature extraction difficulty, low accuracy of long-term prediction, and poor parallel computing capability of general data-driven methods, this paper proposes a SOH estimation method for lithium-ion batteries based on incremental capacity analysis (ICA) and Transformer. First, the original incremental capacity (IC) curve of the battery is extracted based on the ICA method, and the original IC curve is processed using the dual filtering method of moving average smoothing filter plus Gaussian smoothing filter, which in turn extracts the peak features of the curve. Then, the Transformer network model based on the multi-head attention mechanism is built. Finally, the extracted peak features of the IC curve are used as model inputs, and the Transformer model is utilized to realize the SOH estimation of lithium-ion batteries. In this paper, experiments based on different input features, prediction starting points, and ambient temperatures are conducted using experimental data of lithium-ion batteries from three sources and analyzed in comparison with commonly used machine learning methods. The experimental results show that the SOH estimation method proposed in this paper has higher long-term prediction accuracy and better temperature adaptability than commonly used machine learning methods. • The dual filtering method of MASF plus GSF is used to smooth the IC curve. • Transformer model is proposed to estimate the SOH of lithium-ion batteries. • The peak features of the IC curves are used as model inputs. • The advantages of the proposed method were verified by several experiments. [ABSTRACT FROM AUTHOR]

The spread of COVID-19 has caused a great series of negative effects on countries around the world. To curb the pandemic, many governments apply a variety of strict measures including the closure of public places, transportation systems, etc. This paper proposes an algorithm that searches through a set of government policies and selects the policies that minimize the growth rate of the pandemic with minimum cost to society. The proposed algorithm first builds a surrogate model of the pandemic and then searches through the set of government policies via an evolutionary algorithm. The surrogate model in this paper is an ensemble of several learning algorithms that uses a meta-learning algorithm that estimates where in the feature space each of the learning algorithms performs better and then gives a higher weight to the learning algorithm in the voting process. Because using the surrogate models as the fitness function induces uncertainty, in this paper we use an uncertainty reduction mechanism to perform a search through the search space better. Data from 124 countries are used in this paper which contains information about the policies each government has taken since the beginning of the pandemic and 75 environmental factors like climate, religion, politics, economy, etc. Several experimental studies are performed on the output of the modeling system and the optimization algorithms and comparisons are performed with state-of-the-art methods. • A novel ensemble learning algorithm is proposed to build a model of the COVID-19 pandemic. • A list of factors that may affect the behavior of the pandemic are studied. • The best set of features for the modeling algorithm is found. • An evolutionary algorithm is proposed that optimizes the set of government policies. • An uncertainty reduction is proposed to remove approximation uncertainty. [ABSTRACT FROM AUTHOR]

Research on time series online modeling has received much attention in recent years. The existing advanced methods are often accompanied by the problem of low model established efficiency, so it is difficult to realize online modeling in practical applications. This paper proposed a sequence decomposition learning networks (SDLN) to solve time series online forecasting problems. In contrast to most time series forecasting methods, SDLN is a novel time series online forecasting method without capturing sliding time window features. The SDLN consists of five layers of neurons, namely, the input layer, decomposition layer, sparse layer, hidden layer, and output layer, where the additions of the decomposition layer and sparse layer enable a fast model calculation speed and do not require the capture of sliding time window features. In addition, it can establish time series forecasting models with better model accuracy and model stability in a very short model calculation time. To verify the effectiveness of SDLN, eight publicly available time series datasets are applied. Experimental results show that SDLN can obtain better model performance than other state-of-the-art methods, especially in terms of model computation speed. It can achieve a model accuracy of 10-4 to 10-2 and a model computation time of 10-2 seconds for short- and medium-term forecasting on all eight datasets. Therefore, SDLN is an effective time series online modeling method. • For solving the problem of time series online forecasting, a kind of novel model that does not need to continuously capture specific time window features, call Sequence Decomposition Learning Networks (SDLN) is proposed in this paper. It is applied to eight publicly available time series datasets. The experimental results show that it not only has a forecasting accuracy of 10-2 to 10-4, but also has a forecasting time of 10-2s. [ABSTRACT FROM AUTHOR]

Recently, deep learning has proven to be a successful technique especially in medical image analysis. This paper aims to highlight the importance of deep learning architectures in lung disease diagnosis using CXR images. Related articles were identified through searches of electronic resources, including IEEE, Springer, Elsevier, PubMed, Nature and, Hindawi digital library. The inclusion of articles was based on high-performance artificial intelligence models, developed for the classification of possible findings in CXR images published from 2018 to 2023. After the quality assessment of papers, 129 articles were included according to PRISMA guidelines. Papers were studied by types of lung disease, data source, algorithm type, and outcome metrics. Three main categories of computer-aided lung disease detection were covered: traditional machine learning, deep learning-based methods, and combination of aforementioned methods for all lung diseases. The results showed that various pre-trained networks including ResNet, VGG, and DenseNet, are the most frequently used CNN architectures and would result in a notable increase in sensitivity and accuracy. Recent research suggests that utilizing a combination of deep networks with a robust machine learning classifier can outperform deep learning approaches that rely solely on fully connected neural networks as their classifier. Finally, the limitations of the existing literature and potential future research opportunities in possible findings in CXR images using deep learning architectures are discussed in this systematic review. • A systematic review based on PRISMA framework on machine learning applications to analyze chest X-ray images. • Trends of architectures used in last 5 years. • Preprocessing and processing techniques used by the researchers. • Future trends and open problems. [ABSTRACT FROM AUTHOR]

Alzheimer's is a dangerous disease prevalent in human societies, and unfortunately, its incidence is increasing daily. The number of patients is on the rise, while the availability of physical doctors has become limited and their schedules are packed. Consequently, the adoption of digital healthcare systems for Alzheimer's disease (AD) has become more common, aiming to alleviate the burden on both AD patients and doctors. AD digital healthcare is a highly complex domain that incorporates various technologies, including fog computing, cloud computing, and deep learning algorithms. However, the implementation of these fog, cloud, and deep learning technologies has encountered challenges related to high computational time during AD detection processes. To address these challenges, this paper focuses on the convex optimization problem, which aims to optimize computation time and accuracy constraints in digital healthcare applications for AD. Convex optimization necessitates the use of an evolutionary algorithm that can enhance different AD constraints in distinct phases. The paper introduces a novel scheme called Evolutionary Deep Convolutional Neural Network Scheme (EDCNNS), designed to minimize computation time and achieve the highest prediction accuracy criteria for AD. EDCNNS comprises several phases, including feature extraction, selection, execution, and scheduling on the fog cloud nodes. The simulation results demonstrate that EDCNNS optimized security by 38%, reduced the deadline failure ratio by 29%, and improved selection accuracy by 50% across different Alzheimer's classes compared to existing studies. • An evolutionary deep neural network scheme (EDNNS) is proposed. • The EDNNS is applied to detect Alzheimer's disease using different modalities. • Federated deep learning is used to minimize computational load. • Evolutionary algorithm is used for local search and global search. [ABSTRACT FROM AUTHOR]

This paper presents a new training algorithm that can update the situation of layers' network, and therefore, connections, neurons, and firing rate of neurons based on FORCE (first-order reduced and controlled error) training algorithm. The Quantized Full FORCE algorithm (QFF) also updates the number of neurons and connections between different layers in the network per iteration in a way that the whole firing rate of each layer is updated via selecting the best neurons and combining strong features. The update method is sequential, so that with each instance passing through the network, the network structure is updated with the Full FORCE algorithm. The algorithm updates the structure of networks with a multiple/single middle layer of the supervised version of feed forward networks such as Multilayer perceptron (MLP), changing them into partially-connected networks. A combination of principal component analysis PCA and Linear Discriminant Analysis (LDA) algorithms has been used to cluster the network input features. The paper focuses on the deep supervised MLP network with backpropagation (BP) and various datasets and its comparison with other MLP based stat of art methods and hybrid evolutionary algorithms. We achieved 98.15 percent accuracy for facial expression 98.6 and 97.7 percent for Wisconsin breast Cancer and Iris Flower in respectively. The training algorithm employed in the study enjoys a lower computational complexity while yielding faster and more accurate convergence, starting with a very low level of errors of 0.009 in comparison with the full connection network and it solves the challenge of getting stuck in local minima and poor convergence of Gradient Decent with BP. • Providing the new training algorithm to update the connections and firing rate of neurons in each iteration. • The proposed training model is based on Quantized Full FORCE (first-order reduced and controlled error) training. • Providing an accurate framework to update the number of neurons, converting the network to a partial connection network. • In the new training model, the best features have the highest number of connections and neurons within the middle layers. • Being the decision speed of the network in the proposed method much faster than a fully connected network. • Cost-effectiveness and much lower resource requirements for implementation of deep networks on hardware such as FPGAs. [ABSTRACT FROM AUTHOR]

This paper presents the concept of the computationally-efficient reduced order control performance assessment (RO-CPA) system for automatic detection of industrial single closed loop control systems whose performance could be not acceptable due to a poor tuning of the controller. The proposed machine learning (ML) based classification system is derived for a broad class of industrial control closed loops based on the proportional–integral–derivative (PID) controller. Apart from a high accuracy of the classification of the closed loop performance, the proposed RO-CPA system is derived to meet the requirements of low computational complexity and low memory usage. These requirements allow for practical implementation of RO-CPA in devices with low computational and memory resources, such as industrial programmable logic controllers (PLC) operating in manufacturing systems or embedded autonomous systems with dedicated PID-based control loops. This paper shows the rigorous ML-based design of the proposed RO-CPA system and the results of the validation of its classification accuracy. Additionally, an example of a practical implementation within a PLC in the form of a general purpose library function block is presented and the final experimental validation is made using a part of laboratory heat exchange and distribution setup. • Computationally efficient approach to closed loop control performance assessment. • Machine-learning based derivation of control performance assessment system. • High classification accuracy demonstrated by simulation. • Practical implementation in Programmable Logic Controller. • Successful validation based on experimental heating system. [ABSTRACT FROM AUTHOR]

A perishable product has a limited shelf life, and inefficient management often leads to waste. This paper focuses on dynamic pricing and inventory management strategies for perishable products. By implementing effective inventory control and the right pricing policy, it is possible to maximize expected revenue. However, the exponential growth of the problem size due to the shelf life of the products makes it impractical to use methods that guarantee optimal solutions, such as Dynamic Programming (DP). Therefore, approximate solution algorithms become necessary. We use Deep Reinforcement Learning (DRL) algorithms to address the dynamic pricing and ordering problem for perishable products, considering price and age-dependent stochastic demand. We investigate Deep Q Learning (DQL) solutions for discrete action spaces and Soft Actor-Critic (SAC) solutions for continuous action spaces. To mitigate the negative impact of the stochastic environment inherent in the problem, we propose two different DQL approaches. Our results show that the proposed DQL and SAC algorithms effectively address inventory control and dynamic pricing for perishable products, even when products of different ages are offered simultaneously. Compared to dynamic programming, our proposed DQL approaches achieve an average approximation of 95.5 % and 96.6 %, and reduce solution times by 71.5 % and 79.9 %, respectively, for the largest problem. In addition, the SAC algorithm achieves on average 4.6 % and 1.7 % better results and completes the task 56.1 % and 48.2 % faster than the proposed DQL algorithms. • Deep Reinforcement Learning for dynamic system control. • Dynamic pricing and inventory control of perishable products. • Optimization with intelligent agents in large sized state and action spaces. • Tackling random data-driven stochastic environment with robust algorithms. [ABSTRACT FROM AUTHOR]

This paper proposes a new Bayesian inference-based learning automaton with a multi-objective brainstorm algorithm to find the best solutions for combinatorial optimization problems such as vehicle routing problems. Initially, a robust deep k-means algorithm is used to solve the consumer allocation problem to various clusters. Then, a multi-objective brainstorm optimization algorithm is used to determine non-dominated solutions and thereby, the Pareto front of the solution space can be formed. The proposed method uses an external archive to store the attained non-dominated solutions. In addition, a reinforcement learning algorithm is used for improving the overall convergence of the brainstorm optimization algorithm, and it can be used to discover good designs for the parameters of the brainstorm algorithm and ensure the best transition decisions by balancing the exploration and exploitation phases. Here, the control of two phases can be handled by Q-table in reinforcement learning. Subsequently, the Bayesian inference-based learning automaton is employed to learn the optimal actions by placing the agent in a complex environment. The proposed method is implemented in the Python working platform, and the performance of the proposed approach is validated based on Solomon's benchmark problems. Additionally, the effectiveness of the proposed method can be validated by comparing it with other algorithms in terms of Pareto optimal results, hypervolume, generational distance, inverted generation distance, space and spread. As a result, the proposed method has obtained better generational distance and inverted generation distance of 0.0669 and 0.0004 than other competitive methods, and it is superior for solving vehicle routing problems. ● Integrate the machine learning techniques into meta-heuristics for solving CVRP with Time Windows. ● Formulate a MOP for Vehicle Routing based on the constraints of total route distance and total customer waiting time. ● Modify the CRLA using learning automaton to improve the overall meta-heuristics convergence by learning the actions optimally. ● Validate the performance of the proposed algorithm on VRP benchmarks. [ABSTRACT FROM AUTHOR]

Financial time series forecasting is crucial in empowering investors to make well-informed decisions, manage risks effectively, and strategically plan their investment activities. However, the non-stationary and non-linear characteristics inherent in time series data pose significant challenges when accurately predicting future forecasts. This paper proposes a novel Recurrent ensemble deep Random Vector Functional Link (RedRVFL) network for financial time series forecasting. The proposed model leverages randomly initialized and fixed weights for the recurrent hidden layers, ensuring stability during training. Furthermore, incorporating stacked hidden layers enables deep representation learning, facilitating the extraction of complex patterns from the data. The proposed model generates the forecast by combining the outputs of each layer through an ensemble approach. A comparative analysis was conducted against several state-of-the-art models over financial time-series datasets, and the results demonstrated the superior performance of our proposed model in terms of forecasting accuracy and predictive capability. [ABSTRACT FROM AUTHOR]

Interpretation of the prediction is vital for mission critical tasks. Accurate interpretation relies upon the generalization accuracy of the model. In this paper, we propose a modified gradient descent learning algorithm to improve the generalization ability of a Spiking Neural Network with time-varying weights (SNN-t). This algorithm is referred to as GradST, can help towards improving the interpretation of multiclass classification problems. We have transformed the SNN-t to a Generalized Additive Model (GAM) to provide interpretation. The resultant Spiking Additive Model (SAM) has the generalization ability of SNN-t and the interpretable characteristics of GAM for multiclass problems. We also propose a post-processing method to enable better visualization of multiple shape functions of GAMs, towards better relative interpretation for multiclass classification problems. The post-processing method utilizes the properties of multiclass GAMs to visually modify the shape functions to establish the importance of the feature in multiclass setting. We first evaluate the performance of SNN-t, trained with GradST and the SAM generated from it, on large public datasets. The SNN-t trained with GradST has better generalization accuracy than other SNN-t classifier and consequently, the SAM generated from it has better generalization accuracy than other state-of-the-art multiclass GAMs. Improved accuracy in SAM implies more reliable interpretation. Then, we evaluate the proposed post-processing method for multiclass GAMs to provide relative interpretation. It is observed that relative interpretation of multiclass GAM is more meaningful and reliable. • Modified gradient descent-based learning algorithm for a spiking neural classifier. • Transformation of spiking neural classifier to an interpretable classifier. • Improve generalization ability for reliable interpretation. • Post-processing method for relative interpretation. [ABSTRACT FROM AUTHOR]

Compared to white-box adversarial attacks, black-box adversarial attacks are more applicable in practical scenarios and have received significant attention. However, most existing black-box attacks are optimized at the output layer, and the generated adversarial samples are difficult to transfer from the surrogate model to the target model. Existing adversarial attacks indiscriminately perturb features in intermediate layers, which are prone to fall into local optima of surrogate models and have limited transferability. In this paper, we propose a transferable targeted adversarial attack framework based on a feature-aware triplet, which achieves a better tradeoff between attack ability and transferability by perturbing salient features and constructing representative sample features. The importance-weighted optimization objective interferes with the salient object-aware features of images in a targeted manner and guides adversarial perturbations in the optimization process to pull the intermediate features of samples toward a target class while pushing them far from a source class to achieve targeted transferable attacks. Moreover, a construction method of a feature library with the weighted average of feature importance is built to obtain more expressive intermediate features of the target and source classes. The modified target features and source features are fed into the triplets to guide the optimization objectives to find more transferable adversarial samples. Extensive experiments on the ImageNet-compatible dataset verify the effectiveness of the proposed method, e.g., improving the untargeted success rate by 1.8% and the targeted success rate by 2.3% against normally trained models as compared to the existing methods. • An attack framework based on a feature-aware triplet is proposed. • Achieve a better tradeoff between attack capability and transferability. • A weighted average method for constructing a feature library is designed. • Define more expressive sample spaces. • Extended experiments exhibit superior transferability of the proposed FTA. [ABSTRACT FROM AUTHOR]

This paper develops a new method for lifelong learning referred to as Sparsify Dynamically Expandable Network (SDEN) via Variational Dropout, which explores a sparse model while preserving the performance. Dynamically Expandable Network (DEN) can learn a sequence of tasks via performing network retraining, network expansion by adding only the necessary neurons, and network split to effectively prevent semantic drift in an online manner. To overcome point estimation of parameters in DEN, Bayesian Compression for DEN is developed under the Bayesian framework. However, this method demands more time for model training and testing. To improve the model efficiency, we propose SDEN under the efficient sparse learning framework. We validate our SDEN in the lifelong learning scenarios with multiple frequently used benchmarks, on which it can obtain comparable classification accuracy, and less training and testing time compared with the comparison methods. Furthermore, our method can also learn a more sparse network structure which means fewer network parameters. • Comprehensively consider model performance, parameter storage space, and model training and test time. • Dynamically increase network capabilities with necessary number of neurons via Variational Dropout. • Achieve similar performance with comparative lifelong learning algorithms. • Obtain almost the minimum training and test time while considering the parameter quantity. • Learn the probability of removing a parameter or neuron from the dataset. [ABSTRACT FROM AUTHOR]

Topological data analysis (TDA) is a powerful mathematical framework that extracts valuable insights about the shape and structure of complex datasets by identifying and analyzing underlying topological features, including connected components, holes, voids, and higher-dimensional counterparts. TDA achieves this by constructing a simplicial complex from the data, comprising simple shapes like points, lines, triangles, and higher-dimensional counterparts. The Vietoris–Rips complex (VRC), a widely used method, constructs simplicial complexes by measuring the distances between data points in a metric space. While the link between Intuitionistic Fuzzy Sets (IFSs) and TDA is clear through intuitionistic fuzzy distance measures, the literature has not explored the application of TDA for extracting topological features from uncertain and vague datasets using IFSs. In this paper, we propose a novel approach that leverages Intuitionistic Fuzzy Distance Measures (IFDMs) to generate the Intuitionistic Fuzzy Vietoris–Rips Complex (IFVRC) of IFSs. Specifically, we investigate the persistence of invariant relations of topological features between IFVRCs constructed using different distances, including Atanassov's Euclidean distance (l 2 I F S), Boran and Akay's distance (D), and Liu's distance measure (d L), over artificially generated IFSs. Furthermore, we demonstrate that the generation of IFVRC can be viewed as a Natural Hierarchical Clustering Algorithm (NHCA) for IFSs. Finally, we apply the proposed IFVRC as a Natural Hierarchical Clustering Algorithm to cluster intuitionistic fuzzy datasets related to Car and Building materials. Moreover, we incorporate a relatively large dataset, the GTZAN dataset from the Kaggle datasets repository, to classify music genres based on their topological information. • Topological analysis of intuitionistic fuzzy distance measures (IFDMs) have been conducted. • Utilizing IFDMs, Intuition Fuzzy Vietoris Rips Complex (IFVRC) generated. • It is shown how to extract topological features from IFVRC. • The persistence relations of topological features between IFVRCs have been investigated. • Finally, IFVRC is treated as a natural hierarchical clustering algorithm. [ABSTRACT FROM AUTHOR]

Human-Inspired Control (HIC) holds promise for endowing machines with human-like cognition, decision-making, and adaptability. In this study, we employ a fusion of cognitive modeling, machine learning, and control theory as the foundational architecture and empirically validate its suitability for robot control within the realm of HIC. Fuzzy logic stands out as a viable approach for HIC control, wherein control rules can be devised drawing from human intuitive inspiration. Specifically, this study explores similarity inference control systems in robotics, with the objective of enhancing line-following control as an alternative to fuzzy systems. The experimental optimization results provide insights into the advantages and limitations of the similarity inference control system. Despite achieving performance comparable to that of traditional two-stage fuzzy control systems, careful consideration of noise sensitivity is paramount. While the similarity inference approach streamlines implementation and obviates the necessity for expert-designed fuzzy rules, its susceptibility to noise can compromise performance, particularly in noisy environments. These considerations are pivotal for the development of control systems aimed at mitigating noise sensitivity, enhancing task-specific performance, and ensuring the adaptability and robustness of line-following robots. To tackle this challenge, we both discuss and experimentally evaluate potential solutions and their applicability in this paper. • The study focuses on integrating cognitive modeling, machine learning, and control theory in HIC, exploring methods like fuzzy logic. • It delves into robotics' similarity inference control systems, optimizing line-following, tackling noise susceptibility, and suggesting enhancements via sensor data and machine learning. [ABSTRACT FROM AUTHOR]

Density-based clustering is a widely explored domain in the field of machine learning, with numerous methods proposed to address challenges in clustering under varying density conditions. These methods primarily categorize samples based on their local density. However, challenges arise when dealing with boundary samples. The density of certain boundary samples is typically lower, which can lead to misclassification as noise. Moreover, in situations where two clusters have similar densities and are in close proximity, accurately classifying the boundary points between them becomes a challenging task. In this paper, we introduce an enhanced approach based on k -nearest neighbors to address this challenge within density-based clustering. In our method, upon the formation of a new cluster, we identify boundary samples by examining the spatial relationships between each sample and its k -nearest neighbors, as well as their connections to the newly established cluster. Once all clusters are formed, we refine the classification of these boundary samples by adjusting or retaining their assigned labels based on their k -nearest neighbors. Experimental evaluations on synthetic and real-world datasets demonstrate the effectiveness of our proposed method. • A new density clustering method based on boundary samples verification is proposed. • The method can handle datasets with different densities and shapes. • The method only requires one integer parameter, easy to use. • Outperforms existing methods on various benchmark datasets. [ABSTRACT FROM AUTHOR]

The scope of this paper is to test the efficiency of three different neural networks in the area of cloud-to-ground/intra-cloud lightning classification using a not very large data set of electric field data. Effective neural network learning verified for a small lightning event dataset but with whole 2-seconds particular record length is an important feature of fast and efficient procedures implemented into lightning location systems among which discrimination between cloud-to-ground and intra-cloud lightning is one of the most important task. Measurement data obtained at the Lightning Observatory in Rzeszow were used for the analysis. About 100 extremely low/medium frequency bandwidth electric field of cloud-to ground lightning registrations and the same number of intra-cloud lightning events were taken into account. During the study 81% of those registrations were dedicated for learning of neural networks while the remaining 19% were put into the testing set. All data were preselected manually with consideration of the electric field variation and information from the lightning location system database. Lightning events were selected uniformly with the well-pronounced presence of their basic components and reported distance to the registration station which ensured a wide variation of expected lightning electric field signatures. The multilayer perceptron neural network, the radial basis function neural network and the convolutional neural network were tested and compared in different configurations while distinguishing lightning. The results were compared with other conventional classification approaches, such as traditional machine learning methods as well as one a little more contemporary architecture, the long short-term memory neural network. The multilayer perceptron neural network achieved the best detection accuracy overall. Presented research is an extension of lightning identification studies available so far mainly based on the convolutional neural network by the results achieved using the multilayer perceptron neural network, the radial basis function neural network, the long short-term memory neural network and several machine learning-based classification methods. • Research based on electric field data for CG and IC lightning from Rzeszow. • CG/IC lightning discrimination was verified in presence of a small dataset. • MLP, RBF, LSTM and convolutional neural networks as well as ML were verified. • 3-layer MLP neural networks showed the best performance overall. • Analysis results can be adapted for a real-time lightning location system purposes. [ABSTRACT FROM AUTHOR]

Accurate delay evolution prediction plays a pivotal role in train rescheduling decision-making for the railway network. Existing studies on delay prediction predominantly centered around predicting delays for each train in the subsequent stations (i.e., following a train-oriented perspective). Furthermore, train operations in the railway network involve different types of entities (stations, trains, etc.), making the current graph/network models with homogenous nodes (i.e., the same kind of nodes) incapable of effectively capturing the interactions between the entities. This paper develops a network-oriented model to investigate the train delay evolution on railway networks, by predicting the delays of running trains in the network after a given time interval. The proposed model combines the GraphSAGE graph neural network (GNN) and the heterogeneous graph neural network (HetGNN) architecture, thus called SAGE-Het, enabling it to capture interactions between heterogeneous nodes (i.e., different types of nodes) based on different edges (e.g., edges between trains, trains, and stations). Additionally, SAGE-Het allows for flexible inputs in contrast to conventional machine learning techniques, whose inputs must meet the consistent dimension requirement (e.g., in the form of rectangular or grid-like arrays). The performance and robustness of the suggested SAGE-Het model are assessed in experiments on the data from two sub-networks of the China railway network. The experimental results demonstrate that SAGE-Het outperforms the existing delay prediction methods and some advanced HetGNNs used for prediction tasks in other domains; SAGE-Het demonstrates excellent scalability, capable of handling various types of nodes; the predictive performances of SAGE-Het under different prediction time horizons (10/20/30 min ahead) all exhibit better performance over other baselines; the accuracies are over 90 % under the permissible 3-minute errors for the three prediction time horizons. Specifically, the impact of train interactions on delay evolution is investigated based on the flexible input characteristic of the proposed model. The results illustrate that train interactions become subtle with the increase of train headways. This finding directly contributes to decision-making in situations where conflict resolution or train-canceling actions are needed. • A novel network-oriented approach is proposed to investigate the delay evolution in the railway network. • The proposed heterogeneous graph neural network approach allows for diverse heterogeneous nodes and flexible inputs. • The proposed approach has better predictive performance than the commonly used delay prediction methods. • The importance of train interaction on delay evolution is explored based on the proposed approach. [ABSTRACT FROM AUTHOR]

Abstract Regression problems try estimating a continuous variable from a number of characteristics or predictors. Several proposals have been made for regression models based on the use of fuzzy rules; however, all these proposals make use of rule models in which the irrelevance of the input variables in relation to the variable to be approximated is not taken into account. Regression problems share with the ordinal classification the existence of an explicit relationship of order between the values of the variable to be predicted. In a recent paper, the authors have proposed an ordinal classification algorithm that takes into account the detection of the irrelevance of input variables. This algorithm extracts a set of fuzzy rules from an example set, using as the basic model a sequential covering strategy along with a genetic algorithm. In this paper, a proposal for a regression algorithm based on this ordinal classification algorithm is presented. The proposed model can be interpreted as a multiclassifier and multilevel system that learns at each stage using the knowledge gained in previous stages. Due to similarities between regression and ordinal problems as well as the use of a set of ordinal algorithms, an error interval can be returned with the regression output value. Experimental results show the good behavior of the proposal as well as the results of the error interval. Highlights • We present a new fuzzy rule learning algorithm for regression problems. • The algorithm is based on an ordinal one which uses the sequential covering strategy with a genetic algorithm as the search mechanism. • The final regression algorithm is a multiclassifier and multilevel system which learns in an incremental way by using the learnt knowledge in previous stages. • An error interval can be returned with the regression output value. • The results show a good behavior compared with the same category algorithms, so as, other category algorithms, for regression problems. [ABSTRACT FROM AUTHOR]

Abstract Industrial Internet of Things (IIoT) plays an important role in increasing productivity and efficiency in heterogeneous wireless networks. However, different domains such as industrial wireless scenarios, small cell domains and vehicular ad hoc networks (VANET) require an efficient machine learning/intelligent algorithm to process the vertical handover decision that can maintain mobile terminals (MTs) in the preferable networks for a sufficient duration of time. The preferred quality of service parameters can be differentiated from all the other MTs. Hence, in this paper, the problem with the vertical handoff (VHO) decision is articulated as the process of the Markov decision aimed to maximize the anticipated total rewards as well as to minimize the handoffs' average count. A rewards function is designed to evaluate the QoS at the point of when the connections take place, as that is where the policy decision for a stationary deterministic handoff can be established. The proposed hybrid model merges the biogeography-based optimization (BBO) with the Markov decision process (MDP). The MDP is utilized to establish the radio access technology (RAT) selection's probability that behaves as an input to the BBO process. Therefore, the BBO determines the best RAT using the described multi-point algorithm in the heterogeneous network. The numerical findings display the superiority of this paper's proposed schemes in comparison with other available algorithms. The findings shown that the MDP-BBO algorithm is able to outperform other algorithms in terms of number of handoffs, bandwidth availability, and decision delays. Our algorithm displayed better expected total rewards as well as a reduced average account of handoffs compared to current approaches. Simulation results obtained from Monte-Carlo experiments prove validity of the proposed model. Highlights • This works proposes a hybrid intelligent model for network selection in IIoT. • The proposed model merges the BBO with the MDP. • Using MDP, the radio access technology (RAT) selection's probability is established. • With the BBO, the best radio access technology (RAT) is determined. [ABSTRACT FROM AUTHOR]

Abstract Missing values are an unavoidable issue in many real-world datasets. One of the most popular approaches to classification with incomplete data is to use imputation to replace missing values with plausible values. However, powerful imputation methods are too computationally intensive when applying a classifier to a new unknown instance. This paper proposes new approaches to integrating imputation, clustering and feature selection for classification with incomplete data in order to improve efficiency without loss of accuracy. Clustering is used to reduce the number of instances used by the imputation. Feature selection is used to remove redundant and irrelevant features of training data which greatly reduces the cost of imputation. The paper also investigates the ability of Differential Evolution (DE) to search feature subsets with incomplete data. Results show that the integration of imputation, clustering and feature selection not only improves classification accuracy, but also dramatically reduces the computation time required to estimate missing values when classifying new instances. Highlights • Examine clustering and feature selection to improve classification with missing data. • Clustering speeds up imputation for classification with missing data. • Feature selection speeds up imputation and improves classification accuracy. [ABSTRACT FROM AUTHOR]

Imbalanced class distribution is a frequent and problematic issue in the domains of data engineering and machine learning. Traditional classification algorithms or machine learning models frequently fail, in this difficult situation, to provide accurate classification results for the minority class. To improve classification performance for the minority class, some research has advised creating artificial instances in the minority class using the oversampling technique. However, on the support vectors machines (SVM) model, as new minority class instances were created in an original data space, classification improvements made using these popular oversampling techniques may not hold true. In this paper, we develop a novel oversampling method termed DB-MTD-SN (for distance-based mega-trend-diffusion Siamese Network) for generating artificial minority class instances to increase the SVM model's classification accuracy for imbalanced datasets. In the proposed method, we utilize distance-based mega-trend-diffusion (DB-MTD) technique to estimate data domain for tiny support vectors to generate synthetic examples with multi-model distributions of minority class. Further, we construct a novel Siamese network model based on membership function (MF-based SN) to find the most representative synthetic minority class examples. The proposed SN model is used to map the original data onto a high-dimensional space to easily learn complicated patterns between majority and minority classes. In our proposed SN model, the MF-based contrastive loss function is used to measure the similarity between a new synthetic example and original examples to avoid generating noise. To demonstrate efficacy of the proposed DB-MTD-SN approach, ten benchmark datasets are used in this study. On two types of SVM models, we compare the proposed method with three state-of-the-art oversampling methods. Further, three evaluation metrics: G-mean, F1, and index of balanced accuracy (IBA) are utilized to measure SVM classification performance on imbalanced datasets. The experimental datasets are set at different imbalanced ratios (IRs) as 15, 20, and 30 to test classification performance using five methods. For a high IR value of 30, on the two SVM models, the proposed method achieved the best average in terms of G-mean (0.865 and 0.865), F1 (0.785 and 0.772), and IBA (0.681 and 0.689) metrics, respectively. The paired Wilcoxon test is used to evaluate whether the suggested approach has statistically significant differences from the four other methods on the three evaluation metrics. The test results demonstrate that classification results using the proposed DB-MTD-SN method enjoys significant improvements (i.e. p -value <0.05) in terms of G-mean, F1, and IBA indicators when compared to the other four methods. Our experimental results indicate the suggested DB-MTD-SN method outperforms other oversampling methods for imbalanced datasets. [Display omitted] • Imbalanced datasets pose common and important problems in many fields. • A novel distance-based mega-trend-diffusion Siamese Network (DB-MTD-SN) oversampling approach is developed in this paper. • The proposed method's efficacy is verified using several benchmark datasets. • The Wilcoxon test elucidates the significance of differences among all the methods examined. [ABSTRACT FROM AUTHOR]

Due to knowledge alienation, the application of artificial intelligence (AI) techniques in tunnel construction has been greatly stunted in recent years. In order to motivate the development of techniques in tunneling works, we propose an approach that combines automated machine learning, explainable AI, and building information modeling in this paper. Considering geotechnical uncertainties from soil properties and aleatoric uncertainties from the predicting model, this approach is used to evaluate the system risk during a large-diameter tunnel excavation taking into account the two sources of uncertainties simultaneously. The geotechnical uncertainties are evaluated by the confidence interval of the mean of samples in the Monte-Carlo simulation, and the aleatoric uncertainties are evaluated by computing the prediction interval of the machine learning models. It is found that the number of Monte-Carlo samples significantly affects the evaluation of the system's reliability. Based on our results, a generalized setup with 1000 samples performs the best and is recommended considering the efficiency of computation, as well as the reliability and conservativeness of the result. This paper effectively assists with system reliability evaluation in large-diameter tunnel constructions, where the interactive and explainable AI approach largely motivates the application of AI techniques in the field of tunnel construction. • A BIM based interactive and explainable AI approach assisted by SHAP is developed. • A probabilistic safety risk analysis with automated machine learning is proposed. • The influence from different levels of uncertainty to the design are studied. • Auto-ML is able to establish a more accurate model with much less effort paid. • A higher uncertainty level results in a more moderate design. [ABSTRACT FROM AUTHOR]

Feature selection plays an essential role in data preprocessing, which can extract valuable information from extensive data, thereby enhancing the performance of machine learning classification. However, existing feature selection methods primarily focus on selecting feature subsets without considering the impact of classifier parameters on the optimal subset. Different from these works, this paper considers jointly optimizing the feature subset and classifier parameters to minimize the number of features and achieve a low classification error rate. Since feature selection is an optimization problem with binary solution space while classifier parameters involve both continuous and discrete variables, our formulated problem becomes a complex multi-objective mixed-variable problem. To address this challenge, we consider a single-objective optimization method and a multi-objective optimization approach. Specifically, in the single-objective optimization method, we adopt the linear weight method to convert our multiple objectives into a fitness function and then propose a mixed-variable grey wolf optimizer (MGWO) to optimize the function. The proposed MGWO introduces Chaos-Faure initialization, Log convergence factor adjustment, and optimal solution adaptive update operators to enhance its adaptability and balance the global and local search of the algorithm. Subsequently, an improved multi-objective grey wolf optimizer (IMOGWO) is introduced to directly address the problem. The proposed IMOGWO introduces improved initialization, local search, and binary variable mutation operators to balance its exploration and exploitation abilities, making it more suitable for our mixed-variable problem. Extensive simulation results show that our MGWO and IMOGWO outperform recent and classic baselines. Moreover, we also find that jointly optimizing classifier parameters can significantly improve classification accuracy. • Propose a joint feature selection and classifier parameter tuning method. • Formulate a complex mixed-variable optimization problem. • Introduce a single-objective optimization method called MGWO. • Present a multi-objective optimization approach, namely IMOGWO. • Perform extensive simulations to validate the effectiveness of the proposed methods. [ABSTRACT FROM AUTHOR]

The problem of network attacks is a primary focus in the domain of intrusion detection. Models face significant challenges in recognizing intrusion behaviors, particularly when dealing with high-dimensional and sparse datasets. Traditional machine learning methods often struggle with these dimensionality issues. In contrast, deep learning, a crucial technology in intrusion detection, excels at managing high-dimensional data. However, traditional image coding methods do not adequately address data sparsification and often overlook the spatial continuity among features. The Fourier transform is a promising solution for data sparsity issues, as it effectively mitigates the impact by converting data into a different domain. Inspired by the Fourier transform, this paper proposes a lightweight intrusion detection model called TFTKD, based on Convolutional Neural Networks (CNN) and knowledge distillation. The model applies a two-dimensional Fourier transform to convert grayscale images from the time domain to the frequency domain. This transformation enhances the similarity between neighboring pixels, effectively addressing data sparsification. During the training phase, a teacher network, comprising an 8-layer CNN, is pre-trained. In the distillation phase, a one-layer CNN serves as the student network, employing Self-adaptive Temperature Knowledge Distillation to enhance the student's generalization capabilities. This approach results in a compact student network model with a constrained parameter count, demonstrating superior learning efficiency and accuracy compared to state-of-the-art methods. Experimental validation was conducted using four publicly accessible intrusion detection datasets, demonstrating the effectiveness of the proposed method. • Use a two-dimensional Fourier transform to address the problem of image sparsity. • Propose a self-adaptive temperature knowledge distillation method. • Develop a high-accuracy, lightweight intrusion detection model. [ABSTRACT FROM AUTHOR]

The extensive integration of renewable energies into power systems has led to a challenging and computationally demanding scenario for the system planning. This is due to the increased number of time series involved and the greater complexity of time series data integrated into power systems. In this paper, a new deep learning-based time aggregation method is proposed for selecting representative periods for renewable energy-integrated power system datasets where multiple variable energy resources are present. Relying on the capabilities of deep generative models, especially Generative Adversarial Network (GAN), the proposed method selects a representative period, including one or more representative days, obtained from multiple time series with spatio-temporal correlations among them. The proposed approach contains the Long Short-Term Memory (LSTM) Network in both generator and discriminator parts. Also, it includes a loss term, specifically designed for clustering tasks, in the minimax objective of the vanilla GAN loss function to enhance the clustering performance in the latent space. Furthermore, the learning rate of the proposed model, as the most important parameter of the learning algorithm, is adaptively fine-tuned during the training process, based on the training error, to enhance its learning performance. Two real-world test cases with various datasets and time series are used for data-based and model-based evaluations. Results obtained on these two real-world test cases confirm the superiority of the proposed model with respect to the state-of-the-art conventional and deep learning-based models, obtaining the performance improvement in the range of [38.55 %, 46.14 %] in terms of Mean Absolute Percentage Error (MAPE), in the range of [45.75 %, 70.16 %] in terms of Root Mean Squared Error (RMSE), and in the range of [65.81 %, 74.58 %] in terms of Mean Absolute Error (MAE). The proposed model can significantly enhance the tractability and scalability of the planning problem of renewable energy-integrated power systems, facilitating renewable energy integration into power systems which is a key issue for net zero transitioning of communities. • We propose a generative time aggregation method for selecting representative periods. • The proposed approach contains the LSTM in both generator and discriminator parts. • We propose a clustering specific loss term to enhance the clustering performance. • Multiple power time series with spatio-temporal correlations among them are used. • Extensive results on both the data-based and model-based evaluations are presented. [ABSTRACT FROM AUTHOR]

Distributed storage systems deploy data centers across multiple geographical locations to share peak loads and store massive amounts of data. By integrating edge AI technology, these systems elevate edge servers to intelligent edge nodes, deeply analyze user request patterns and data center loads, and intelligently allocate requests, thereby reducing service latency and enhancing overall service quality. However, cross-geographic distribution brings challenges in maintaining data consistency and controlling cross-node communication costs for edge AI. Additionally, a single consistency level struggles to meet diverse requirements, affecting system performance and flexibility. We explore these challenges based on a novel key–value storage system, FlexSync. Firstly, we propose a Writeless-Consistency strategy, combining machine learning to predict replica synchronization requirements, reducing redundant data transmission, and improving synchronization efficiency. Building upon this, we optimize traditional consistency algorithms and leverage Conflict-free Replicated Data Types (CRDTs) to construct a flexible multi-consistency level model, better suited for diverse application scenarios. We deployed a FlexSync cluster on the Alibaba Cloud platform and conducted comprehensive performance evaluations. Experimental results demonstrate that compared to Raft, FlexSync's strong consistency algorithm improves performance by 1.8X, while compared to Vector Clock-based algorithms, causal consistency algorithm performance enhances by 2.5X. These results validate FlexSync's efficiency and stability in geographically distributed environments. • This paper proposes a new key–value storage system named FlexSync. • Efficient sync strategy for replicas is proposed. • Introducing machine learning to optimize replication sync. • Real-world workloads are used to evaluate the performance of the proposed FlexSync. • Results demonstrated efficiency of FlexSync in terms of sync strategy. [ABSTRACT FROM AUTHOR]

This paper presents a causal discovery model for time series analysis, which can find several causal lags of one variable over another, presenting the results by means of a temporal causal graph. The proposed model is based on multi-head attention neural networks and includes convolutional networks, allowing high prediction performance and the ability to estimate confounders. The developed framework presents an innovative methodology for validating potential causes: through a second neural network model, a direct and inverse model is trained with the potential causal lag between the causal variable under study and the target variable. In addition, another potential causal discovery method based on a non-linear correlation test is implemented. The main results show that the proposed model outperforms other causal algorithms concerning the F1 Score and achieves a considerable performance in finding temporal lags. In addition, the network fit (evaluated through the mean squared error) is significantly improved when only the causal variables predicted by the proposed algorithm (with their corresponding causal lags) are used for model training, relative to the incorporation of the entire time series dataset for the prediction of the target variable. • Causal discovery model able to find several causal lags of one variable over another. • Proposed model is based on multi-head attention and convolutional networks. • Proposed model allows high prediction performance and finding confounders. • The framework presents methodology for validating potential causes. [ABSTRACT FROM AUTHOR]

Feature selection is a crucial data preprocessing technique extensively employed in machine learning and image processing. However, feature selection encounters significant challenges when addressing high-dimensional data due to the huge and discrete decision space. This paper proposes a hierarchical learning multi-objective firefly algorithm (HMOFA) for solving the feature selection task in high-dimensional data. The main contributions are as follows: 1) Features are clustered based on the evaluation of multiple metrics, which are used to initialize the population and improve the quality of the initial population; 2) A hierarchy-guided learning model is proposed, where individuals move toward superior solutions while moving away from inferior solutions, avoiding the oscillation phenomenon that occurs under the full attraction model, and reducing the likelihood of the population being trapped in a local optimum; 3) Use duplicate solution modification mechanism to reduce the number of duplicate individuals in the population. The proposed method is compared with 8 competitive feature selection methods using 15 datasets, and the results demonstrate that HMOFA can achieve higher classification accuracy while selecting fewer features. • HMOFA is proposed to solve high-dimensional feature selection tasks. • A clustering initialization method is introduced to reduce redundant features and improve the quality of initial population. • The population updates its position using a hierarchy-guided learning model. • A duplicate solution modification mechanism is applied to enhance the diversity of the population. [ABSTRACT FROM AUTHOR]

In recent times, yoga has gained global prominence due to its wide-ranging health benefits. Heart rate variability (HRV) analysis has become a widely embraced tool to assess the impact of yoga and meditation on individuals. The effects of yoga can vary significantly among individuals based on their prior expertise, skills, and the interplay of external and internal dynamics. This variability complicates the task of extracting distinctive characteristics from raw data to categorize yogic and pre-yogic states. This paper introduces a new approach that enhances the screening of yogic states through multi-resolution analysis of HRV data combined with a machine learning module. The empirical Fourier decomposition (EFD) technique is harnessed to decompose input data into distinct multi-resolution modes. Various statistical characteristics are then extracted from these modes, allowing for a comprehensive analysis of the immediate effects of yoga breathing on HRV signals. To discern between yogic and pre-yogic HRV signals, a selection of significant features is ranked using a chi-square score and subsequently inputted into machine-learning modules. Notably, employing normalized features with the decision medium tree classifier yields an area under the curve (AUC) value of 0.99, while using direct features with an ensemble bagged tree classifier results in an AUC value of 0.82. Furthermore, this research underscores the importance of assessing outcomes from various metrics to compute the underlying dynamics of HRV data during breathing practices. This approach facilitates precise monitoring, personalized feedback, and intervention in yoga machine interfacing systems, ultimately enriching the overall experience and well-being of yoga practitioners. • In recent times, yoga has gained global prominence. • HRV analysis has become a widely embraced tool to assess the impact of yoga. • The EFD technique is harnessed to decompose data into multi-resolution modes. • Statistical characteristics are extracted and ranked using a chi-square score. • A decision medium tree classifier yields an AUC value of 0.99. [ABSTRACT FROM AUTHOR]

Sentiment analysis of customer feedback is a pivotal component of Natural Language Processing (NLP), enabling businesses to gauge consumer emotions towards their products and services. The inherent variability of language introduces substantial challenges in the analysis of unstructured customer data. Sentiment analysis techniques generally fall into two categories: traditional methods, which involve the extraction of hand-crafted features and the application of machine learning algorithms, and end-to-end deep learning approaches that process raw data, transforming them layer by layer into high-level textual features. Each method presents its own balance between simplicity and speed versus analytical power and flexibility, and they all grapple with the need for extensive labeled data and computational resources. To address these challenges, this paper presents an innovative hybrid model that amalgamates RoBERTa, ResNeXt, BiLSTM, and self-attention mechanisms. This integration capitalizes on the collective strengths of these models to surmount the limitations inherent in each. The proposed model proves to be a powerful and efficient tool for sentiment analysis, demonstrating proficiency across various data types and tasks. We undertake a comprehensive evaluation of the proposed model's accuracy and training efficiency using four benchmark datasets. Our investigation also explores the impact of different similarity measures and convolutional neural network architectures on the model's performance. The results affirm that our model not only achieves high accuracy but also significantly reduces training time compared to RoBERTa's fine-tuning. Furthermore, the model exhibits exceptional domain adaptability, particularly when fine-tuned on the IMDb dataset following initial training on the Yelp Review Full dataset. The practicality of the proposed model is underscored by its reduced computational complexity and its adeptness at navigating semantic and syntactic nuances. • A novel hybrid sentiment model using RoBERTa, ResNeXt, BiLSTM, and self-attention. • Outperforms other methods on two datasets; matches RoBERTa base on two others. • Fewer parameters and faster fine-tuning compared to RoBERTa base. • Analysis of document similarity methods and text feature extractors. • TS-SS is viable for low-dimensional features but not high-dimensional ones. [ABSTRACT FROM AUTHOR]

Minimizing the tardiness of a parallel machine scheduling problem has been actively studied in modern manufacturing systems. In particular, dynamic parallel machine scheduling problems (DPMSPs) have gained much attention since rescheduling is required to address unpredictable events such as unexpected job arrivals and machine breakdowns, which usually occur in real-world manufacturing systems. To deal with such unpredictable events, many researchers have employed deep supervised learning (SL). However, it is still challenging to solve a DPMSP since SL requires a large number of high-quality schedules for the training neural networks. In this paper, we propose an incremental learning-based scheduling method (ILS) in which neural networks (NNs) are periodically trained by utilizing schedules built from the updated NNs in previous training intervals. Furthermore, to perform training without reducing solution space within a dynamic environment, the proposed method is designed to consider the dependency between consecutive allocations on a machine for the sum of setup time and tardiness. To verify the effectiveness of the proposed method, extensive experiments are carried out in static and dynamic scheduling problems. The experiment results demonstrate that ILS outperforms the existing methods such as dispatching rules, metaheuristics, and DRL-based methods. Additionally, we show that the proposed input features are effective in appropriately selecting job-machine pairs for assigning jobs on eligible machines through deep Shapley additive explanations analysis. • Address dynamic parallel machine scheduling problem with eligibility restrictions. • Incremental learning is employed to overcome limitations of supervised learning. • Propose a novel input design that can consider all waiting jobs and machine pairs. • Higher performance with proposed model than existing scheduling methods. • Describe the novelty of input features through XAI using SHAP. [ABSTRACT FROM AUTHOR]

In this paper soft computing fuzzy proposition/rule based adaptive extreme learning is proposed. This article illustrates the impact of agricultural solar panel (AgSP) and its operating condition/state during the farming seasons and hence safety of energy system sustained. Extreme learning machine (ExLM) used as learning machine (LM) for artificial intelligence (AI) training to predict the operating state of applied AgSP green energy unit. soft computing Fuzzy rules are framed for every instant of leaning machine to identifying the lower mismatch conditions and avoid the false prediction. Higher and lower limits of adaptive resonance theory (ART) support LM (ART-LM) is properly used by providing unusual dust formation (UnDFR) data and usual/normal dust formation (UsDFR) data for AI training. Outcome of AI neural net is determined by resonant neuron net (RSNN). Predicted variable from ART-LM is calculated with higher and lower limits of dataset captured from applied energy common node (CN). Many test conditions of dust formation on solar photovoltaic (PV) has been validated to predict the state of applied energy system. For any UnDFR confirmation, entire AgSP based distributed generator (DG) unit will isolate from common node/utility grid. Proposed ART-LM based AI algorithm has been validated even for charging state events of electrical vehicle (EV) during the abnormal conditions. • Operating state of PV energy system is identified using ART-LM for all dust conditions. • Accurate mismatch extracted by fuzzy rule extreme learning to protect PV energy system panels. • Fuzzy propositions are confirmed by fractional fuzzy inference reasoning during identification. • SOC of electrical vehicle during abnormal conditions is assessed to obtain energy system safety. [ABSTRACT FROM AUTHOR]

This review paper provides a detailed evaluation of the existing landscape and future trends in applying machine learning and deep learning approaches for predicting concrete strength in construction engineering. The study contextualizes the investigation of machine learning and deep learning in concrete strength prediction, emphasizing the need for precise strength forecasting in construction. This hybrid review uses quantitative analysis of an extensive collection of 1005 research publications from the Scopus database (2010−2023) to identify clusters, hotspots, and gaps in this area, giving a systematic way to analyze the field's dynamics. This review reveals major research clusters such as concrete characteristics, sustainability, error analysis, and optimization. It identifies research hotspots like compressive strength prediction, reinforced concrete, and neural networks. The review illuminates future research paths, ethical concerns, and environmental implications. It emphasizes the relevance of fairness, bias reduction, and sustainability in developing and deploying machine and deep learning models in the construction sector and the necessity for specialized models in forecasting concrete durability, sustainable concrete strength, and shear strength. • Thorough review of on the use of ML and DL in concrete strength prediction is offered. • Critical insights and patterns in the prediction using ML and DL are identified. • Research directions in ML for concrete strength prediction are elaborated on. • Future role of deep learning in advancing concrete strength prediction is revealed. • Notable research paths, ethical concerns, and environmental implications are given. [ABSTRACT FROM AUTHOR]

This paper introduces a novel approach for deriving an equation aimed at predicting concrete compressive strength, leveraging two non-destructive test results: ultrasonic pulse velocity and rebound hammer value. Traditional regression methods often fail to capture the complex non-linear relationships crucial for accurate concrete CS estimation, leading to limited effectiveness. In contrast, while machine learning models excel at mapping these non-linear relationships from non-destructive test results, their intricate internal mechanisms can be opaque, posing challenges for practical application and comprehension by engineers. This study aims to bridge these gaps by developing an estimation equation derived from the decoded weights of a neural network specifically designed to capture the intricate nonlinear mapping relationship. Initially, the equation is intricate and lengthy, comprising multiple terms. To streamline and refine it, a genetic algorithm is employed, focusing solely on the most pivotal terms. Consequently, the refined prediction equation demonstrates superior estimation performance with an RMSE of 3.40 MPa, an MAE of 2.70 MPa, an R2 of 0.92, and an R of 0.97 compared to several existing formulations. Furthermore, a comparative analysis is undertaken to assess the impact of the degree of equation simplification on its predictive accuracy. The findings offer insights into the pivotal roles of exponential function terms in enhancing prediction performance, as well as elucidating trigonometric function terms contribute the model's complex non-linear mapping capability. • A novel scheme for derivation of equations to predict concrete compressive strength is proposed. • Ultrasonic pulse velocity and rebound number are utilized for assessing concrete compressive strength. • A neural network learns the non-linear relationship between non-destructive test results and concrete compressive strength. • To translate the learned non-linear mapping relationship into the mathematical equation, the neural network was customized. • A genetic algorithm refines the neural network-derived estimation equation by selectively preserving key terms. [ABSTRACT FROM AUTHOR]

Determining the correct mushroom species with the necessary ecological characteristics is critical to continue mushroom production, which is essential in gastronomy. The mushroom farmers and collectors technique may help identify toxic mushrooms by detecting poisonous mushrooms using images of different mushroom species with distinctive morphological features. However, it can be not easy to distinguish between species. This paper used a dataset of 6714 mushroom images obtained from nine different mushroom species to classify the mushroom species. For a more straightforward comprehension of mushroom images and feature extraction by reanalysis of data sets, data visualization was performed using Grad-CAM, LIME, and Heatmap methods. Residual block-based Convolutional Neural Network (CNN) architectures are trained to automatically classify the concatenated feature map obtained from the Grad-CAM, LIME, and Heatmap methods. After extracting the deep features of the images from each architecture, the Atom Search Optimization (ASO) algorithm has been used to select the most distinctive features. The 6714×9000 size of the concatenated feature map was reduced to 6714×600 using the ASO algorithm. Classification results were evaluated using six different classifiers based on the feature map obtained to determine the mushroom species. The nine classes of mushroom species were classified successfully with 95.45 % accuracy using the proposed model with the ASO algorithm and KNN classifier. The methodology introduces novel visualization techniques for interpreting CNN-based models' decisions in mushroom species classification tasks. Using metaheuristics-based CNN models with multi-feature fusion techniques allows the model to leverage diverse sources of information, potentially enhancing its ability to discriminate between mushroom species and achieve higher classification accuracy than existing methods. This study can advance the mushroom species classification field by introducing new methodologies, improving classification accuracy, providing insights into model interpretability, and facilitating knowledge transfer to related fields. • The proposed model automatically detects mushroom species. • Grad-CAM, LIME, and Heatmap methods based on CNN architectures are used to visualize mushroom images. • The ASO algorithm has been successfully applied to residual block-based CNN models for feature selection. • The hybrid model classification accuracy for different species of mushrooms is 95.45 %. [ABSTRACT FROM AUTHOR]

The analysis of Medical Images (MI), particularly the detection and classification of anomalies in 3D MRI (Magnetic Resonance Imaging) scans, plays a critical part in timely intervention and personalized therapy plans. In our paper, a comprehensive methodology for anomaly detection in 3D MRI scans of the brain is proposed that combines advanced Deep Learning (DL) techniques, including Convolutional Long Short-Term Memory (ConvLSTM) model, with efficient statistical feature extraction from segmented anomaly regions. The data collection phase utilizes three main datasets namely the Brain Tumor Segmentation Challenge (BRATS), the Federated Tumor Segmentation Challenge (FETS), and the Medical Segmentation Decathlon (MSD). The research begins with preprocessing steps including image resizing, intensity normalization, and alignment of anomaly region segmentation masks. The targeted anomaly regions within the samples are segmented using U-Net architecture and then follow statistical feature extraction procedure. Dimensionality reduction methods such as Principal Component Analysis (PCA) and Recursive Feature Elimination (RFE) are utilized to streamline the feature space. The ConvLSTM is then used to classify the anomalies using both convolution and recurrent layers to capture spatiotemporal patterns in MRI data. The model is fine-tuned and iterated for better classification performance using Adam optimizer. The statistical evaluation results with accuracy of 98.9 % showed that the designed ConvLSTM method is more suitable for clinical diagnosis and treatment design in detecting anomalies in 3D MRI images. • Advanced Anomaly Detection in 3D MRI. • Deep Learning Integration. • Crucial for ensuring data quality and consistency. • This optimizes feature space while preserving discriminative information. [ABSTRACT FROM AUTHOR]

Highlights • We studied a set of two-agent scheduling problem with a single machine. • The paper proposes two heuristics and one metaheuristic to solve the problem. • Lower bound has been developed for checking the effectiveness of the heuristics. • The efficiency of the heuristic is tested through randomly generated instances. Abstract This paper considers two-agent scheduling problem with a single machine which is responsible for processing jobs from two agents. The objective is to minimize the objective function of one agent, subject to an upper bound on the objective function of the other agent. The objectives considered in this paper are, (1) the minimization of total completion time and (2) the minimization of total weighted completion time. To solve these problems, one heuristic and an Ant Colony Optimization algorithm are proposed. The heuristic suggested in the paper are motivated by the Weighted Shortest Processing Time first (WSPT) rule. A numerical experiment is performed on randomly generated problem instances. The performance of the algorithm is evaluated by comparing it with the lower bound value of all three problems considered in the present paper. [ABSTRACT FROM AUTHOR]

Highlights • An in-depth review of deep learning methods for semantic segmentation applied to various areas. • An overview of background concepts and formulation for newcomers. • A structured and logical review of datasets and methods, highlighting their contributions and significance. • Quantitative comparison of performance and accuracy on common datasets. • A discussion of future works and promising research lines and conclusions about the state of the art of the field. Abstract Image semantic segmentation is more and more being of interest for computer vision and machine learning researchers. Many applications on the rise need accurate and efficient segmentation mechanisms: autonomous driving, indoor navigation, and even virtual or augmented reality systems to name a few. This demand coincides with the rise of deep learning approaches in almost every field or application target related to computer vision, including semantic segmentation or scene understanding. This paper provides a review on deep learning methods for semantic segmentation applied to various application areas. Firstly, we formulate the semantic segmentation problem and define the terminology of this field as well as interesting background concepts. Next, the main datasets and challenges are exposed to help researchers decide which are the ones that best suit their needs and goals. Then, existing methods are reviewed, highlighting their contributions and their significance in the field. We also devote a part of the paper to review common loss functions and error metrics for this problem. Finally, quantitative results are given for the described methods and the datasets in which they were evaluated, following up with a discussion of the results. At last, we point out a set of promising future works and draw our own conclusions about the state of the art of semantic segmentation using deep learning techniques. [ABSTRACT FROM AUTHOR]

Sparse coding is based on the concept of having a large dictionary of candidate basis vectors. Any given vector is expressed as a sparse linear combination of the dictionary vectors. It has been developed in the signal processing field, and has many applications in data compression and image processing. In this paper we propose applying sparse coding to the time series forecasting field. Specifically, the paper investigates different dictionary based local learning techniques for building predictive models for the time series forecasting problem. The proposed methodology is based on a local learning framework whereby the query point is embedded and coded in terms of a sparse combination of the training dictionary atoms (vectors). Then this embedding is used for estimating the target value of the query point, by applying the same embedding to the target vectors of the dictionary training atoms. We present an experimental study of several sparse coding algorithms. Experiments are performed on the large monthly time series benchmark from the M3 competition, and these experiments showed that the sparse methods Lasso and Elastic-Net presented the best results among the sparse coding algorithms. Moreover, they outperformed the K -nearest neighbor (KNN) regression and most of the compared machine learning and statistical forecasting techniques, especially for higher horizons. [ABSTRACT FROM AUTHOR]

In this paper we present a novel approach for solving the consequent part of neuro-fuzzy modeling with an emphasis on the forgetting factor in the multi-class learning problem. Our solution is based on recursive least squares (RLS) for online and incremental learning applications, where the data stream is not necessarily uniformly distributed over time. Such a setup can lead to forgetting of specific classes that have not been used for a period of time. In this work we present a reasoned and detailed description of elastic memory learning (EML) and EML with the use of confidence interval (EML+) to avoid unnecessary treatment of the forgetting factor. We present the experimental results and evaluation of our methods in order to show their usefulness not only against forgetting of unused classes, but also for dealing with the lowered recognition rate after all classes have been learned. We note that by using EML, forgetting is significantly eliminated and the recognition rate is slightly affected as well, while EML+ puts more emphasis on keeping the recognition rate higher than the forgetting. Thus, this paper presents two methods that significantly eliminate the forgetting factor for incremental learning with a different focus on its importance, i.e. high recognition rate vs high immunity to forgetting; both of these methods perform significantly better than RLS for these aspects. [ABSTRACT FROM AUTHOR]

Multi-view clustering has achieved many applications in recent years. But the existing multi-view clustering methods face two problems, firstly, the traditional multi-view clustering uses a hard clustering method, which cannot describe the uncertainty between the samples and the clusters, and secondly how to perform effective incremental learning on multi-view data when the number of views increases. Therefore to address the above two problems, we firstly propose a three-way fuzzy spectral clustering algorithm based on fuzzy clustering, which can perform three-way clustering on multi-views and get soft clustering results, thus describing the uncertainty between samples and clusters. Then based on the sequential decision making approach, an incremental learning mechanism is designed for multi-view clustering when the number of views increases. Finally, these two works are combined to propose a multi-view clustering algorithm based on sequential three-way decision making. The experimental results demonstrate that the method proposed in this paper has better clustering accuracy and efficiency compared to the traditional multi-view clustering algorithm. • As incremental data comes in, the performance of the strategy goes up. • Methods can delay decisions when there is insufficient data for multiple views. • Methods are suitable for saving the results obtained by strategies when there is not enough information. • Strategies can be more time efficient than traditional methods. [ABSTRACT FROM AUTHOR]

During the synthesis of acceleration autopilots for high dynamic flight vehicles (HDFV), autopilots with feedback of angular acceleration (AFAA) have become more perspective with stringent requirements on response speed and high maneuverability, compared with autopilots with feedback of angular rate (AFAR). Integral reinforcement learning (IRL) method has now proved to be an effective technique for adaptive optimal control of partially unknown nonlinear systems. In this paper, a novel data-driven IRL algorithm with "actor–critic" structure is proposed for HDFV utilizing AFAA. As an advanced model-free approach, "actor–critic" based IRL algorithm learns optimal behaviors by observing the real-time responses from the environment under the action of nonoptimal control policies. Instead of solving algebraic Riccati equation directly, the control policy updates online via the solution of proposed IRL Bellman equation with sensed quantities. Numerical simulation is carried out to validate the effectiveness of proposed online IRL-based angular acceleration autopilot for HDFVs. Besides, the tracking performance under different wave commands, the robustness against parameter uncertainties and the noise attenuation capacity between classical optimal tracking approach and proposed IRL method are analyzed for AFAR and AFAA, respectively. Simulation results show that, angular acceleration autopilot with proposed integral RL algorithm possesses better tracking performance against various disturbances. • An angular acceleration feedback HDFV autopilot is proposed for higher performances. • A data–driven integral RL algorithm with "actor–critic" structure is proposed. • Tracking performances of autopilots with different algorithms are analyzed. [ABSTRACT FROM AUTHOR]

Genetic Programming (GP) can be used to design effective classifiers due to its built-in feature selection and feature construction characteristics. Unbalanced data distributions affect the classification performance of GP classifiers. Some fitness functions have been proposed to solve the class imbalance problem of GP classifiers. However, with the evolution of GP, single-objective GP classifiers evaluated by a single fitness function have poor generalization ability. Moreover, using the best evolved GP classifier for decision-making can easily lead to the possibility of misclassification. In this paper, multi-objective GP is used to optimize multiple fitness functions including AUC approximation (Wmw), Distance (Dist), and Complexity to evolve ensemble classifiers, which jointly determines the class labels of unknown instances. Experiments on sixteen datasets show that our multi-objective GP can significantly improve classification performance compared with single-objective GP, and our proposed ensemble classifiers evolved by multi-objective GP can further improve the classification performance than the single best GP classifier. Comparisons with six GP-based and five traditional machine learning algorithms show that our proposed approaches can achieve significantly better classification performance on most cases. • Proposed multi-objective GP can significantly improve the classification performance compared with single-objective GP. • Proposed ensemble classifiers can further improve the classification performance. • Proposed approaches can achieve better performance than six GP-based and five traditional machine learning algorithms. [ABSTRACT FROM AUTHOR]

Chimp optimization algorithm (ChOA), which simulates the social behaviors of chimps, is a novel swarm intelligence algorithm for solving global optimization problems. ChOA has the advantages of fast convergence and avoiding falling into local optimum. However, the global search capability is weakened and the time overhead is too large when solving complex optimization problems. In order to improve the overall performance of ChOA, a parallel chimp optimization algorithm based on tracking-learning and fuzzy opposition-learning behaviors (PChOA) is proposed in this paper. First, a tracking-learning behavior is designed to improve the search accuracy. Second, a fuzzy opposition-learning behavior is adopted to enhance the global search capability. Third, a parallel computing architecture is developed to accelerate computational speed. Moreover, the convergence of our proposed PChOA has been analyzed theoretically. To validate the effectiveness of PChOA, it is applied to solve classification problem. The experimental results demonstrate that the classification performance of our proposed algorithm outperforms six other state of the art algorithms on most used datasets. Meanwhile, the time overhead of PChOA is significantly reduced in the environment of parallel computing. When the number of processors is increased to 16, PChOA costs less time than NBTree which is the fastest comparison algorithm in the experiment. • We propose parallel chimp optimization algorithm with two learning behaviors for data classification. • We design a tracking-learning behavior which helps improve the search accuracy. • A fuzzy opposition-learning behavior is adopted to enhance the global search capability. • A parallel computing architecture is developed to accelerate computation speed. • The convergence is analyzed theoretically. [ABSTRACT FROM AUTHOR]

The field of machine learning (ML) has witnessed significant advancements in recent years. However, many existing algorithms lack interpretability and struggle with high-dimensional and imbalanced data. This paper proposes SPINEX, a novel similarity-based interpretable neighbor exploration algorithm designed to address these limitations. This algorithm combines ensemble learning and feature interaction analysis to achieve accurate predictions and meaningful insights by quantifying each feature's contribution to predictions and identifying interactions between features, thereby enhancing the interpretability of the algorithm. To evaluate the performance of SPINEX, extensive experiments on 59 synthetic and real datasets were conducted for both regression and classification tasks. The results demonstrate that SPINEX achieves comparative performance and, in some scenarios, may outperform commonly adopted ML algorithms. The same findings demonstrate the effectiveness and competitiveness of SPINEX, making it a promising approach for various real-world applications. • SPINEX: An interpretable neighbor exploration algorithm is presented. • Combines ensemble learning and feature interaction for meaningful insights. • Quantifies feature contributions and interactions to enhance interpretability. • 59 experiments were conducted to validate SPINEX's comparative performance. [ABSTRACT FROM AUTHOR]