In numerous episodic reinforcement learning (RL) environments, SARSA-based methodologies are employed to enhance policies aimed at maximizing returns over long horizons. Traditional SARSA algorithms face challenges in achieving an optimal balance between bias and variation, primarily due to their dependence on a single, constant discount factor (η). This study enhances the temporal difference decomposition method, TD(Δ), by applying it to the SARSA algorithm, wherein the action-value function is segmented into several components based on the differences between action-value functions linked to specific discount factors. Each component, referred to as a delta estimator (D), is linked to a specific discount factor and learned independently. This modified technique is referred to as SARSA(Δ). SARSA is a widely used on-policy RL method that enhances action-value functions via temporal difference updates. This decomposition, namely SARSA(Δ), facilitates learning across a range of time scales. This analysis makes learning more effective and guarantees consistency, especially in situations where long-horizon improvement is needed. The results of this research show that the proposed technique works to lower bias in SARSA’s updates and speed up convergence in both deterministic and stochastic settings, even in dense-reward Atari environments. Experimental results from a variety of benchmark settings show that the proposed SARSA(Δ) outperforms existing TD learning techniques in both tabular and deep RL environments.

The ability to accurately predict and analyze student performance in online education, both at the outset and throughout the semester, is vital. Most of the published studies focus on binary classification (Fail or Pass) but there is still a significant research shortcoming in predicting performance of students across multiple categories. This study introduces a novel neural network-based approach capable of accurately predicting student performance and identifying vulnerable students at early stages of the online courses. The open university learning analytics (OULA) dataset is employed to develop and test the proposed model, which predicts outcomes in Distinction, Fail, Pass, and Withdrawn categories. The OULA dataset is preprocessed to extract features from demographic data, assessment data, and clickstream interactions within a virtual learning environment (VLE). Novel features engineering has been utilized to predict students' performance across multiple categories at early stages of courses. Specially, students' VLE interactions are aggregated by total clicks to represent daily engagement and assess online activity. Comparative simulations indicate that the proposed model significantly outperforms existing baseline models including artificial neural network long short-term memory (ANN-LSTM), random forest (RF) 'gini', RF 'entropy' and deep feed forward neural network (DFFNN) in terms of accuracy, precision, recall, and F1-score. The results indicate that the prediction accuracy of the proposed method is about [Formula: see text] more than the existing state-of-the-art methods. Furthermore, compared to existing methodologies, the model demonstrates superior predictive capability across temporal course progression, achieving superior accuracy even at the initial [Formula: see text] phase of course completion.

Optimization techniques are widely used for their simplicity and robust performance, making them a better choice for solving complex problems across various domains. Among these techniques, metaheuristic algorithms have shown great promise due to their flexibility and ability to handle diverse optimization challenges. This study introduces a Chaotic Gorilla Troops Optimizer (CGTO), which enhances the Gorilla Troops Optimizer (GTO) based on the Logistic map to improve its search capabilities. While GTO is a promising algorithm, it suffers from certain limitations, such as restricted global search capability, slow convergence speed, and a tendency to fall into local optima when addressing complex optimization problems. To address these problems, initially, the proposed CGTO incorporates enhancements designed to improve convergence behavior and exploitability, making it more effective for a broad range of optimization tasks. The CGTO achieves these improvements by modifying the position update mechanism of the GTO algorithm through the integration of chaotic maps. Secondly, the performance of the e algorithm's performance is with PSO and the original GTO algorithm. Then the CGTO is verified through, unimodal, multimodal, and fixed-dimensional multimodal functions which shows that CGTO consistently outperformed its counterparts in most cases. Finally, the obtained result shows CGTO's superiority in terms of solution quality and convergence rate, establishing its effectiveness as a robust optimization technique for solving complex problems. These findings demonstrate the potential of CGTO to address the limitations of traditional algorithms while delivering reliable and efficient optimization performance.

Metaheuristic algorithms (MAs) are powerful tools for solving complex optimization problems across diverse domains. This comprehensive study analyzes 162 MAs through a unified multi-criteria taxonomy classifying algorithms by control parameters (parameter-free, low-parameter, high-parameter), inspiration sources (biological, physical, human-based), search space scope, and exploration–exploitation balance—alongside bibliometric assessment of publications from 2000 to 2024 (including articles, reviews, books, and conference papers). We evaluate the time complexity of 24 highly cited MAs and their real-world applications in engineering, healthcare, and energy systems. Results demonstrate that different MAs exhibit algorithmic simplicity, often requiring only a small number of control parameters, and are capable of efficient global search. Additionally, several algorithms demonstrate strong adaptability for hybridization with other techniques. However, certain methods are prone to premature convergence, primarily due to unbalanced exploration and exploitation dynamics. The proposed work also critically examines metaphor-inspired algorithms, whose contributions are critically evaluated in light of their conceptual complexity. The rise of such methods has led to redundancy and fragmentation in the field, as many reframe familiar optimization principles using superficial metaphors rather than advancing core algorithmic mechanisms. Bibliometric analysis reveals accelerated growth in MA research (64

To address complex optimization challenges in engineering, such as the welded beam design (WBD) problem, this research work introduces the Artificial Gorilla Troops Optimizer (GTO), an innovative technique inspired by the social intelligence demonstrated by real-life gorilla troops due to certain limitation in traditional optimization methods such as slow convergence. This work presents mathematical formulations capturing gorilla group dynamics and introduces exploration and exploitation mechanisms. A comparative analysis with the traditional Particle Swarm Optimization (PSO) algorithm shows that GTO achieves a 19 % lower optimal value than PSO. Furthermore, statistical analysis further validates the GTO accuracy, robustness, and competitiveness, demonstrating a better solution and convergence rate than PSO.

Efficient network coverage and connectivity in wireless sensor networks (WSNs) is critical for modern data-driven applications requiring seamless data collection and transmission. One of the key challenges is the optimal placement of sensor nodes, which directly impacts network performance and deployment costs. This study presents an Improved Chaotic Grey Wolf Optimization (ICGWO) algorithm to enhance WSN coverage and connectivity while addressing challenges like high deployment costs, limited coverage, and insufficient connectivity. A mathematical model for the WSN coverage and connectivity optimization problem is developed as the foundation. The Grey Wolf Optimizer (GWO) is enhanced using a chaotic map, improving its ability to find the best solutions and achieve faster convergence, resulting in the ICGWO algorithm. The performance of ICGWO is evaluated using CEC_22 benchmark functions and compared with other optimization methods, demonstrating clear improvements in efficiency. In practical applications, the proposed ICGWO obtained superior results for sensor node placement. For example, with 20 sensor nodes in Case 1, the coverage rate reaches 95.9077%, while for 30 nodes in Case 2, it achieves 98.2211%. Similarly, in Case 3, with 40 sensor nodes, the coverage rate is 91.6875%, and in Case 4, with 50 sensor nodes, it is 99.4940%. In addition, in Case 5 and Case 6, with 60 and 70 sensor nodes, the coverage rates are 99.7801% and 99.7822%, respectively. These outcomes reflect average improvements of 16.41%, 5.36%, 3.45%,2.371%,2.80%, and 2.18%, respectively, compared to other state-of-the-art methods. These metrics emphasize the effectiveness of ICGWO in maximizing network coverage and connectivity. The findings confirm that ICGWO efficiently improves the coverage and connectivity, making it a reliable solution for addressing deployment challenges in diverse scenarios. By maximizing the coverage and connectivity, ICGWO significantly contributes to the advancement of WSN technology.

Interdisciplinary teaching is an important way to cultivate innovative talents and abilities. The theme of this project is to enhance the interdisciplinary innovation ability of junior high school students, and to design and develop a micro course titled "Ideal to Reality? IFR Helps Me!". Through interviews with frontline junior high school teachers, it was found that most teachers focus more on designing innovative and interdisciplinary case studies to teach subject knowledge, rather than cultivating students' interdisciplinary innovation abilities. The interviewed teachers, based on their interdisciplinary teaching experience, have put forward requirements and design concepts for micro courses that cultivate interdisciplinary innovation abilities, including emphasizing interdisciplinary ability enhancement, animation, case-based learning, as well as dialogue between teachers' real faces and virtual characters. This project is based on the above design concept, starting from the analysis, design, development, implementation, self-evaluation, feedback and other aspects of micro courses, proposing practical and feasible micro course design and production strategies to enhance the interdisciplinary innovation ability of middle school students. The micro course design case has been completed, aiming to provide useful references for middle school teachers to design interdisciplinary teaching cases and micro courses, fundamentally improve the interdisciplinary innovation ability of middle school students, and promote the cultivation process of national innovative talents.

The concept of STEAM education integrates the disciplinary concepts of science, technology, engineering, art and mathematics, aiming to cultivate students' practical hands-on skills and improve their Core Literacy through an interdisciplinary approach. The purpose of this study is to investigate the current situation and problems faced by the Information technology at the compulsory education level, and to redesign the interdisciplinary Information technology based on the concepts of the New Curriculum Standards and in conjunction with the intangible cultural heritage of the Chaoshan region, in order to cultivate students' Core Literacy and practical problem-solving abilities.

A University-Enterprise Deep Integration-Oriented Talent Training Approach (UED-IOTTA) was designed to improve the post-competency and employment prospects of software engineering graduates from colleges and universities. This approach is tightly aligned with the demand for talent in enterprises and the development of talent in academic in-stitutions on a variety of dimensions, including employment requirements, professional ethics, and occupational skills. Building on this paradigm, a specific implementation strategy has been developed. This implementation strategy's key components include establishing industry-specific training classes known as UED-IOTTA classes, providing software engineering vocational training, putting industry-academic collaboration supervision techniques into practice, developing progressive pedagogy, and defining particular university-enterprise partnership strategies. Empirical results indicate the noteworthy influence of this cus-tomized talent training approach on improving the caliber of talent advancement in the field of software engineering.

Nowadays, cluster analyses are widely used in mental health research to categorize student stress levels. However, conventional clustering methods experience challenges with large datasets and complex issues, such as converging to local optima and sensitivity to initial random states. To address these limitations, this research work introduces an Improved Grey Wolf Clustering Algorithm (iGWCA). This improved approach aims to adjust the convergence rate and mitigate the risk of being trapped in local optima. The iGWCA algorithm provides a balanced technique for exploration and exploitation phases, alongside a local search mechanism around the optimal solution. To assess its efficiency, the proposed algorithm is verified on two different datasets. The dataset-I comprises 1100 individuals obtained from the Kaggle database, while dataset-II is based on 824 individuals obtained from the Mendeley database. The results demonstrate the competence of iGWCA in classifying student stress levels. The algorithm outperforms other methods in terms of lower intra-cluster distances, obtaining a reduction rate of 1.48% compared to Grey Wolf Optimization (GWO), 8.69% compared to Mayfly Optimization (MOA), 8.45% compared to the Firefly Algorithm (FFO), 2.45% Particle Swarm Optimization (PSO), 3.65%, Hybrid Sine Cosine with Cuckoo search (HSCCS), 8.20%, Hybrid Firefly and Genetic Algorithm (FAGA) and 8.68% Gravitational Search Algorithm (GSA). This demonstrates the effectiveness of the proposed algorithm in minimizing intra-cluster distances, making it a better choice for student stress classification. This research contributes to the advancement of understanding and managing student well-being within academic communities by providing a robust tool for stress level classification.

In this paper, we introduce an efficient and effective algorithm for Graph-based Semi-Supervised Learning (GSSL). Unlike other GSSL methods, our proposed algorithm achieves efficiency by constructing a bipartite graph, which connects a small number of representative points to a large volume of raw data by capturing their underlying manifold structures. This bipartite graph, with a sparse and anti-diagonal affinity matrix which is symmetrical, serves as a low-rank approximation of the original graph. Consequently, our algorithm accelerates both the graph construction and label propagation steps. In particular, on the one hand, our algorithm computes the label propagation in closed-form, reducing its computational complexity from cubic to approximately linear with respect to the number of data points; on the other hand, our algorithm calculates the soft label matrix for unlabeled data using a closed-form solution, thereby gaining additional acceleration. Comprehensive experiments performed on six real-world datasets demonstrate the efficiency and effectiveness of our algorithm in comparison to five state-of-the-art algorithms.

Anxiety is an important issue that affects their academic performance, mental health, and overall educational journey. To address this issue, it is important to accurately assess anxiety levels and provide evidence-based techniques. However, due to the complexity of anxiety and individual differences, analyzing clustering algorithms to efficiently classify psychological levels is challenging. Traditional clustering techniques face certain challenges in accurately classifying anxiety levels, such as slow convergence, sensitivity to initial conditions, and difficulties in handling constraints. To address these issues, clustering with an improved Mayfly-based optimization algorithm (IMOA) is proposed based on the dynamic variable for better performance to classify psychological levels. Initially, IMOA is validated using 23 standard benchmark functions, confirming its ability to find optimal solutions. Then, IMOA is applied to the student dataset, classifying them into Cluster A and Cluster B. The average scores for both clusters across all test cases are 76.7% and 53.07%, respectively. These results demonstrate the formation of dissimilar student groups with homogeneous emotions and performance, highlighting the importance of addressing emotional stress. Finally, by assigning students to clusters, educators and mental health professionals can better support those who may struggle, ensuring they receive the attention and resources they need. The obtained results show that IMOA with a dynamic variable effectively classifies student anxiety, improving the learning environment and helping teachers better understand students’ needs. This identification allows them to provide more effective support and adapt their teaching to meet the specific needs of those seeking support.

This paper discusses a reduction in the optimal time due to the presence of input redundancy in time-optimal control problems. By introducing a non-idle channel to represent an active input channel, we establish the necessary and sufficient conditions that ensure a strict reduction in the optimal time for affine nonlinear systems. In cases of identical input redundancy, its impact varies according to the type of input constraint, and certain types may not lead to a reduction in the optimal time. Ultimately, in linear time-invariant (LTI) systems, the extent of the optimal time reduction depends on the system’s controllability.

This paper introduces the Group Forward–Backward Orthogonal Matching Pursuit (Group-FoBa-OMP) algorithm, a novel approach for sparse feature selection. The core innovations of this algorithm include (1) an integrated backward elimination process to correct earlier misidentified groups; (2) a versatile convex smooth model that generalizes previous research; (3) the strategic use of gradient information to expedite the group selection phase; and (4) a theoretical validation of its performance in terms of support set recovery, variable estimation accuracy, and objective function optimization. These advancements are supported by experimental evidence from both synthetic and real-world data, demonstrating the algorithm’s effectiveness.

A variety of reasons have made it more difficult for educators and tutors to anticipate students' performance. Numerous researchers have used various predictive models to identify students who may be at-risk of dropping out early. Additionally, these methods were used to forecast final semester grades based on various datasets. However, these prediction models still fall short of meeting educational management requirements. In this paper, we propose the deep learning (DL) based model named students academic performance prediction network (SAPPNet) to predict the students' grades. We consider the questionnaire-based Jordan University dataset which contains demographic information, usage of digital tools before and after COVID-19, sleep times before and after COVID-19, social interaction, psychological state, and academic performance. SAPPNet consists of spatial convolution modules which are designed to extract spatial dependencies includes categorical and numerical attributes that represent static features (gender, level/year, age, digital tools used before and after COVID-19, psychological condition using prolonged e-learning tools) and temporal module for temporal dependencies involves sequences that capture changes before and after COVID-19. Additionally, we also try to implement classical machine learning (ML) models including support vector machine, k nearest neighbor, decision tree, and random forest, and DL models named artificial neural network, convolutional neural network, long short-term memory, and students learning prediction network. Simulation results show that SAPPNet achieved the best performance compared to state-of-the-art methods, with an accuracy, precision, recall, and an F1-score of 93%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$93\%$$\end{document}. The proposed model with spatial and temporal modules improves the prediction performance, and it implies new aspect of the educational dataset.

Transmission line (TL) parameters, particularly capacitance, are important for ensuring the efficient and reliable operation of power systems. As power networks become increasingly complex, accurately determining TL parameters faces certain challenges, particularly capacitance, which involves overcoming computational challenges due to bundling configurations. The interconnected nature of modern TL and the use of advanced technologies further add to the complexity. Effective estimation requires advanced measurement techniques and sophisticated computational tools. Recently, optimization techniques have become widely used for TL parameter calculations. However, traditional methods struggle with challenges like limited exploration and slow convergence. To address these issues, this research introduces a hybrid algorithm called HGWPSO, which combines the Grey Wolf Optimizer (GWO) with Particle Swarm Optimization (PSO). The main objective is to enhance the exploitation ability of GWO with the exploration capability of PSO, maximizing the strengths of both variants. Initially, to verify the efficiency of HGWPSO, CEC_19 benchmark functions are utilized to evaluate its performance. Secondly, the main focus of this study is to calculate TL parameters such as capacitance considering two, three, and four-bundle conductors using the HGWPSO algorithm and comparing its performance with other optimization techniques. According to the obtained result, the average percentage reduction for HGWPSO is 0.15 % in test case 1, 4.85 % in test case 2, and 2.84 % in test case 3, compared to others. It shows that the HGWPSO has better performance than other methods regarding convergence speed, and ability to locate the global optimum. Finally, experimental analysis confirms the superiority of the HGWPSO in accurately estimating TL capacitance for different bundle conductor configurations, obtaining lower average values, and effectively addressing the characteristic complexities in the TL parameter.

The prediction of students’ performance has become challenging for tutors and management in education because of various factors. Many researchers have deployed different predictive models to predict students who are at risk of dropping out early or estimate grades at the end of the semester on different datasets. Nevertheless, prediction models cannot fulfill the requirements of educational management. In this paper, we propose a deep learning (DL) model named the student learning performance prediction network (SLPNet) to predict students’ grades, where we consider the Quaternaries-based Jordan University dataset, which contains demographic information, digital tools, sleep habits, social interaction, psychological state, and academic performance (assignments, quizzes, and other tasks). Additionally, we have attempted to implement other DL and machine learning (ML) models, including artificial neural networks (ANNs), support vector machines (SVMs), k nearest neighbors (K-NNs), decision trees (DTs), and random forests (RFs). Furthermore, the simulation results show that the proposed SLPNet model achieves better performance than do the ANN and ML baseline methods, with 89

Human action recognition methods based on skeleton data have been widely studied owing to their strong robustness to illumination and complex backgrounds. Existing methods have achieved good recognition results; however, they have certain challenges, such as the fixed topological structure of the graph, the omission of nonphysical joint correlation, and the inability to extract local spatial–temporal features. Herein, we propose spatial–temporal mixing of global and local self-attention graph convolutional networks (STGL-GCN) using skeleton data. The global self-attention matrix captures the potential dependencies of nonphysical correlations between joints, and the local self-attention matrix determines the connection strength of the physical edges of joints. The matrices are updated together with the convolution parameters in each network layer as the model is trained for optimal graph structure to achieve accurate action expressions Experiments on the NTU-RGBD dataset demonstrate that our model accurately recognizes actions.

Causal discovery is one of the most important research directions in the field of machine learning, aiming to discover the underlying causal relationships in the observed data. In practice, the time complexity of causal discovery will grow exponentially with increasing variables. To alleviate this problem, many methods based on divide-and-conquer strategies have been proposed. Existing methods usually partition the variables heuristically using scattered variables to achieve the dividing process, which makes it difficult to minimize vertex cut-set C and then leads to diminished causal discovery performance. In this work, we design an elaborated causal partition strategy called Causal Partition Base Graph (CPBG) to solve this problem. CPBG uses a set of low-order conditional independence (CI) tests to construct a rough skeleton S corresponding to the observed data and takes a heuristic method to search S for the optimal vertex cut-set C . Then the observed data can be partitioned into multiple variable subsets. We therefore can run a causal discovery method on each part and finally obtain the complete causal structure by merging the partial results. The proposed method is evaluated by various real-world causal datasets. Experimental results show that the CPBG method outperforms its existing counterparts, which proves that the method can support more effective and efficient causal discovery. The source code of the proposed method and all experimental results are available at https://github.com/DreamEdm/Causal .