72,032 results on '"LEARNING strategies"'
Search Results
2. Typologies of Teaching Strategies in Classrooms and Students' Metacognition and Motivation: A Latent Profile Analysis of the Greek PISA 2018 Data
- Author
-
Ioannis G. Katsantonis
- Abstract
Classical conceptualisations of self-regulated learning typically ignore the role of teaching strategies in real-world classrooms. Therefore, the present exploratory study aimed to examine the different clusters of perceived teaching strategies and students' metacognitive knowledge and experiences, and motivation. The data came from 6365 (49.63% female) Greek secondary school students who participated in PISA 2018. Latent profile analyses revealed four patterns of perceived teachers' strategies and students' self-regulated learning, namely 'struggling learners' (18.60%), 'resilient autonomous learners' (19.79%), 'average learners' (32.73%), and 'self-regulated thriving learners' (28.86%). Comparisons with reading achievement competencies revealed that 'struggling learners' were the worst off followed by the 'average learners', whereas 'self-regulated thriving learners' and 'resilient autonomous learners' had the best reading scores. Good metacognitive knowledge and lower feelings of difficulty along with a higher self-concept could compensate for a reduced perceived student-centric instruction and lead to notable achievement gains. Perceived teacher-directed instruction complemented enthusiastic, student-centric instruction, and a good disciplinary climate were associated with higher metacognition and motivation, and optimal learning. In conclusion, the findings suggest that high metacognition and motivation can overcome perceived instructional barriers and, when combined with a perceived academically supportive teaching environment, can lead to maximum success in reading achievement.
- Published
- 2025
- Full Text
- View/download PDF
3. When Two Learners Are Better than One: Using Flashcards with a Partner Improves Metacognitive Accuracy
- Author
-
Megan N. Imundo, Inez Zung, Mary C. Whatley, and Steven C. Pan
- Abstract
We investigated the benefits of two ways to use flashcards to perform retrieval practice: alone versus with a partner. In three experiments, undergraduate students learned word-definition pairs using flashcards alone (Individual condition) or with another student (Paired condition). Participants then made global judgments of learning (gJOLs; Experiments 1-3), and item-level judgments of learning (iJOLs; Experiment 3). Finally, participants took a cued-recall test after a 5-min delay (Experiments 1-3) and a 24-h delay (Experiments 2-3). In Experiment 1, students in the Paired condition dropped flashcards less often than in the Individual condition (dropping was prohibited in Experiments 2-3). In addition, although final test performance tended to be similar across conditions, inaccurate gJOLs for the immediate test--inflated by ~ 20% relative to actual immediate test performance--were common in the Individual condition but not in the Paired condition in Experiments 1-2. In Experiment 3, we tested whether this difference in metacognitive calibration was due to the Paired condition requiring overt retrieval by instructing participants in the Individual condition to retrieve out loud. With this change, participants in the Individual and Paired conditions reported similarly accurate gJOLs and iJOLs. Taken together, these findings suggest that although performing retrieval practice with flashcards alone versus with a partner yields comparable amounts of learning, doing so with a partner can increase metacognitive accuracy, a benefit possibly driven by the facilitation of overt retrieval. Overall, these findings have implications for self-regulated learning and effective exam preparation.
- Published
- 2025
- Full Text
- View/download PDF
4. Grit and self-regulated learning: evaluating achievement goals as mediators
- Author
-
Won, Sungjun, Wolters, Christopher A., Brady, Anna C., and Hensley, Lauren C.
- Published
- 2025
- Full Text
- View/download PDF
5. Critical vector based evolutionary algorithm for large-scale multi-objective optimization.
- Author
-
Zhu, Shuwei, Wang, Wenping, Fang, Wei, and Cui, Meiji
- Subjects
- *
MULTI-objective optimization , *COMPUTATIONAL mathematics , *MATHEMATICAL optimization , *BENCHMARK problems (Computer science) , *LEARNING strategies - Abstract
In this work, we propose a method for solving large-scale multi-objective problems based on problem transformation strategy. The key point of this method lies in how to construct the search subspace. First, the algorithm obtains a set of direction vectors in the decision space, which are combined in pairs to construct a set of subspaces. To obtain direction vectors with a uniform distribution as much as possible, we introduce the opposition-based learning strategy. Then, based on these subspaces, the original high-dimensional problem is transformed into a relatively lower-dimensional problem. A multi-objective evolutionary algorithm is used to quickly obtain a set of quasi-optimal solutions for the transformed lower-dimensional problem, and this set of solutions is further optimized in the original high-dimensional decision space. To validate its performance, the proposed algorithm is compared with six state-of-the-art large-scale multi-objective algorithms on various benchmark test problems, including one practical application. The experimental results demonstrate that the proposed algorithm shows competitive performance for dealing with large-scale multi-objective optimization problems. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
6. A survey of latent factor models in recommender systems.
- Author
-
Alshbanat, Hind I., Benhidour, Hafida, and Kerrache, Said
- Subjects
- *
ARTIFICIAL neural networks , *OPTIMIZATION algorithms , *DIGITAL technology , *LEARNING strategies , *MATHEMATICAL optimization , *RECOMMENDER systems - Abstract
Recommender systems are essential tools in the digital era, providing personalized content to users across various domains, including e-commerce, entertainment, and social media. Among the numerous approaches developed for these systems, latent factor models have proven to be particularly effective. This survey systematically reviews latent factor models in recommender systems, highlighting their core principles, methodologies, and recent advancements. The literature is analyzed through a structured framework that organizes prior work into a well-defined taxonomy based on four key aspects where advancements in latent factor models have occurred: learning data, model architecture, learning strategies, and optimization techniques. The analysis includes a taxonomy of contributions and detailed discussions on the types of learning data used, such as implicit feedback, trust data, and content data. Additionally, it explores various models, including probabilistic, nonlinear, and neural models, and examines diverse learning strategies like online learning, transfer learning, and active learning. Furthermore, the survey addresses the optimization strategies employed to train latent factor models, which enhance their performance and scalability. By identifying trends, gaps, and potential research directions, this survey aims to provide valuable insights for researchers and practitioners seeking to advance the field of recommender systems. [Display omitted] • Provides a comprehensive survey of the field of latent factor recommender systems. • Taxonomizes the field by data, model, learning strategy, and optimization algorithm. • Reviews principles, methods, and recent advances in the field. • Identifies trends, gaps, and future research directions. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
7. IF-USOD: Multimodal information fusion interactive feature enhancement architecture for underwater salient object detection.
- Author
-
Yuan, Genji, Song, Jintao, and Li, Jinjiang
- Subjects
- *
CONVOLUTIONAL neural networks , *UNDERWATER construction , *TRANSFORMER models , *LEARNING strategies - Abstract
Underwater salient object detection (USOD) has garnered increasing attention due to its superior performance in various underwater visual tasks. Despite the growing interest, research on USOD remains in its nascent stages, with existing methods often struggling to capture long-range contextual features of salient objects. Additionally, these methods frequently overlook the complementary nature of multimodal information. The multimodal information fusion can render previously indiscernible objects more detectable, as capturing complementary features from diverse source images enables a more accurate depiction of objects. In this work, we explore an innovative approach that integrates RGB and depth information, coupled with interactive feature enhancement, to advance the detection of underwater salient objects. Our method first leverages the strengths of both transformer and convolutional neural network architectures to extract features from source images. Here, we employ a two-stage training strategy designed to optimize feature fusion. Subsequently, we utilize self-attention and cross-attention mechanisms to model the correlations among the extracted features, thereby amplifying the relevant features. Finally, to fully exploit features across different network layers, we introduce a cross-scale learning strategy to facilitate multi-scale feature fusion, which improves the detection accuracy of underwater salient objects by generating both coarse and fine salient predictions. Extensive experimental evaluations demonstrate the state-of-the-art model performance of our proposed method. • Introduced IF-USOD for depth and RGB fusion in underwater object detection. • Proposed multimodal cross-scale learning to improve detection precision. • Enhanced feature fusion with a full-perception cross-attention module. • Achieved state-of-the-art performance in underwater detection tasks. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
8. Active in-context learning for cross-domain entity resolution.
- Author
-
Zhang, Ziheng, Zeng, Weixin, Tang, Jiuyang, Huang, Hongbin, and Zhao, Xiang
- Subjects
- *
LANGUAGE models , *DEEP learning , *LEARNING strategies , *LANGUAGE ability - Abstract
Entity resolution (ER) is the task of determining the equivalence between two entity descriptions. In traditional settings, the testing data and training data come from the same domain, e.g., sharing the same attribute structure. Nevertheless, in practical situations, the testing and training data often span different domains, hence calling for the study of the cross-domain ER problem. To tackle the domain shift in cross-domain ER, state-of-the-art solutions devise neural models to utilize the information from the entity pairs in the target domain to guide the feature modeling in the source domain and also the model training. Nevertheless, these approaches require excessive computational resources and fine-tuning efforts to achieve effective matching. To mitigate these issues, in this work, we for the first time investigate the in-context learning (ICL) capabilities of large language models (LLMs) for cross-domain ER and introduce a new framework, CiDER. CiDER consists of three main modules, i.e., active candidate source data generation, in-context demonstration selection, and prompt generation, which can select optimal demonstrations from the source data to enhance LLM inference performance on ER in the target domain. Comprehensive experiments on multiple benchmarks demonstrate that CiDER offers significant improvements over existing methods on cross-domain ER. • We are among the first attempts to explore the ability of large language model (LLM) for performing cross-domain entity resolution tasks, which avoids the extensive fine-tuning efforts required by existing methods. • We devise an active in-context learning strategy and a domain-invariant similarity calculation method to improve the demonstration selection and unleash the power of LLM. • We have thoroughly evaluated our proposal by comparing with other LLM-based and state-of-the-art deep-learning based methods on different datasets, validating the effectiveness of our method. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
9. Hierarchical bipartite graph based multi-view subspace clustering.
- Author
-
Zhou, Jie, Nie, Feiping, Luo, Xinglong, and He, Xingshi
- Subjects
- *
HIERARCHICAL clustering (Cluster analysis) , *LEARNING strategies - Abstract
Multi-view subspace clustering has attracted much attention because of its effectiveness in unsupervised learning. The high time consumption and hyper-parameters are the main obstacles to its development. In this paper, we present a novel method to effectively solve these two defects. First, we employ the bisecting k-means method to generate anchors and construct the hierarchical bipartite graph, which greatly reduce the time consumption. Moreover, we adopt an auto-weighted allocation strategy to learn appropriate weight factors for each view, which can avoid the influence of hyper-parameters. Furthermore, by imposing low rank constraints on the fusion graph, our proposed method can directly obtained the cluster indicators without any post-processing operations. Finally, numerous experiments verify the superiority of proposed method. • We propose a multi-view subspace clustering based on hierarchical bipartite graph to effectively handle large-scale data clustering task. • The bisecting k-means as anchors selection method is adopted to enhance the stability of anchor-based clustering algorithms. • An auto-weighted allocation strategy is employed to automatically learn appropriate weight factors for each view in fusion. • The clustering results can be obtained directly by imposing low rank constraints without any post-processing operations. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
10. Practice and reflection of differentiated learning in sociology at senior high school.
- Author
-
Hati, Kusuma, Rahmayanti, Ayu Afriliani, Aprilia, Cinta Widi Happy, Nisa, Luthfiyah An, Anggraeni, Meisita, Trinugraha, Yosafat Hermawan, and Parahita, Bagas Narendra
- Subjects
HIGH school students ,COGNITIVE styles ,CURRICULUM ,SOCIAL skills ,SOCIOLOGY - Abstract
Global challenges require the government to organize learning in schools that are more up-to-date following episodes in the times. The independent curriculum learning promises flexibility for teachers and students and replaces the relatively new curriculum-13. In implementing the independent curriculum, differentiated learning is a concept that focuses on the diversity of students' learning styles. This research was conducted to obtain an overview of the implementation and reflection of differentiated learning in senior high schools in Central Java. Teachers, especially those teaching sociology subjects, the assistant principal of academic affairs and curriculum, and students are the research subjects. It is qualitative research with a phenomenological research type and purposive sampling technique. The results of this study show: i) the practice of differentiated learning in sociology subjects in high schools has been well implemented with the independent curriculum approach; ii) the challenges and obstacles faced by sociology teachers when implementing differentiated learning are the teacher's skills in preparing learning; and iii) sociology teachers who face challenges and obstacles when implementing differentiated learning must pay more attention to the readiness of differentiated skills. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
11. Modeling and multi-objective optimal state-dependent control of a continuous double-bioreactor in series fermentation.
- Author
-
Wang, Juan, Zhao, Feiyan, Wang, Jichao, and Li, An
- Subjects
- *
NONLINEAR dynamical systems , *PARETO optimum , *LEARNING strategies , *FERMENTATION , *PARAMETERIZATION - Abstract
This paper models a continuous double-bioreactor in series fermentation of glycerol to 1,3-propanediol by a nonlinear dynamic system and formulates its process control by a multi-objective optimal control problem formulating the dilution rates as varying-coefficient state-dependent controls. Control parameterization and time scale transformation are firstly applied to transform the proposed optimal control problem into a large-scale parameter optimization problem, which is then solved by a novel numerical algorithm based on an improved dynamic neighborhood learning strategy and a classified pairwise competition mechanism. Numerical results suggest that the proposed algorithm has good diversity of solutions and convergence to the Pareto optimal front for complex multi-objective problems. Numerical comparisons indicate that the proposed control has the characters of shorter computation time, higher calculation accuracy, and poorer stability when compared to two closed-loop controls, and is better in system stability and improving mean productivity compared to two other open-loop controls. Simulation curves also show the potential application of double-bioreactor in series fermentation. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
12. Dynamics of self-regulated learning: The effectiveness of students' strategies across course periods.
- Author
-
Cristea, T.S., Heikkinen, S., Snijders, C., Saqr, M., Matzat, U., Conijn, R., and Kleingeld, A.
- Subjects
- *
SELF-regulated learning , *DIGITAL learning , *LEARNING management system , *DEEP learning , *TRACE analysis , *LEARNING strategies - Abstract
Proper self-regulating skills are essential in the new reality of digital learning in higher education. Research has shown that the trace data of students' learning management system activity can identify various online learning tactics and strategies, but also their transitional dynamics, which are linked to academic performance. This study builds on this work by examining how learning tactics and strategies change within individual courses and how this relates to academic performance. A substantial dataset of 41 courses over two academic years at one university is analyzed. Employing Markov models on trace data, we identify prevalent tactics and strategies students use throughout courses. Our study examines shifts in strategy usage, comparing patterns between the initial and latter stages of the courses. The results reveal distinct clusters of learning strategies and their impact on academic achievement. Notably, deep learning strategies show significantly superior performance to surface approaches, especially when maintained over time. Moreover, students who consistently apply the same strategy score higher than those who are inconsistent. However, consistent surface learners score significantly lower than inconsistent learners. Underscoring such longitudinal trends could help interventions, aiding educators in targeting students with weaker strategies at specific times to boost their effectiveness and efficiency. This research contributes to a nuanced understanding of self-regulated learning behaviors in online educational contexts by showing the importance of dynamic transition of learning strategies for educators, instructional designers, and policymakers to enhance student learning experiences and outcomes. • Used pattern analysis on trace data to identify four tactics and four strategies. • Explored the strategies students used per course, course halves, and course quarters. • Identified distinct clusters of surface and deep learning strategies at all levels. • Deep learning strategies outperformed surface approaches when applied consistently. • Consistent surface learning scored worse even when compared to inconsistent approaches. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
13. Sparse holographic tomography reconstruction method based on self-supervised neural network with learning to synthesize strategy.
- Author
-
Liu, Yakun, Xiao, Wen, and Pan, Feng
- Subjects
- *
NUMERICAL apertures , *TOMOGRAPHY , *LEARNING strategies , *INFORMATION networks , *ANGLES - Abstract
• Sparse-angle digital holographic tomography reconstruction can be achieved. • Predict the phase images at unmeasured angles in an unsupervised approach. • Enhance high-frequency information by synthetic network structure. • Has good generalization and robustness. This research proposes a novel method for sparse digital holographic tomography reconstruction. Due to the limitations of numerical aperture and sampling time, the development of a high-precision sparse digital holographic tomography reconstruction techniques is necessitated. Our main innovation is the developing a composite coordinate-based implicit neural network with learning to synthesize strategy. It addresses the information limitations of limited angle by directly mapping the sample's rotation angle and coordinates to the phase images, allowing for the prediction of phase images at unmeasured angles without requiring external training dataset. Furthermore, it avoids the issue of high-frequency suppression caused by the uneven distribution of frequency information in the images and the network's characteristics using separately processing low-frequency and high-frequency information in different channels, resulting in higher fidelity of the predicted phase images and the tomographic results. We validated the effectiveness of the proposed method on four different fiber structures at various sampling intervals. This method provides a new perspective for tomographic reconstruction at sparse angles. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
14. Breaking the confinement of fixed nodes: A causality-guided adaptive and interpretable graph neural network architecture.
- Author
-
Wang, Chao, Zhou, Xuancheng, Wang, Zihao, and Zhou, Yang
- Subjects
- *
GRAPH neural networks , *VERNACULAR architecture , *CAUSAL inference , *LEARNING strategies , *NEIGHBORHOODS - Abstract
Graph neural networks (GNNs) have significantly advanced the processing of graph-structured data, where objects exhibit complex relationships and interdependencies. The graph convolutional network (GCN), as a representative technology, enables end-to-end learning of such data. However, as GNN technology continues to evolve, certain entrenched research paradigms have created bottlenecks in further development. A key issue arises from the common practice of predefining the graph structure, such as fixing the node degrees before learning the underlying graph structure. While this approach is often employed to constrain the learning process, it does not guarantee the optimal discovery of the graph's potential structure. Specifically, the fixed node degree can limit the adaptability of the neighborhood, thereby influencing the model's performance. In this paper, we provide an in-depth analysis of this limitation. From a theoretical perspective, we rigorously examine the constraints of traditional GNN architectures and highlight the importance of considering the dynamic relationship between input features and node degrees. Furthermore, we propose an optimization strategy for GNN learning architectures, utilizing causal inference techniques, and introduce an enhanced model, termed Causality-guided Graph Neural Network (C-GNN). Our theoretical contributions are supported by experimental validation, where comprehensive quantitative and qualitative evaluations demonstrate the superiority of the C-GNN model over traditional GNN architectures. • Consider input-feature correlation with node degree to enhance GNN performance. • A causality-guided GNN architecture with self-adaptation is presented. • Theoretical analyses and experiments validate the sophistication and effectiveness of C-GNN. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
15. Uncertainty quantification in Bayesian inverse problems with neutron and gamma time correlation measurements.
- Author
-
Lartaud, Paul, Humbert, Philippe, and Garnier, Josselin
- Subjects
- *
NUCLEAR reactor cores , *INVERSE problems , *LEARNING strategies , *NEUTRONS , *TIME measurements - Abstract
Neutron noise analysis is a predominant technique for fissile matter identification with passive methods. Quantifying the uncertainties associated with the estimated nuclear parameters is crucial for decision-making. A conservative uncertainty quantification procedure is possible by solving a Bayesian inverse problem with the help of statistical surrogate models but generally leads to large uncertainties due to the surrogate models' errors. In this work, we develop two methods for robust uncertainty quantification in neutron and gamma noise analysis based on the resolution of Bayesian inverse problems. We show that the uncertainties can be reduced by including information on gamma correlations. The investigation of a joint analysis of the neutron and gamma observations is also conducted with the help of active learning strategies to fine-tune surrogate models. We test our methods on a model of the SILENE reactor core, using simulated and real-world measurements. • Two methods for uncertainty quantification in neutron and gamma noise analysis. • Inclusion of gamma correlation measurements for matter identification. • Design strategy for fine-tuning high-dimensional surrogate models. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
16. Predictor-based event-triggered learning control of networked control systems with false data injection attacks and output delay.
- Author
-
Yang, Meng and Zhai, Junyong
- Subjects
- *
LEARNING strategies , *LYAPUNOV functions , *SIMULATION methods & models , *BANDWIDTHS , *ALGORITHMS - Abstract
This article is concerned with the predictor-based event-triggered learning control of networked control systems (NCSs) with false data injection attacks (FDIAs) and output delay. Firstly, by applying the prediction method, a new state observer including an output predictor is employed to get the estimation of delayed sampled-data output in the context of sampling. To improve the efficiency of limited networked resources, an intelligent periodic event-triggered scheme (PETS) is established, in which the triggered threshold can be optimized by the asynchronous advantage actor-critic (A3C) algorithm. Then, a predictor-based event-triggered learning control strategy is developed to handle the FDIAs occurring in the controller-to-actuator channel, and the neural network (NN) technique is introduced to approximate the false data. By applying the Lyapunov function, some sufficient conditions are given to guarantee the boundedness of the NCSs. At last, a simulation of a satellite system is given to confirm the superiorities of the presented predictor-based learning control strategy. • This article addresses the predictor-based event-triggered learning control problem. • Using predictor-based observer to actively compensate the impact of output delay. • Using intelligent periodic event-triggered scheme to save network bandwidth. • Using neural network technique to approximate false data affected by FDIAs. • Some sufficient conditions can ensure the boundedness of NCSs. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
17. Intelligent strategy for severity scoring of skin diseases based on clinical decision-making thinking with lesion-aware transformer: Intelligent strategy for severity scoring of skin diseases: K. Huang et al.
- Author
-
Huang, Kai, Sun, Kai, Li, Jiayi, Wu, Zhe, Wu, Xian, Duan, Yuping, Chen, Xiang, and Zhao, Shuang
- Subjects
SKIN disease diagnosis ,SKIN diseases ,ATOPIC dermatitis ,LEARNING strategies ,DERMATOLOGISTS - Abstract
Skin diseases are numerous in types and high in incidence, posing a serious threat to human health. Accurately assessing the severity of skin diseases helps dermatologists in making personalized treatment decisions. However, focusing solely on the skin lesion itself and ignoring the true state of the surrounding skin can lead to distorted results. Assessing the severity of the condition should be a holistic process. Specifically, dermatologists need to compare the abnormal skin with surrounding skin to conduct the diagnosis. To imitate such diagnosis practice of dermatologists, we propose LSATrans, a Transformer based framework customized for severity scoring of skin diseases. Different from the Standard Self-Attention module, we propose the Lesion-aware Self-Attention (LSA) module. LSA can capture the visual features of both lesion and normal surrounding skin areas and include their relationship in modeling. In addition to LSA, the proposed LSATrans also introduces a contrastive learning strategy for further optimization. We first evaluated the performance of LSATrans in scar, atopic dermatitis, and psoriasis scoring tasks, and it achieved mean absolute errors of 0.5895, 0.5614, and 0.5416 respectively in these three tasks. Furthermore, we conducted additional validation of LSATrans's performance in two distinct skin disease diagnosis tasks, where it demonstrated remarkable outcomes with AUCs of 0.9774 and 0.9801, respectively, in the classification of common skin diseases and subtypes of skin diseases. These results are better than existing methods, indicating that LSATrans is expected to become a universal, accurate and objective intelligent tool for scoring the severity of skin diseases. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
18. Cross-dataset motor imagery decoding — A transfer learning assisted graph convolutional network approach.
- Author
-
Zhang, Jiayang, Li, Kang, Yang, Banghua, and Zhao, Zhengrun
- Subjects
MOTOR imagery (Cognition) ,SIGNAL-to-noise ratio ,LEARNING strategies ,ELECTRODES ,ELECTROENCEPHALOGRAPHY - Abstract
The proliferation of portable electroencephalogram (EEG) recording devices has made it practically feasible to develop the motor imagery (MI) based brain–computer interfaces (BCIs). However, the low signal-to-noise ratio of EEG signals for abstract MI tasks, limited data, limited EEG channels, and strong inter- and intra-subject variability pose significant challenges for MI-task recognition. This paper proposes a transfer learning assisted graph convolutional network (GCN) modeling approach for cross-dataset MI decoding, one of the most challenging issues in this field. In the experiments, a multi-channel dataset with 62 electrodes and a few-channel dataset with 8 electrodes are utilized for cross-dataset modeling. To harness multi-channel information, we utilize the GCN module to aggregate topological features. The pre-trained model is guided with few-channel signals as inputs through a knowledge distillation framework. Subsequently, the pre-trained model is adapted to the few-channel dataset using a transfer learning strategy with minimal data training. Experiment results show that the proposed model achieves 3.92% and 3.83% more accuracy improvement compared with state-of-the-art models in the cross-validation and cross-session scenario respectively, demonstrating the effectiveness of the proposed approach in cross-dataset MI-EEG decoding, thus enabling more effective MI-BCI applications. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
19. VVBPNet: Deep learning model in view-by-view backprojection (VVBP) domain for sparse-view CBCT reconstruction.
- Author
-
Zhao, Xuzhi, Du, Yi, and Peng, Yahui
- Subjects
CONE beam computed tomography ,SIGNAL-to-noise ratio ,LEARNING strategies ,DEEP learning - Abstract
• This is the first study on CBCT reconstruction in the VVBP domain. • VVBPNet is the first DL model for VVBP domain-based CBCT reconstruction. • VVBPNet is tested on 1/4, 1/8, and 1/16 sparse-view reconstruction tasks. • VVBPNet outperforms 14 SOTA models for the 1/8 and 1/16 tasks. • VVBPNet's effectiveness increases as the projection views become sparser. This study proposes a deep learning model in the view-by-view backprojection (VVBP) domain, named VVBPNet, to improve the quality of sparse-view cone-beam computed tomography (CBCT) images. The VVBPNet model adopted a content-noise complementary learning strategy, featuring two parallel attention Res-UNet sub-networks and a fusion mechanism. It processed VVBP-Tensors, which were intermediate results generated during the execution of the Feldkamp-Davis-Kress algorithm, to produce denoised and artifact-reduced axial images. The model was trained, validated, and tested on CBCT data from 163, 30, and 30 real patients, respectively. Quantitative metrics including root-mean-square error (RMSE), peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and feature similarity (FSIM) were calculated to evaluate the model performance. The VVBPNet model was compared with 3 state-of-the-art (SOTA) projection domain models and 11 SOTA image domain models on three reconstruction tasks with varying sparsity levels: 1/4 moderate-sparse-view, 1/8 high-sparse-view, and 1/16 ultra-sparse-view. For the 1/4 task, VVBPNet showed differences of −0.00001 for RMSE, +0.1 dB for PSNR, −0.002 for SSIM, and –0.003 for FSIM compared to the best metrics from all 14 SOTA models. For the 1/8 and 1/16 tasks, VVBPNet demonstrated consistent improvements over all 14 SOTA models, with differences of −0.00003 and −0.00009 for RMSE, +0.3 dB and +0.6 dB for PSNR, +0.005 and +0.009 for SSIM, and +0.001 and +0.008 for FSIM, respectively. The proposed VVBPNet model in the VVBP domain effectively improves the quality of sparse-view CBCT images. As the projection views become sparser, the VVBPNet model exhibits greater performance advantages over the SOTA models. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
20. Mediating learning with mobile devices through pedagogical innovation: Teachers' perceptions of K-12 students' learning experiences.
- Author
-
Burke, Paul F., Schuck, Sandy, Burden, Kevin, and Kearney, Matthew
- Subjects
- *
DIGITAL technology , *STRUCTURAL equation modeling , *INNOVATION adoption , *RESEARCH personnel , *LEARNING strategies , *MOBILE learning - Abstract
This article presents a quantitative study demonstrating that digital pedagogy using mobile devices (e.g., laptops, mobile phones, tablets) impacts teachers' perceptions of pedagogical innovation. We further find that innovative mobile pedagogy adoption positively impacts teachers' perceptions of improved student learning in K-12 settings. Examining teachers' perceptions of pedagogical innovation in digital practice is important as researchers have questioned whether the use of digital devices constitutes an innovative break with traditional pedagogical practice or serves merely as a digital replica of non-digital practices. Further, providing a link between pedagogical innovation and perceived improvements in student learning provides further support for removing barriers to innovation adoption. The study uses the validated iPAC Framework (referring to Personalisation, Authenticity and Collaboration in mobile teaching pedagogies) as the basis of an international survey. The survey measured teachers' perceptions of their adopted pedagogies using mobile technologies during students' completion of a digital task, whether they viewed these practices as innovative, and how such approaches impacted teachers' perceptions of their students' learning experiences. Results from a Structural Equation Model (SEM) demonstrate that when teachers adopt innovative pedagogical tasks into their teaching with digital technologies, they perceive an improvement in student learning experiences. This is the first study that considers both the contribution of digital pedagogies and the contribution of innovation as direct and indirect effects on teachers' perceptions of student learning experiences. • Teachers indicate innovation occurs when their mobile learning tasks include personalisation, authenticity, and co-creation. • Teachers adopting innovative m-learning pedagogies leads to their perceived improvement of student learning experiences. • Support for teachers to develop and increase their adoption of innovative digital pedagogies should be encouraged. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
21. The effects of different metacognitive patterns on students' self-regulated learning in blended learning.
- Author
-
Geng, Xingyu and Su, Yu-Sheng
- Subjects
- *
COMPUTER assisted instruction , *SELF-regulated learning , *BLENDED learning , *LEARNING strategies , *METACOGNITION , *CLUSTER analysis (Statistics) - Abstract
Self-regulated learning has significant importance in blended learning, necessitating an exploration into the effects of metacognition on SRL. Furthermore, SRL exhibits interdependence, thus highlighting the urgent need for research that can capture the temporal processes of SRL in multi-task activities during blended learning. Over 18 weeks, 44 students participated in three SRL tasks designed for blended learning. Students completed a questionnaire assessing their metacognitive awareness at the end of the course. A two-step cluster analysis was employed to explore different metacognitive patterns among students: high metacognitive knowledge and regulation students and low metacognitive knowledge and regulation students. Furthermore, data from student's learning diaries were collected and coded based on the SRL model specific to blended learning. Epistemic Network Analysis, a computer-assisted learning analysis method, was employed to investigate the SRL process of students with different metacognitive patterns in various tasks of blended learning. First, the findings indicate that both types of students strive for performance achievement; however, students with high metacognitive knowledge and regulation primarily employ task strategies, while those with low metacognitive knowledge and regulation focus on time management. Moreover, a detailed centroid analysis conducted for each task revealed that students with high metacognitive knowledge and regulation initiate their SRL process in computer-assisted blended learning with self-efficacy and conclude it with goal setting. Conversely, students with low metacognitive knowledge and regulation commence their SRL process with self-efficacy and conclude it by employing strategies for self-evaluation. Finally, implications, limitations, and future research directions are discussed. • Metacognition plays a crucial role in SRL for computer-assisted blended learning. • Metacognition includes two interconnected components: regulation and knowledge. • Metacognition influences the distribution of SRL elements in blended learning. • Metacognition influences the SRL process pattern in blended learning. • It is crucial for the development of computer-assisted instruction. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
22. Effectiveness of gamified intelligent tutoring in physical education through the lens of self-determination theory.
- Author
-
Hsia, Lu-Ho, Lin, Yen-Nan, Lin, Chung-Hisenh, and Hwang, Gwo-Jen
- Subjects
- *
SELF-determination theory , *INTERACTIVE learning , *ANALYSIS of covariance , *LEARNING strategies , *ACADEMIC motivation - Abstract
Scholars have recommended the application of an intelligent tutoring and instant feedback system (ITIFS) to enhance students' motor skills performance by automatically evaluating their learning performance and providing personalized guidance and feedback. However, solely providing personalized evaluation and feedback may not necessarily attract students' active and sustained engagement in practice. In particular, it is difficult to arouse students' enthusiasm to participate in sports that require repetitive practice to improve their physical abilities and which involve less interaction with the environment and their opponents. To address this issue, grounded in self-determination theory (SDT), the present study integrated a gamification mechanism that aligned with students' psychological needs into an ITIFS. The gamification features included avatars, achievements (personal ratings and rankings), badges, levels, and social networks (group ratings and rankings). It aimed to attract students to engage continuously in practice, and to address the issue of students lacking motivation to engage in repeated practice. To investigate the effectiveness of the proposed method, a quasi-experimental research design was adopted, and the collected data were analyzed with analysis of covariance (ANCOVA), independent samples t tests and qualitative coding. Four classes of university students participated in the experiment. Two classes (N = 80) were the experimental group adopting the SDT-based gamified ITIFS (G-ITIFS), and the other two classes (N = 76) were the control group adopting the conventional ITIFS (C-ITIFS). The findings indicated that the experimental group showed significantly better yoga skills performance and learning engagement compared to the control group. Feedback from students also revealed that the gamification mechanism provided more excitement and had positive impacts, satisfying students' psychological needs and reinforcing the learning benefits. The findings of the present study revealed that, from the perspective of SDT, incorporating gamification elements into the development of ITIFS could be a promising approach for physical education. Therefore, it is strongly encouraged that educators promote such a gamified intelligent tutoring mode in physical education curriculums as it is crucial to the development of students' physical and mental health, as well as to their enthusiasm to participate in sports. • A Gamified intelligent tutoring and instant feedback system is proposed. • The system is proposed based on self-determination theory. • An experiment was conducted in a university yoga course. • The system improved students' yoga skills performance and learning engagement. • The system improved students' autonomy, competence, and relatedness. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
23. Self-supervised contrast learning based UAV fault detection and interpretation with spatial–temporal information of multivariate flight data.
- Author
-
Wang, Shengdong, Jia, Zhen, Liu, Zhenbao, Tang, Yong, Qin, Xinshang, and Wang, Xiao
- Subjects
- *
GRAPH neural networks , *DRONE aircraft , *SUPERVISED learning , *LEARNING strategies , *GAUSSIAN distribution - Abstract
Precise fault detection and interpretation can effectively enhance the safety of unmanned aerial vehicles (UAV) flight missions. However, sufficient fault data covering all fault modes is generally inaccessible, which poses a formidable challenge to the traditional supervised learning strategy. In this study, a novel UAV fault detection approach based on self-supervised contrast learning and spatial–temporal information of multivariate flight data is proposed. In the contrast learning task, a series of specific sample transformations are first designed, and the feature distribution of normal data can be modeled in self-supervised manner through comparing the similarity of different transformed samples. In above process, an auxiliary classification task that distinguishes different sample transformations is further introduced to facilitate the learning of critical features. In order to extract comprehensive spatial–temporal information from multivariate flight data, a multi-channel spatial–temporal encoder is designed in which two independent graph multi-head attention neural networks (GMAT) are implemented to mine the temporal features and multivariate spatial features, respectively. The extracted spatial–temporal features are then fused with the designed locally-enhanced token fusion module and the powerful multi-headed self-attention module. Finally, the occurrence of faults can be detected by comparing the reconstruction error with the fault threshold. With the box-plot analysis, the flight variables whose reconstruction errors are far from the overall distribution will be regarded as the possible fault sources to implement fault interpretation. Experimental results on the self-developed fixed-wing UAVs demonstrated the prominent performance of our method on fault detection and interpretation. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
24. Semi-supervised suppressed possibilistic Gustafsan-Kessel clustering algorithm based on local information and knowledge propagation.
- Author
-
Yu, Haiyan, Liu, Junnan, and Gong, Kaiming
- Subjects
- *
CLUSTERING algorithms , *IMAGE segmentation , *COVARIANCE matrices , *LOCAL knowledge , *LEARNING strategies - Abstract
Traditional clustering algorithms always suffer from tough issues in the clustering of complex multi-dimensional data with multiple characteristics, such as significantly imbalanced sizes, strong feature correlation, and noisy corruption. Semi-supervised possibilistic c-means clustering (SS-PCM) algorithm can alleviate center offset to some extent for imbalanced datasets by introducing prior information into possibilistic clustering. However, the SS-PCM is still incapable of handling the above-mentioned complex data with multiple characteristics. Thus, a semi-supervised suppressed possibilistic Gustafon-Kessel algorithm based on local information and knowledge propagation (SS-S-PLIGK) is proposed. Firstly, an improved Mahalanobis distance by adding prior information into its covariance matrix is introduced to improve the clustering performance of ellipsoidal data with feature correlation and imbalanced sizes. Secondly, a new local factor G ki based on Mahalanobis distance which contains the shape distribution is designed to enhance the anti-noise robustness of ellipsoid data. Thirdly, two schemes of neighborhood knowledge propagation are designed to expand the impact of prior information. Finally, a suppressed competitive learning strategy is improved based on shadow sets to overcome center overlapping of SS-PCM and decrease the iteration number. Meanwhile, an adaptively determination method of the suppressed rate is presented based on the local factor G ki to further enhance the anti-noise robustness. Experiments conducted on synthetic data, real data, and color images with imbalanced sizes and strong noise show that the proposed SS-S-PLIGK achieves higher accuracy and efficiency than several state-of-the-art semi-supervised clustering and unsupervised clustering algorithms. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
25. Multi-layer multi-level comprehensive learning for deep multi-view clustering.
- Author
-
Chen, Zhe, Wu, Xiao-Jun, Xu, Tianyang, Li, Hui, and Kittler, Josef
- Subjects
- *
LEARNING strategies , *INFORMATION sharing - Abstract
Multi-view clustering has attracted widespread attention because of its capability to identify the common semantics shared by the data captured from different views of data, objects or phenomena. This is a challenging problem but with the emergence of deep auto-encoder networks, the performance of multi-view clustering methods has considerably improved. However, it is notable that most existing methods merely utilize the features outputted by the last encoder layer to carry out the clustering task. Such approach neglects potentially useful information conveyed by the features of the previous layers. To address the this problem, we propose a novel m ulti-layer m ulti-level comprehensive learning framework for deep m ulti-view c lustering (3MC). 3MC firstly conducts a contrastive learning involving different views based on deep features in each encoder layer separately, so as to achieve multi-view feature consistency. The next step is to construct layer-specific label MLPs to transform the features in each layer to high-level semantic labels. Finally, 3MC conducts an inter-layer contrastive learning using the high-level semantic labels in order to obtain multi-layer consistent clustering assignments. We demonstrate that the proposed comprehensive learning strategy, commencing from layer specific inter-view feature comparison to inter-layer high-level label comparison extracts and utilizes the underlying multi-view complementary information very successfully and achieves more accurate clustering. An extensive experimental comparison with the state-of-the-art methods demonstrates the effectiveness of the proposed framework. The code of this paper is available at https://github.com/chenzhe207/3MC. • A novel multi-layer feature learning method is designed to solve deep MvC problem. • A double contrastive learning strategy is proposed to realize multi-layer learning. • Detailed ablation study demonstrates the effectiveness of multi-layer learning. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
26. QTypeMix: Enhancing multi-agent cooperative strategies through heterogeneous and homogeneous value decomposition.
- Author
-
Fu, Songchen, Zhao, Shaojing, Li, Ta, and Yan, Yonghong
- Subjects
- *
MULTIAGENT systems , *LEARNING strategies , *MARKOV processes , *TASK performance , *DEEP learning - Abstract
In multi-agent cooperative tasks, the presence of heterogeneous agents is familiar. Compared to cooperation among homogeneous agents, collaboration requires considering the best-suited sub-tasks for each agent. However, the operation of multi-agent systems often involves a large amount of complex interaction information, making it more challenging to learn heterogeneous strategies. Related multi-agent reinforcement learning methods sometimes use grouping mechanisms to form smaller cooperative groups or leverage prior domain knowledge to learn strategies for different roles. In contrast, agents should learn deeper role features without relying on additional information. Therefore, we propose QTypeMix, which divides the value decomposition process into homogeneous and heterogeneous stages. QTypeMix learns to extract type features from local historical observations through the TE loss. In addition, we introduce advanced network structures containing attention mechanisms and hypernets to enhance the representation capability and achieve the value decomposition process. The results of testing the proposed method on 14 maps from SMAC and SMACv2 show that QTypeMix achieves state-of-the-art performance in tasks of varying difficulty. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
27. Dynamic graph consistency and self-contrast learning for semi-supervised medical image segmentation.
- Author
-
Li, Gang, Xie, Jinjie, Zhang, Ling, Cheng, Guijuan, Zhang, Kairu, and Bai, Mingqi
- Subjects
- *
SUPERVISED learning , *DIAGNOSTIC imaging , *LEARNING strategies , *PIXELS , *CORPORA , *IMAGE segmentation - Abstract
Semi-supervised medical image segmentation endeavors to exploit a limited set of labeled data in conjunction with a substantial corpus of unlabeled data, with the aim of training models that can match or even exceed the efficacy of fully supervised segmentation models. Despite the potential of this approach, most existing semi-supervised medical image segmentation techniques that employ consistency regularization predominantly focus on spatial consistency at the image level, often neglecting the crucial role of feature-level channel information. To address this limitation, we propose an innovative method that integrates graph convolutional networks with a consistency regularization framework to develop a dynamic graph consistency approach. This method imposes channel-level constraints across different decoders by leveraging high-level features within the network. Furthermore, we introduce a novel self-contrast learning strategy, which performs image-level comparison within the same batch and engages in pixel-level contrast learning based on pixel positions. This approach effectively overcomes traditional contrast learning challenges related to identifying positive and negative samples, reduces computational resource consumption, and significantly improves model performance. Our experimental evaluation on three distinct medical image segmentation datasets indicates that the proposed method demonstrates superior performance across a variety of test scenarios. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
28. Noise-robust consistency regularization for semi-supervised semantic segmentation.
- Author
-
Zhang, HaiKuan, Li, Haitao, Zhang, Xiufeng, Yang, Guanyu, Li, Atao, Du, Weisheng, Xue, Shanshan, and Liu, Chi
- Subjects
- *
SUPERVISED learning , *LEARNING strategies , *NOISE - Abstract
The essential of semi-supervised semantic segmentation (SSSS) is to learn more helpful information from unlabeled data, which can be achieved by assigning adequate quality pseudo-labels or managing noisy pseudo-labels during training. However, most relevant state-of-the-art (SOTA) methods are mainly devoted to improving one aspect. By revisiting the representative SSSS methods from a robust learning view, this paper discovers that the appropriate combination of multiple noise-robust methods contributes both to assigning sufficient quality pseudo labels and managing noisy labels. Therefore, from five different noise management perspectives, we summarize the reasons why noise-robust techniques can successfully harvest performance gains in SSSS. Subsequently, we present a novel feature perturbation method, multi-view learning strategy, and robust loss function to exploit the advantages of different noise-robust techniques. The outcome of this paper is a new SSSS approach with noise-robust consistency regularization called NRCR that can simultaneously produce adequate quality pseudo-labels and manage noisy pseudo-labels. Abundant experiments on public benchmarks demonstrate the performance superiority of our approach compared with previous SOTA methods and the correctness of our analytical viewpoints. Code is available at https://github.com/zhanghk1996/NRCR. • The first work revisiting semi-supervised semantic segmentation from a robust learning view. • Three novel noise-robust techniques for semi-supervised semantic segmentation. • A novel semi-supervised semantic segmentation approach with noise-robust consistency regularization. • The presented method outperform state-of-the-art methods on different benchmarks. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
29. Dual Contrastive Label Enhancement.
- Author
-
Guan, Ren, Wang, Yifei, Liu, Xinyuan, Chen, Bin, and Zhu, Jihua
- Subjects
- *
LEARNING strategies , *LOW density lipoproteins - Abstract
Label Enhancement (LE) strives to convert logical labels of instances into label distributions to provide data preparation for label distribution learning (LDL). Existing LE methods ordinarily neglect to consider original features and logical labels as two complementary descriptive views of instances for extracting implicit related information across views, resulting in insufficient utilization of the feature and logical label information of the instances. To address this issue, we propose a novel method named Dual Contrastive Label Enhancement (DCLE). This method regards original features and logical labels as two view-specific descriptions and encodes them into a unified projection space. We employ dual contrastive learning strategy at both instance-level and class-level to excavate cross-view consensus information and distinguish instance representations by exploring inherent correlations among features, thereby generating high-level representations of the instances. Subsequently, to recover label distributions from obtained high-level representations, we design a distance-minimized and margin-penalized training strategy and preserve the consistency of label attributes. Extensive experiments conducted on 13 benchmark datasets of LDL validate the efficacy and competitiveness of DCLE. • Unify features and logical labels as dual-view descriptions in a projection space. • Dual contrastive learning is used to obtain high-level representations for LE. • Consider label consistency to guide recovering label distributions. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
30. Count, decompose and correct: A new approach to handwritten Chinese character error correction.
- Author
-
Hu, Pengfei, Ma, Jiefeng, Zhang, Zhenrong, Du, Jun, and Zhang, Jianshu
- Subjects
- *
CHINESE characters , *LEARNING strategies , *ANNOTATIONS , *SPELLING errors , *GENERALIZATION - Abstract
Recently, handwritten Chinese character error correction has been greatly improved by employing encoder–decoder methods to decompose a Chinese character into an ideographic description sequence (IDS). However, existing methods implicitly capture and encode linguistic information inherent in IDS sequences, leading to a tendency to generate IDS sequences that match seen characters. This poses a challenge when dealing with an unseen misspelled character, as the decoder may generate an IDS sequence that matches a seen character instead. Therefore, we introduce Count, Decompose and Correct (CDC), a novel approach that exhibits better generalization towards unseen misspelled characters. CDC is mainly composed of three parts: the Counter, the Decomposer, and the Corrector. In the first stage, the Counter predicts the number of each radical class without the symbol-level position annotations. In the second stage, the Decomposer employs the counting information and generates the IDS sequence step by step. Moreover, by updating the counting information at each time step, the Decomposer becomes aware of the existence of each radical. With the decomposed IDS sequence, we can determine whether the given character is misspelled. If it is misspelled, the Corrector under the transductive transfer learning strategy predicts the ideal character that the user originally intended to write. We integrate our method into existing encoder–decoder models and significantly enhance their performance. • We spot the impact of linguistic information on Chinese character error correction. • We introduce a 3-step method named Count, Decompose and Correct. • Our method can be generalized to various encoder–decoder models. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
31. Time and frequency synergy for source-free time-series domain adaptations.
- Author
-
Furqon, Muhammad Tanzil, Pratama, Mahardhika, Shiddiqi, Ary, Liu, Lin, Habibullah, Habibullah, and Dogancay, Kutluyil
- Subjects
- *
TIME series analysis , *LEARNING strategies , *GENERALIZATION , *PRIVACY , *CURRICULUM - Abstract
Notwithstanding that source-free domain adaptation (SFDA) has gained its prominence due to privacy protections of source samples in domain adaptations, very few works have been devoted to address time-series problems possessing temporal characteristics. Besides, existing works of source-free time-series domain adaptation (SFTSDA) have not exploited the potential of frequency features offering complementary information to boost the performances. This paper proposes time frequency domain adaptation (TFDA) to overcome the SFTSDA problems. TFDA fully utilizes time and frequency information via a dual branch network structure comprising time and frequency encoders. Consistencies of time and frequency domains are forced via the contrastive learning strategies in the time, frequency and time-frequency domains while applying the self-distillation concept to maintain the same. To further improve the performance, TFDA implements the uncertainty reduction strategy to combat the issue of domain shift and the curriculum learning strategy to deal with the noisy pseudo labels. Rigorous experiments with 3 time series datasets of different application domains confirm the advantage of TFDA over prior arts with noticeable margins. In addition, the theoretical analysis is provided to show the generalization bound of our approach. • This paper proposes Time Frequency Domain Adaptation (TFDA) to address the source free time-series domain adaptation problem. • TFDA puts forward the concept of a dual-branch network structure mining both temporal and frequency features. • TFDA proposes the idea of time and frequency consistencies for the source free time-series domain adaptation problem. • Rigorous experiments have been performed where TFDA demonstrate the most encouraging performance over the prior arts. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
32. The deep continual learning framework for prediction of blast-induced overbreak in tunnel construction.
- Author
-
He, Biao, Li, Jialu, Armaghani, Danial Jahed, Hashim, Huzaifa, He, Xuzhen, Pradhan, Biswajeet, and Sheng, Daichao
- Subjects
- *
TUNNEL design & construction , *LEARNING strategies , *PREDICTION models , *BLASTING , *LEARNING ability , *DEEP learning - Abstract
Blast-induced overbreak, characterized by the excessive removal of rock mass beyond the planned tunnel profile, poses significant safety risks, increases costs, and causes project delays during tunneling. Traditional machine/deep learning models have been developed to predict overbreak. However, these models are often inadequate because they are static and lack the flexibility to adapt to new, real-world data continuously. This study addresses this limitation by introducing a novel data-driven approach based on deep continual learning. The primary objective is to develop an adaptable predictive model with the ability of continual learning, which is particularly advantageous in dynamic environments like tunnel blasting. To achieve this, a self-attention multi-layer perceptron (MLP) model for overbreak prediction, integrated with two continual learning strategies (elastic weight consolidation (EWC) and memory replay (MR)), is developed. This step enables the overbreak prediction model to possess the ability to continuously learn real-world scenarios and adapt to the dynamic environment of tunnel blasting. The findings show that the continuous MLP model, empowered by EWC and MR, demonstrates superior adaptability and accuracy in predicting overbreak. Compared with the standard MLP model, which achieves a predictive accuracy of 0.831, the continuous MLP model achieves a predictive accuracy of 0.845 on unseen data. The integration of EWC and MR strategies proves to be a pivotal factor in developing deep learning models for the dynamic task of predicting overbreak. The continual learning strategies ensure that the models remain adaptable and accurate over time, which is essential for practical applications in dynamic environments of tunnel blasting operations. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
33. CCL-MPC: Semi-supervised medical image segmentation via collaborative intra-inter contrastive learning and multi-perspective consistency.
- Author
-
Du, Xiaogang, Zou, Yibin, Lei, Tao, Zhang, Weichuan, Wang, Yingbo, and Nandi, Asoke K.
- Subjects
- *
IMAGE segmentation , *COMPUTER-assisted image analysis (Medicine) , *DEEP learning , *LEARNING strategies , *SOURCE code - Abstract
Semi-supervised image segmentation extracts specific regions and tissues by utilizing extensive unlabeled images and limited labeled images, which can alleviate the dependence on plenty of accurately labeled data. However, it is difficult to learn robust feature representations due to the potential noise in pseudo-labels caused by inefficient consistency learning, and poor class diversity in feature spaces. To address this issue, we propose a semi-supervised medical image segmentation method using Collaborative intra-inter Contrastive Learning and Multi-Perspective Consistency (CCL-MPC). First, we propose a collaborative intra-inter contrastive learning strategy that includes symmetric bidirectional contrastive learning and certainty-guided contrastive learning, to exploit the intrinsic differences between inter-image and intra-image feature representations. Second, we design a multi-perspective consistency learning strategy to improve the class diversity by utilizing a dual-branch network and two augmented views. Additionally, we dynamically partition the pseudo-label certainty area for auxiliary consistency learning to reduce the potential noise during the training process. Experimental results on the publicly available datasets demonstrate that CCL-MPC can achieve better segmentation performance than the state-of-the-art methods for semi-supervised medical image segmentation tasks. The source code is available at https://github.com/EmarkZOU/CCL-MPC. • Compare the pixels within the same category across various images to discern their analogous intrinsic characteristics. • Enrich positive and negative samples by cross sampling from labeled and unlabeled images. • Leverage certainty pseudo-labels for enhancing the reliability of consistency supervision. • Use the feature differences inside and outside the image can enable more comprehensive representation learning. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
34. TNPNet: An approach to Few-shot open-set recognition via contextual transductive learning.
- Author
-
Wu, Shaoling, Luo, Huilan, and Lin, Xiaoming
- Subjects
- *
IMAGE recognition (Computer vision) , *SUPERVISED learning , *CONTEXTUAL learning , *LEARNING strategies , *PROTOTYPES - Abstract
In the field of Few-Shot Open-Set Recognition (FSOSR), identifying images from known and unknown categories with limited data is a pressing challenge. Existing approaches often falter due to complex learning strategies or reliance on pseudo-negative sample features. To address this, we introduce the Transductive Negative Prototype learning Network (TNPNet), a straightforward yet effective architecture for FSOSR. Leveraging both local and global context, our novel Contextual Transductive Learning (CTL) method enhances prototypes for both known and unknown categories. TNPNet also features a parameter-free classifier designed to optimize CTL-enhanced prototypes and introduces a multi-line loss for robust supervision. Through extensive evaluations, TNPNet demonstrates superior AUROC metrics across multiple scenarios and datasets, affirming its robustness and practicality in real-world applications. Code is available at: https://github.com/wushaoling540/TNPNet/ • Contextual transductive learning for robust negative prototypes. • RPAF Classifier to enhance CTL-enhanced prototype performance. • Multi-line loss to fortify model's supervisory capabilities. • TNPNet's excellent validation across three FSOSR benchmark datasets. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
35. Category-integrated Dual-Task Graph Neural Networks for session-based recommendation.
- Author
-
Ding, Yuhan, Zhang, Zizhuo, and Wang, Bang
- Subjects
- *
GRAPH neural networks , *RECOMMENDER systems , *LEARNING strategies , *TIME measurements - Abstract
Session-based recommendation (SBR) aims to predict the subsequent item a user may be interested in based on their behavior within a limited timeframe. Most existing approaches primarily harness item relations and overlook the significance of attribute information (e.g. category). Users' interests in specific items could change frequently within a single session, yet may exhibit more stability at the category level. We argue that integrating category information into SBR models can help mitigating data sparsity challenges for promoting next-item prediction. In this paper, we propose a novel SBR methodology named C ategory-integrated D ual- T ask G raph N eural N etworks (CDT-GNN). It constructs a heterogeneous global graph encompassing all sessions and individual heterogeneous local graphs for each session to learn items' and categories' representations. A dual-task learning strategy is employed to involve next-category prediction which serves as an auxiliary task to bolster the major task of next-item prediction. Additionally, a user-selectable accessory feature is developed to enhance the utilization of the predicted category. Extensive experimental results on three real-world datasets validate the effectiveness of CDT-GNN. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
36. Toward the ensemble consistency: Condition-driven ensemble balance representation learning and nonstationary anomaly detection for battery energy storage system.
- Author
-
Yang, Jiayang, Chen, Xu, and Zhao, Chunhui
- Subjects
- *
BATTERY storage plants , *LEARNING strategies , *CELL analysis - Abstract
In the battery energy storage systems (BESS), multiple lithium-ion battery (LIB) cells are consolidated into a LIB module for scalable management. Normally, LIB cells within the same module are deemed to exhibit consistency acting as an ensemble. For the reliable monitoring of LIB cells, it is considerably challenging to capture the overall working status of LIB cells meanwhile maintaining the awareness of the consistency among each cell. Additionally, the nonstationary characteristics of LIB cells arising from charging, discharging, and other behaviors pose more difficulties for anomaly detection. In this study, we propose a condition-driven ensemble balance representation learning and anomaly detection method to address those challenges, introducing the concept of ensemble analysis for the first time in the field of LIB anomaly detection. Specifically, an ensemble balance representation learning strategy is developed for LIB cells, primarily consisting of two aspects. First, a dual-layer health (DLH) feature learning approach is proposed to provide a representation of the status of LIB cells, which considers LIB cell's operation characteristics and the interaction with others. Second, an ensemble balance component analysis (EBCA) method is designed for DLH features to uncover the inherent balance relationship between LIB cells. This approach allows us to monitor the overall working status of LIB cells within the module while maintaining sensitivity to detecting individual LIB cell anomaly. Further, considering the influence of nonstationary characteristics, we develop a condition-driven mode partition strategy to uncover multiple condition modes from the nonstationary operation process of the LIB cells, where the EBCA model is established for each mode. The effectiveness of the proposed method is demonstrated through real operation processes of LIB cells in a BESS. • A novel concept of ensemble analysis is introduced for anomaly detection of LIB. • The inherent balance relationship among LIB cells is revealed. • An ensemble balance representation learning strategy is developed. • The nonstationary operation process of LIB cells is tackled. • Multiple statistics are designed to provide explainable anomaly detection results. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
37. PDCA-Net: Parallel dual-channel attention network for polyp segmentation.
- Author
-
Chen, Gang, Zhang, Minmin, Zhu, Junmin, and Meng, Yao
- Subjects
DEEP learning ,COLORECTAL cancer ,LEARNING strategies ,POLYPS ,PIXELS - Abstract
Accurate segmentation of polyps in colonoscopy images is crucial for the diagnosis and cure of colorectal cancer. Although various deep learning methods have been proposed and have shown promising performance, accurately distinguishing between polyp and mucosal boundaries remains a challenge. In this work, we propose a Parallel Dual-Channel Attention Network (PDCA-Net) for polyp segmentation. This method utilizes the mapping transformations to adaptively encapsulate the global dependency from superpixel into pixels, enhancing the model's ability to localize foreground and background regions. Specifically, we first design a parallel spatial and channel attention fusion module to capture the global dependencies at the superpixel level from the spatial and channel dimensions. Furthermore, an adaptive associative mapping module is proposed to encapsulate the global dependencies of superpixels into each pixel through a coarse-to-fine learning strategy. Extensive experiments demonstrate that the proposed PDCA-Net effectively improves the segmentation performance and achieves new state-of-the-art results (i.e., 0.815, 0.936, 0.945, and 0.838 mDice, 0.744, 0.891, 0.900, and 0.765 mIoU on the ETIS, Kvasir-SEG, CVC-ClinicDB, and CVC-ColonDB). Our code is available at https://github.com/lzucg/PDCA-Net. • A novel parallel dual-channel attention network for polyp segmentation is designed. • A parallel attention feature fusion strategy is proposed to promote the interaction between spatial and channel features. • An adaptive associative mapping approach is introduced to encapsulate the global dependencies of superpixels into each pixel. • Extensive experiments are performed on the ETIS, Kvasir-SEG, CVC-ClinicDB, and CVC-ColonDB datasets to validate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
38. Pseudo-label-assisted subdomain adaptation network with coordinate attention for EEG-based driver drowsiness detection.
- Author
-
Feng, Xiao, Dai, Shaosheng, and Guo, Zhongyuan
- Subjects
FEATURE extraction ,TRAFFIC accidents ,DROWSINESS ,LEARNING strategies ,MODEL validation ,DEEP learning - Abstract
• We propose the first EEG feature extraction network framework with a coordinate attention mechanism, to capture spatial long-range dependencies and channel-wise relationships to efficiently augment the EEG feature representations. • We propose a subdomain adaptation method with a LMMD loss and dynamic weighted pseudo-label learning strategy, which can enhance cross-domain generalization capability for new target subjects. • We propose the interpretable convolutional coordinate attention network framework, giving an interpretable classification decision-making via the coordinate attention maps instead of black-box results. • Our proposed model is evaluated on two publicly available driver drowsiness dataset with remarkable classification performance. • Visualizing the learned attention map and feature distribution provides interpretable insights for model validation. Accurate detection of driver drowsiness using Electroencephalography (EEG) is crucial for reducing traffic accidents. Although recent deep learning-based approaches have shown promising results, two significant challenges still exist: how to explicitly model EEG features interdependencies and augment representations learning for better classification and interpretation; how to improve generalization performance on the calibration-free EEG system for new subjects in practice. To address these issues, we propose a pseudo-label-assisted subdomain adaptation network with coordinate attention (PASAN-CA) for EEG-based driver drowsiness detection. In our method, the feature extractor employs coordinate attention to enhance the discriminative features representation by effectively modeling long-range dependencies between features. For subdomain adaptation, a local maximum mean discrepancy (LMMD) is constructed to align the distribution of relevant subdomains with the same class between source and target domains, so as to learn domain-invariant discriminative features. In addition, a curriculum pseudo labeling (CPL) strategy is introduced to dynamically pick up high-quality pseudo labels of target subdomain for model training, assisting subdomain adaptation. Extensive experiments on two publicly available driver drowsiness datasets demonstrate that the proposed framework outperforms state-of-the-art baselines in overall performance. Moreover, visualizing the learned attention map and feature distribution provides interpretable insights for model validation. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
39. Rewiring Co-creation: Towards Transition Arenas with Urban Transformative Capacity.
- Author
-
Escario-Chust, Ana, Vogelzang, Fenna, Palau-Salvador, Guillermo, and Segura-Calero, Sergio
- Subjects
- *
SWARM intelligence , *CLIMATE change mitigation , *CITIES & towns , *TRANSFORMATIVE learning , *LEARNING strategies - Abstract
Cities, as significant contributors to global warming, have a crucial responsibility to implement climate mitigation measures. Addressing this complex challenge requires innovative governance and societal involvement. Transition Arenas (TAs) are seen as transformative tools for tackling such issues, but practical pathways for cities to implement them and foster their transformative capacities are often unclear. In order to provide this, this research delves into the Valencia's Energy Transition Arena in Spain, a two-year initiative that gathered stakeholders monthly to co-create the city's energy transition roadmap. The Urban Transformative Capacity framework provides the needed conceptual lens to analyze and enhance the transformative potential of this TA through its strategic, tactical, operational, and reflexive stages. It also brings a new "relational" governance level that integrates TAs into wider systems, promoting systemic alliances. The study also emphasizes the importance of autonomous processes free from external influence for innovation and engagement, structured strategies for effective collective intelligence, action-focused processes in resource-limited contexts, and formal learning strategies for systemic impact. • Introduces Urban Transformative Capacities in Transition Arenas • Urban Transformative Capacities brings strategic guidance to the Transition Arena • Introducing a relational level ensures Transition Anrenas amplify their impact and facilitate a more systemic approach • Highlights the need to foster autonomy, innovation, action and learning for transformative Transition Arenas [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
40. An integrated reinforcement learning framework for simultaneous generation, design, and control of chemical process flowsheets.
- Author
-
Reynoso-Donzelli, Simone and Ricardez-Sandoval, Luis A.
- Subjects
- *
CHEMICAL process control , *LEARNING strategies , *DYNAMIC models , *INLETS - Abstract
• A reinforcement learning strategy is presented for optimal process integration. • Framework simultaneously generates, design and control chemical process flowsheets. • Dynamic process behavior is captured using surrogate models identified from data. • Framework includes a new step-reward system that ensures effective agent learning. • Key features of the framework are illustrated through case studies. This study introduces a Reinforcement Learning (RL) approach for synthesis, design, and control of chemical process flowsheets (CPFs). The proposed RL framework makes use of an inlet stream and a set of unit operations (UOs) available in the RL environment to build, evaluate and test CPFs. Moreover, the framework harnesses the power of surrogate models, specifically Neural Networks (NNs), to expedite the learning process of the RL agent and avoid reliance on mechanistic dynamic models embedded within the RL environment. These surrogate models approximate key process variables and descriptive closed-loop performance metrics for complex dynamic UO models. The proposed framework is evaluated through case studies, including a system where more than one type of UO is considered for simultaneous synthesis, design and control. The results show that the RL agent effectively learns to maintain the dynamic operability of the UOs under disturbances, adhere to equipment design and operational constraints, and generate viable and economically attractive CPFs. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
41. Exploring the impact of integrated design on employee learning engagement in the ubiquitous learning context: A deep learning-based hybrid multistage approach.
- Author
-
SHANG, Dawei, ZHANG, Caiyi, and JIN, Li
- Subjects
- *
EMPLOYEE psychology , *SCHOOL environment , *STRUCTURAL equation modeling , *DESCRIPTIVE statistics , *DEEP learning , *CONCEPTUAL structures , *ARTIFICIAL neural networks , *RESEARCH methodology , *PROFESSIONAL employee training , *LEARNING strategies , *COMPARATIVE studies , *EMPLOYEE attitudes , *USER interfaces - Abstract
Learning engagement has received the attention of academics and practitioners; however, studies on employee learning engagement are limited. Based on an integrated hardware-software-value-design perspective and domain-specific innovativeness theory, we developed and tested a theoretical framework using a novel and hybrid multistage approach combining a partial least squares (PLS) structural equation model (SEM) and artificial neural networks from deep learning. We used multigroup analysis (PLS-MGA-ANN), which examines key integrated design elements and domain-specific innovativeness drivers of employee learning engagement in ubiquitous learning context. According to a sample of learners' responses, the linear PLS-SEM results demonstrated that (a) integrating design elements, including perceived compatibility, familiarity, value, and user interface design, had a direct impact on domain-specific innovativeness; (b) domain-specific innovativeness had a direct impact on employee learning engagement and played a mediating role in the relationship between integrating design elements and employee learning engagement; and (c) copresence moderated the relationships between domain-specific innovativeness and employee learning engagement. Furthermore, through the evaluation of nonlinear models of the neural network, perceived compatibility and value revealed nonlinear average importance. Practical and theoretical implications are discussed. • Propose a framework on the exploration of employee learning engagement (ELE). • Use a novel and hybrid multistage approach based on statistics and machine learning. • Innovativeness plays the mediating roles between integrates design elements and ELE. • Copresence moderates the relationships between DSI and ELE. • Nonlinear results also show the differentiation average importance on ELE. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
42. Doctoral capstone theories as indicators of university rankings: Insights from a machine learning approach.
- Author
-
Stanciu, Ionut Dorin and Nistor, Nicolae
- Subjects
- *
UNIVERSITY rankings , *SCHOOL environment , *INTERDISCIPLINARY education , *SCHOLARLY method , *DOCTORAL programs , *CLINICAL medicine research , *EMPIRICAL research , *MULTIVARIATE analysis , *ACADEMIC dissertations , *ACADEMIC achievement , *CONCEPTUAL structures , *RESEARCH , *ANALYSIS of variance , *MACHINE learning , *LEARNING strategies - Abstract
Although journal articles dominate visibility and recognition in scholarly output, doctoral theses or capstones represent a significant, yet often overlooked, component of university research. This study takes a learning analytics perspective to explore the relationship between university rankings and the theoretical frameworks used in doctoral capstones within the education field, an area largely underexamined in prior research. Using the 2023 Academic Ranking of World Universities (ARWU) for education, a dataset of 9770 doctoral capstone abstracts, and a curated list of 59 learning theories, we investigated theory prevalence relative to university ranking. Employing machine learning to calculate cosine similarity between capstones and learning theories, followed by multivariate ANOVA, we identified statistically significant differences in theory usage across ranking groups. These findings suggest that theoretical choices in capstones may contribute to the external evaluations underpinning university rankings, offering insights for aligning doctoral programs with ranking criteria. However, this study's limitations, mainly its correlational nature and the U.S.-exclusive dataset, suggest the need for further research on interdisciplinarity and theory clustering across global institutions. The study makes headway in the empirical investigation into how theoretical frameworks of doctoral research may be related to university rankings, and its findings pertain to the management of educational and psychological research at doctoral level by means of learning analytics. • Doctoral capstones reveal underexplored insights into university rankings. • The choice of theories used in doctoral capstones is linked to the university rankings. • Machine learning links theory use in capstones with university rankings. • Cosine similarity can be used to detect the prevalence of theories in doctoral theses. • Exploiting the differences in theory choice might inform the direction of doctoral research. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
43. Enhancing non-intrusive load monitoring through transfer learning with transformer models.
- Author
-
Rong, Jing, Wang, Cong, Zhou, Qiuzhan, He, Yunxue, and Wu, Huinan
- Subjects
- *
TRANSFORMER models , *ENERGY management , *LEARNING strategies , *ENERGY consumption , *GENERALIZATION - Abstract
As global energy demands continue to rise, the importance of efficient energy management systems becomes increasingly clear. Non-invasive load monitoring (NILM) technologies, which identify the energy consumption of individual loads through the analysis of aggregate mains data without physical alterations to the electrical system, are gaining widespread attention. To address persistent challenges of low prediction accuracy and weak model generalization in NILM, this paper introduces TransDisNILM—an optimized transformer based NILM model enhanced by transfer learning. First, sinusoidal encoding and improved multi-head transformer encoder layers are employed to capture richer temporal features, thereby improving prediction accuracy in complex multi-load scenarios. Second, transfer learning strategies are applied to systematically select source tasks and fine-tune the model, enabling robust generalization across diverse environments and load types. Evaluation results on multiple public datasets demonstrate that TransDisNILM significantly reduces mean absolute error and normalized signal aggregate error, outperforming state-of-the-art methods. Moreover, TransDisNILM's transfer learning strategies allow effective training across different load types without starting from scratch, thus reducing the reliance on large-scale labeled datasets. Overall, TransDisNILM not only achieves higher accuracy but also exhibits stronger generalization capabilities, advancing the practical deployment of NILM technologies. • Introduced the transformer disaggregator for NILM (TransDisNILM), a transformer based model enhancing NILM system accuracy. • Developed and validated innovative source task selection methods for NILM transfer learning. • Implemented and tested pre-training and fine-tuning strategies to improve NILM adaptability. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
44. Identifying University Students' Online Self-Regulated Learning Profiles: Predictors, Outcomes, and Differentiated Instructional Strategies
- Author
-
Hyejoo Yun, Hae-Deok Song, and YeonKyoung Kim
- Abstract
Self-regulated learning (SRL) is critical in online learning, and profiling learners' SRL patterns is needed to provide personalized support. However, little research has examined how each learner performs the cyclical phases of SRL based on trace data. To fill the gap, this study attempts to derive SRL profiles encompassing all cyclical phases of forethought, performance, and self-reflection based on learning analytics and establish specific SRL support by exploring profile membership predictors and distal outcomes. Through profiling 106 students in a university online course using Latent profile analysis (LPA), four distinctive SRL profile types emerged: "Super Self-Regulated Learners," "All-around Self-Regulated Learners," "Unbalanced Self-Regulated Learners," and "Minimally Self-Regulated Learners." Multinomial logistic regression analysis revealed that task value and teaching presence significantly predicted profile membership. Additionally, multivariate analysis of variance (MANOVA) showed that cognitive, affective, behavioral, and agentic engagement and learning achievement differed significantly among the four profiles. More instructional strategies for supporting SRL are described in the paper.
- Published
- 2025
- Full Text
- View/download PDF
45. Pre-Enrollment Tests to Predict First-Year Academic Achievement in Open-Admission Higher STEM Education
- Author
-
Jolan Hanssens, Carolien Van Soom, and Greet Langie
- Abstract
The societal demand for graduates with expertise in Science, Technology, Engineering, and Mathematics (STEM) stands in contrast with prevalent issues of diminished interest among high school students in STEM subjects and low completion rates in STEM academic programs. In the Flemish educational context, heterogeneity in STEM preparedness of incoming students, caused by open admission to higher education (HE) and lack of centralized exams at the end of secondary education (SE), contributes to low completion rates. Pre-enrollment, low-stakes positioning tests aim to inform students of their preparedness. This is an in-depth study of the predictive validity for first-year academic achievement in STEM of these tests. We used data from three academic years, four universities, and six study programs (n = 1973). Using nested linear and logistic regression analyses, we found incremental predictive validity of positioning tests over SE outcomes, and learning and study strategies, especially of the mathematics problems on the tests. Furthermore, we found no evidence for differences in predictive validity between relevant sub-populations (related HE programs, or different SE programs). Finally, we introduced utility functions to determine optimal cut-offs for positioning test scores.
- Published
- 2025
- Full Text
- View/download PDF
46. Maximizing Mathematics Achievement with Less Homework Time: The Role of Self-Regulated Learning Components
- Author
-
Kexin Qin and Yehui Wang
- Abstract
Homework is a self-regulated activity that constitutes a large proportion of learning time. In mathematics, how long is the optimal homework time for achievement development and how to improve homework time efficiency have long been questions. The study examined how mathematics homework time was related to mathematics achievement among Chinese students in Grade 8 (N = 2440). The study also investigated how this relationship was moderated by self-regulated learning components (mathematics confidence, mathematics interest, mathematics anxiety, memorization strategies, elaboration strategies, and control strategies). After accounting for covariates, an inverted-U relationship between mathematics homework time and achievement was found by a generalized propensity score analysis. The optimal time that maximized achievement was 45 min per day. By increasing students' mathematics interest, the optimal time was shortened to 35 min, and thus, mathematics achievement can be maximized with less homework time. The study provides guidance for teachers to assign the appropriate amount of homework and sheds light on the ways of improving homework time efficiency.
- Published
- 2025
- Full Text
- View/download PDF
47. Peer- and Self-Assessment in Collaborative Online Language-Learning Tasks: The Role of Modes and Phases of Regulation of Learning
- Author
-
Rebecca Clayton Bernard and Gilles Kermarrec
- Abstract
This study examined how peer assessment and self-assessment influence higher education students' use of three modes of regulation of learning (self-regulation, co-regulation, and socially shared regulation) in an online collaborative task during the COVID-19 epidemic. Twenty-one first-year undergraduate students were assigned to one of three assessment conditions: self-assessment, written peer assessment, or oral peer assessment. Interview data was coded into 709 meaningful segments using content analysis and a deductive coding matrix. Segments were then investigated both qualitatively and using Chi2 quantitative analyses to explore five research questions. Results suggest (a) significant emotional difficulty was associated with peer assessment, both written and oral; (b) although negative affect was associated with the online learning context, this was mitigated in part by the self-assessment condition; (c) self-regulatory processes were more prevalent in the self-assessment condition than in the synchronous (oral) peer-assessment condition; (d) peer assessment was not associated with higher levels of socially shared regulation than self-assessment; (e) socially shared regulatory processes were not more prevalent when peer assessment was provided orally rather than in writing; and (f) modes of regulation were not equally distributed across the cyclical phases of regulation, with socially shared regulation being the predominant mode in the forethought phase and self-regulation in the reflection phase. Findings provide insight into students' affective experience of self- and peer assessment. They also shed light on the association of modes of regulation with self- and peer-assessment activities and raise new questions about emotional context and variations in modes of regulation throughout the cyclical phases model of self-regulation.
- Published
- 2025
- Full Text
- View/download PDF
48. Extracting optimal fuel cell parameters using dynamic Fick's Law algorithm with cooperative learning strategy and k-means clustering.
- Author
-
Ghetas, Mohamed and Issa, Mohamed
- Subjects
- *
PROTON exchange membrane fuel cells , *MACHINE learning , *GROUP work in education , *FUEL cells , *LEARNING strategies - Abstract
This article introduces an enhanced stochastic search method tailored for optimizing the parameters of fuel cells (FCs), which hold significant relevance across various applications. The nonlinear nature of FCs poses a modeling challenge, prompting the proposal of an advanced Dynamic Fick's Law Algorithm (DFLA). This improved approach incorporates a cooperative learning strategy, leveraging K-Means clustering, to derive optimal FC parameters. DFLA introduces a dynamic swarm topology by segmenting the population into subswarms, boosting diversity, and enabling broader global exploration. Simultaneously, the cooperative learning strategy fosters information exchange among subswarms, enhancing the algorithm's ability to explore vast solution spaces and exploit diverse subswarm-derived solutions. The evaluation of DFLA utilized real-world datasets from commercial PEMFC stacks: 250-W stack, BCS 500-W, and NedStack PS6. Performance assessment relied on the Sum Squared Error (SSE), a standard evaluation metric, with comparisons drawn against established competing methods. The results underscored DFLA's superior efficiency, showcasing improved performance metrics and convergence behaviors compared to the tested algorithms. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
49. Class activation map-based slicing-concatenation and contrastive learning: A novel strategy for record-level atrial fibrillation detection.
- Author
-
Zhu, Qiang, Zhang, Lingwei, Lu, Fei, Fang, Luping, and Pan, Qing
- Subjects
- *
DATA augmentation , *ATRIAL fibrillation , *LEARNING strategies , *ELECTROCARDIOGRAPHY , *ANNOTATIONS , *DEEP learning - Abstract
Deep learning-based models for atrial fibrillation (AF) detection require extensive training data, which often necessitates labor-intensive professional annotation. While data augmentation techniques have been employed to mitigate the scarcity of annotated electrocardiogram (ECG) data, specific augmentation methods tailored for recording-level ECG annotations are lacking. This gap hampers the development of robust deep learning models for AF detection. We propose a novel strategy, a combination of Class Activation Map-based Slicing-Concatenation (CAM-SC) data augmentation and contrastive learning, to address the current challenges. Initially, a baseline model incorporating a global average pooling layer is trained for classification and to generate class activation maps (CAMs), which highlight indicative ECG segments. After that, in each recording, indicative and non-indicative segments are sliced. These segments are subsequently concatenated randomly based on starting and ending Q points of QRS complexes, with indicative segments preserved to maintain label correctness. Finally, the augmented dataset undergoes contrastive learning to learn general representations, thereby enhancing AF detection performance. Using ResNet-101 as the baseline model, training with the augmented data yielded the highest F1-score of 0.861 on the Computing in Cardiology (CinC) Challenge 2017 dataset, a typical AF dataset with recording-level annotations. The metrics outperform most previous studies. This study introduces an innovative data augmentation method specifically designed for recording-level ECG annotations, significantly enhancing AF detection using deep learning models. This approach has substantial implications for future AF detection research. [ABSTRACT FROM AUTHOR]
- Published
- 2025
- Full Text
- View/download PDF
50. How Learner Control and Explainable Learning Analytics on Skill Mastery Shape Student Desires to Finish and Avoid Loss in Tutored Practice
- Author
-
Conrad Borchers, Jeroen Ooge, Cindy Peng, and Vincent Aleven
- Abstract
Personalized problem selection enhances student practice in tutoring systems. Prior research has focused on transparent problem selection that supports learner control but rarely engages learners in selecting practice materials. We explored how different levels of control (i.e., full AI control, shared control, and full learner control), combined with showing learning analytics on skill mastery and visual "what-if" explanations, can support students in practice contexts requiring high degrees of self-regulation, such as homework. Semi-structured interviews with six middle school students revealed three key insights: (1) participants highly valued learner control for an enhanced learning experience and better self-regulation, especially because most wanted to avoid losses in skill mastery; (2) only seeing their skill mastery estimates often made participants base problem selection on their weaknesses; and (3) "what-if" explanations stimulated participants to focus more on their strengths and improve skills until they were mastered. These findings show how explainable learning analytics could shape students' selection strategies when they have control over what to practice. They suggest promising avenues for helping students learn to regulate their effort, motivation, and goals during practice with tutoring systems. [This paper will be published in: "LAK25: The 15th International Learning Analytics and Knowledge Conference (LAK 2025), March 3-7, 2025, Dublin, Ireland," ACM, 2025. Additional funding provided by Research Foundation Flanders.]
- Published
- 2025
Catalog
Discovery Service for Jio Institute Digital Library
For full access to our library's resources, please sign in.