Learning Progress Research Articles

Universal Ensemble-Embedding Graph Neural Network for Direct Prediction of Optical Spectra from Crystal Structures.

Optical properties in solids, such as refractive index and absorption, hold vast applications ranging from solar panels to sensors, photodetectors, and transparent displays. However, first-principles computation of optical properties from crystal structures is a complex task due to the high convergence criteria and computational cost. Recent progress in machine learning shows promise in predicting material properties, yet predicting optical properties from crystal structures remains challenging due to the lack of efficient atomic embeddings. Here, Graph Neural Network for Optical spectra prediction (GNNOpt) is introduced, an equivariant graph-neural-network architecture featuring universal embedding with automatic optimization. This enables high-quality optical predictions with a dataset of only 944 materials. GNNOpt predicts all optical properties based on the Kramers-Krönig relations, including absorption coefficient, complex dielectric function, complex refractive index, and reflectance. The trained model is applied to screen photovoltaic materials based on spectroscopic limited maximum efficiency and search for quantum materials based on quantum weight. First-principles calculations validate the efficacy of the GNNOpt model, demonstrating excellent agreement in predicting the optical spectra of unseen materials. The discovery of new quantum materials with high predicted quantum weight, such as SiOs, which host exotic quasiparticles with multifold nontrivial topology, demonstrates the potential of GNNOpt in predicting optical properties across a broad range of materials andapplications.

Self-supervised learning for CT image denoising and reconstruction: a review.

This article reviews the self-supervised learning methods for CT image denoising and reconstruction. Currently, deep learning has become a dominant tool in medical imaging as well as computer vision. In particular, self-supervised learning approaches have attracted great attention as a technique for learning CT images without clean/noisy references. After briefly reviewing the fundamentals of CT image denoising and reconstruction, we examine the progress of deep learning in CT image denoising and reconstruction. Finally, we focus on the theoretical and methodological evolution of self-supervised learning for image denoising and reconstruction.

Teaching methods emphasizing phonological forms enhance L2 vocabulary learning

The present study aims to measure the effects of the teaching of second language (L2) phonological forms on L2 receptive vocabulary learning. Two teaching methods were compared in a pre- and delayed post-test to evaluate their impact on L2 word learning. Participants ( n = 127; mean age = 12;6, i.e. 12 years and 6 months) were randomly divided in two groups that followed either an explicit teaching method focused on L2 phonological forms, or a communicative teaching method focused on meaning, in which L2 phonological forms were taught implicitly. The teaching methods in the two groups aimed to foster the skills and the learning of phonological forms involved in the development of receptive vocabulary. The two teaching methods trained the same skills and relied on the same vocabulary. They both targeted the phonological forms of two difficult phonemic contrasts in French as a foreign language. The two teaching sequences took place during mandatory lessons in French as a foreign language for six weeks (12 lessons), in a Swiss state school. Generalized mixed models were fitted to the data to test for differences across teaching methods in their impact on L2 word learning. Overall, the results indicate that participants made significant progress in word learning, with no significant differences between the two teaching methods. Pronunciation, discrimination, retention in verbal working memory, and the mastery of phoneme–grapheme correspondences are significant factors of vocabulary learning in French as foreign language. The teaching of L2 phonological representations and the training of their processing facilitated the learning of words in L2 French. However, the teaching of vocabulary in French as a foreign language rarely involves a focus on phonological representations.

Structural graph federated learning: Exploiting high-dimensional information of statistical heterogeneity

With the recent progress in graph-federated learning (GFL), it has demonstrated a promising performance in effectively addressing challenges associated with heterogeneous clients. Although the majority of advances in GFL have been focused on techniques for elucidating the intricate relationships among clients, existing GFL methods have two limitations. First, current methods comprising the use of low-dimensional graphs fail to accurately depict the associations between clients, thereby compromising the performance of GFL. Second, these methods may disclose additional information when sharing client-side hidden representations. This paper presents a structural GFL (SGFL) framework and a suite of novel optimization methods. SGFL addresses the limitations of existing GFL approaches with three original contributions. Firstly, our approach advocates the dynamic construction of federated learning (FL) graphs by leveraging the high-dimensional information inherent among clients, while enabling the discovery of hierarchical communities within clients. Secondly, we present SG-FedX, a novel federated stochastic gradient optimization algorithm that mitigates the effects of heterogeneity by intelligently using a global representation. Furthermore, SG-FedX introduces a strict sharing mechanism that protects client privacy more effectively by refraining from sharing client information beyond the model parameters. Our comparative evaluations, conducted against ten representative FL algorithms under challenging non-independently-and-identically-distributed settings, demonstrated the superior performance of SG-FedX. It was noted that, in the cross-dataset scenarios, SG-FedX outperformed the second-best baseline by 8.12% and 7.91% in personalization and generalization performance, respectively.

Highly textured CMOS-compatible hexagonal boron nitride-based neuristor for reservoir computing

Significant advancements in artificial neural networks (ANNs) have driven the rapid progress of artificial intelligence and machine learning. While current feedforward neural networks primarily handle static data, recurrent neural networks (RNNs) are designed for dynamical systems. However, RNNs demand extensive training on specific tasks, limiting their scalability and affordability for edge computing. Physical reservoir computing (RC) offers an alternative approach by mapping inputs into high-dimensional states, allowing for pattern analysis within a fixed reservoir. Unlike RNNs, RC is well-suited for temporal and sequential data processing with rapid speed and low training costs. This makes RC suitable for hardware implementation across various research domains. Nonetheless, existing demonstrations of RC remain constrained to small-scale device arrays. As electronic synapse arrays aim to approach very large-scale and highly complex hardware as in the human brain, managing heat dissipation becomes a formidable challenge. In this work, we successfully developed the neuristors based on textured h-BN films, prepared using a CMOS-compatible technique, and constructed a physical RC system based on as-fabricated devices. Our approach leverages vertically aligned BN to provide aligned diffusion paths for the reproducible migration process of metal ions from the electrodes and offers a potential solution for thermal management in electronic devices. This achievement highlights the promising potential of our neuristors for future high-density and energy-efficient neuromorphic computing.

Linking Education Data in Wales (UK) – Methodology and Lessons Learned

There are separate data collections across education phases in Wales (UK). Each use a different unique learner identifier. The matched education dataset project used advanced data linking methods to produce a set of pseudo identifiers for each learner that can be matched back to the original datasets to undertake specific, de-identified analysis. The first phase of the project involved data cleaning, preparation, and the creation of new derived variables. The second phase established the linking methodology, developing and advanced data linking techniques and algorithms including frequency matching and phonetic string comparators. At each stage of the project the data linking was sequential, ranging from exact to more fuzzy matching. During the code development the approach was uniquely tailored to each data set and constantly fine-tuned to ensure the highest possible match rate while reducing potential for false matches. The resulting linked education data sets are used in a number of ways for statistical and research purposes to support the formation of evidence-based policies. This has included research into raising the compulsory education age and the evaluation of learner journeys during the pandemic. Robust linked data facilitates analysis examining the progression of learners through the education system in Wales. There is a broad scope of future analysis planned and the matched education outputs will be used extensively in the evaluation of learner journeys post pandemic to inform Welsh Government policy. The matched education dataset project involved researching and upskilling in data linkage methodology, we hope to disseminate the lessons learned widely.

Student`s Understanding toward Rural Development Development after Learning from Local Visit by the Faculty of Education Souphanouvong University; During Academic Year of 2018 - 2020

This research aimed to investigate the students’ understanding toward rural development after learning from local visit at rural areas by Faculty of Education Souphanouvong University, which is a subject that belongs to the political science curriculum of Faculty of Education, Souphanouvong University. The research was to gain insights regarding the understanding level by using the pretest and posttest to assess their learning progress, then compared means of those scores with the variable of their residence origins, The sampling students focused on 58 students (27 Women students included) of political science department, the instrument of this research was an objective test with multiple choice with IOC = 1.00 and difficulty level ranged between 0.40-0.86; and the reliability was of 0.84; there are two steps of assessment including of before and after local visiting with the same questionnaire The research findings revealed that 1). The students` understand about rural development policy after local visit was significant higher with the compared mean of posttest remark after local visit was significant different Sig .05, and 2). The students` understand about rural development policy after local visit compared to residence origins were not significant different.

THE APPLICATION OF PROBLEM BASED LEARNING (PBL) MODEL TO IMPROVE STUDENTS' LEARNING OUTCOMES ON SOCIAL STUDIES SUBJECTS IN GRADE IV OF ELEMENTARY SCHOOL II TANJUNGSARI

This research was conducted using the classroom action research (CAR) method. The progress in learning and storytelling skills is still below the Minimum Completeness Criteria and success indicators. The study included 142 students from SDN 2 Tanjungsari, ranging from grade I to grade VI. 36 sixth-grade students were part of the sample, with an equal number of males and females. The researcher collected data using tests and observations. The research instruments were observation sheets and test sheets. The analysis techniques used were qualitative and quantitative analysis. The results of this study showed that the students' learning outcomes in cycle I had a mastery level of 27.78%. In cycle II, there was an improvement in student learning outcomes with a mastery level of 52.78%. In cycle III, the mastery level reached 83.33%. This indicates that the students' learning outcomes had met the applied mastery criteria of 80%. The study also observed students' learning activities. In cycle I, the students' activities scored 19 with a percentage of 1.58%, which was still categorized as poor. In cycle II, there was an improvement in students' learning activities with a score of 27 and a percentage of 2.25%, categorized as moderate. In cycle III, students' learning activities had a score of 41 with a percentage of 3.42%, which was categorized as good. Research using Problem-Based Learning can enhance learning outcomes and student engagement in Social Studies for fourth-grade students at SDN 2 Tanjungsari in Indramayu District.

Research on the Intelligent System of English Teaching from the Perspective of Intercultural Education Based on Deep Learning

In today’s globalized world, cross-cultural communication skills are crucial for English learners. However, traditional English teaching often overlooks cultural differences among students, leading to difficulties in practical communication. The development of deep learning technology has provided new possibilities for English teaching, which can analyze students’ cultural background and learning needs through intelligent systems, and provide more personalized teaching plans. This study focuses on the application of intelligent system in English teaching from the perspective of intercultural education. By analyzing the characteristics of cross-cultural education and the advantages of deep learning technology, an intelligent system model of English teaching based on deep learning is proposed. The model can provide individualized English teaching programs according to students’ cultural backgrounds, learning needs and learning progress. A recommendation algorithm based on a cross-cultural education model comparison is proposed. For each student, retrieve similar teaching cases from the database based on their cultural background and learning needs. Compare the similarity between current students and cases in the database using similarity calculation algorithms. Based on the sustainable development teaching model of college English, the MAP algorithm is analyzed. The experimental results show that the intelligent English teaching system based on deep learning has a remarkable effect on improving students’ English achievement and intercultural communication ability. The model is trained and tested by the measured data. The error of the test results is less than 5%. The results of the experiment prove that the construction of a Chinese college English sustainable development education model can effectively enhance China’s cultural soft power.

EXPLORING STUDENT ENGAGEMENT WITH FORMATIVE ASSESSMENT AND ICT TOOLS

This study investigates the potential of Information and Communication Technology (ICT) tools to enhance formative assessment (FA) practices in English as a Foreign Language (EFL) classrooms. While previous research has focused primarily on teacher feedback in writing courses, this study emphasizes a broader scope of FA. Utilizing a cross-sectional, quantitative analysis, the research examines how ICT tools support FA's goal of facilitating the learning process from students' perspectives, thereby informing effective teaching practices and improving learning outcomes. Participants included 173 university EFL students in Japan, divided into English majors and non-English majors. The study employed a revision of an existing survey for language learning with digital tools and a self-developed survey based on an existing framework for FA. Correlational analysis indicated that participants generally perceived ICT tools as supportive of all five FA principles. Despite issues like connectivity problems, students found the tools valuable for making the learning progress visible and enhancing peer communication. The study underscores the importance of understanding and practicing FA beyond feedback, such as clarifying learning goals and fostering peer learning, to enhance learning. These findings suggest that integrating ICT tools into FA practices can significantly benefit EFL instruction by promoting student engagement and self-regulated learning.

Survey of source code vulnerability analysis based on deep learning

Amidst the rapid development of the software industry and the burgeoning open-source culture, vulnerability detection within the software security domain has emerged as an ever-expanding area of focus. In recent years, the rapid advancement of artificial intelligence, particularly the notable progress in deep learning for pattern recognition and natural language processing, has catalyzed a surge in research endeavors exploring the integration of deep learning for the enhancement of vulnerability detection techniques. In this paper, we investigate contemporary deep learning-based source code analysis methods, with a concentrated emphasis on those pertaining to static code vulnerability detection. We categorize these methods based on various representations of source code employed during the preprocessing stage, including token-based and graph-based representations of source code, and further subdivided based on the types of deep learning algorithms or graph representations employed. We summarize the basic processes of model training and vulnerability detection under these different representation formats. Furthermore, we explore the limitations inherent in current approaches and provide insights into future trends and challenges for research in this field.

The Implementation of Love Sensory Stimulation Method in Teaching Autistic Students

This research aimed to find depth insight into the implementation of love sensory in teaching English to autistic students and subsequently the impact of its implementation on their learning achievement. The research subjects are students who have problems understanding instructions, interacting, and communicating. The data collected in this research is qualitative and was collected through observation and interviews. There are some findings found such as the behavior of autistic students that is seen in the teaching and learning process such as difficulty in understanding instructions and interacting, and ignoring their learning environment are specific symptoms of Autistic Syndrome Disorder. The symptoms shown by autistic students are the cause of failure in the learning process. However, the implementation of love sensory stimulation can change the behavior of autistic students. This method can help autistic students adapt and learn together with normal students. So, autistic students’ progress in learning and have achieved their learning goals. The expectation is to provide contributions to parties in need, including teachers who teach autistic students, parents, and future researchers. Based on the findings can be concluded that autistic students can develop in their learning process and interact in their social environment if they are guided in the right way.

Exploration in 4-year-old children is guided by learning progress and novelty.

Humans are driven by an intrinsic motivation to learn, but the developmental origins of curiosity-driven exploration remain unclear. We investigated the computational principles guiding 4-year-old children's exploration during a touchscreen game (N = 102, F = 49, M = 53, primarily white and middle-class, data collected in the Netherlands from 2021-2023). Children guessed the location of characters that were hiding following predictable (yet noisy) patterns. Children could freely switch characters, which allowed us to quantify when they decided to explore something different and what they chose to explore. Bayesian modeling of their responses revealed that children selected activities that were more novel and offered greater learning progress (LP). Moreover, children's interest in making LP correlated with better learning performance. These findings highlight the importance of novelty and LP in guiding children's exploration.

MCMVDRP: a multi-channel multi-view deep learning framework for cancer drug response prediction.

Drug therapy remains the primary approach to treating tumours. Variability among cancer patients, including variations in genomic profiles, often results in divergent therapeutic responses to analogous anti-cancer drug treatments within the same cohort of cancer patients. Hence, predicting the drug response by analysing the genomic profile characteristics of individual patients holds significant research importance. With the notable progress in machine learning and deep learning, many effective methods have emerged for predicting drug responses utilizing features from both drugs and cell lines. However, these methods are inadequate in capturing a sufficient number of features inherent to drugs. Consequently, we propose a representational approach for drugs that incorporates three distinct types of features: the molecular graph, the SMILE strings, and the molecular fingerprints. In this study, a novel deep learning model, named MCMVDRP, is introduced for the prediction of cancer drug responses. In our proposed model, an amalgamation of these extracted features is performed, followed by the utilization of fully connected layers to predict the drug response based on the IC50 values. Experimental results demonstrate that the presented model outperforms current state-of-the-art models in performance.

SAFLFusionGait: Gait recognition network with separate attention and different granularity feature learnability fusion

Gait recognition, an essential branch of biometric identification, uses walking patterns to identify individuals. Despite its effectiveness, gait recognition faces challenges such as vulnerability to changes in appearance due to factors like angles and clothing conditions. Recent progress in deep learning has greatly enhanced gait recognition, especially through methods like deep convolutional neural networks, which demonstrate impressive performance. However, current approaches often overlook the connection between coarse-grained and fine-grained features, thereby restricting their overall effectiveness. To address this limitation, we propose a new framework for gait recognition framework that combines deep-supervised fine-grained separation with coarse-grained feature learnability. Our framework includes the LFF module, which consists of the SSeg module for fine-grained information extraction and a mechanism for fusing coarse-grained features. Furthermore, we introduce the F-LCM module to extract local disparity features more effectively with learnable weights. Evaluation on CASIA-B and OU-MVLP datasets shows superior performance compared to classical networks.

Quantitative analysis of student learning progress in online curses during the COVID-19 Pandemic

Abstract The pandemic of COVID-19 forced the transition of education to online platforms. This shift allowed for a more in-depth study of the learning process of students, as it provided access to information about their activities and interactions with the course’s contents [1-3]. In this work, we apply quantitative methods to analyze the student learning progress of students for two first year courses of the B.Sc. degree in Physics (Universidad Católica del Norte). The data are collected from activity logs of digital platforms during Academic year 2021, which were employed to characterize the student learning habits along these processes (experience, selectivity and workflow). Statistical analysis is based on the distribution of the time spent on learning activities, such as the daily time distribution and the cumulative attention per online materials. The results show that the learning process of students in online courses is complex and adaptive. We report statistical patterns similar to those found in financial markets. These empirical results provide evidence about the importance of economic principles and methods in the understanding of the teaching-learning process.

Using eye tracking to assess learning of a multifunction prosthetic hand: an exploratory study from a rehabilitation perspective

BackgroundEye tracking technology not only reveals the acquisition of visual information at fixation but also has the potential to unveil underlying cognitive processes involved in learning to use a multifunction prosthetic hand. It also reveals gaze behaviours observed during standardized tasks and self-chosen tasks. The aim of the study was to explore the use of eye tracking to track learning progress of multifunction hands at two different time points in prosthetic rehabilitation.MethodsThree amputees received control training of a multifunction hand with new control strategy. Detailed description of control training was collected first. They wore Tobii Pro2 eye-tracking glasses and performed a set of standardized tasks (required to switch to different grips for each task) after one day of training and at one-year-follow-up (missing data for Subject 3 at the follow up due to socket problem). They also performed a self-chosen task (free to use any grip for any object) and were instructed to perform the task in a way how they would normally do at home. The gaze-overlaid videos were analysed using the Tobii Pro Lab and the following metrics were extracted: fixation duration, saccade amplitude, eye-hand latency, fixation count and time to first fixation.ResultsDuring control training, the subjects learned 3 to 4 grips. Some grips were easier, and others were more difficult because they forgot or were confused with the switching strategies. At the one-year-follow-up, a decrease in performance time, fixation duration, eye-hand latency, and fixation count was observed in Subject 1 and 2, indicating an improvement in the ability to control the multifunction hand and a reduction of cognitive load. An increase in saccade amplitude was observed in both subjects, suggesting a decrease in difficulty to control the prosthetic hand. During the standardized tasks, the first fixation of all three subjects were on the multifunction hand in all objects. During the self-chosen tasks, the first fixations were mostly on the objects first.ConclusionThe qualitative data from control training and the quantitative eye tracking data from clinical standardized tasks provided a rich exploration of cognitive processing in learning to control a multifunction hand. Many prosthesis users prefer multifunction hands and with this study we have demonstrated that a targeted prosthetic training protocol with reliable assessment methods will help to lay the foundation for measuring functional benefits of multifunction hands.

GLAN: Global Local Attention Network for Thoracic Disease Classification to Enhance Heart Failure Diagnosis

ABSTRACTChest x‐ray (CXR) is an effective method for diagnosing heart failure, and identifying important features such as cardiomegaly, effusion, and edema on patient chest x‐rays is significant for aiding the treatment of heart failure. However, manually identifying a vast amount of CXR data places a huge burden on physicians. Deep learning's progression has led to the utilization of this technology in numerous research aimed at tackling these particular challenges. However, many of these studies utilize global learning methods, where the contribution of each pixel to the classification is considered equal, or they overly focus on small areas of the lesion while neglecting the global context. In response to these issues, we propose the Global Local Attention Network (GLAN), which incorporates an improved attention module on a branched structure. This enables the network to capture small lesion areas while also considering both local and global features. We evaluated the effectiveness of the proposed model by testing it on multiple public datasets and real‐world datasets. Compared to the state‐of‐the‐art methods, our network structure demonstrated greater accuracy and effectiveness in the identification of three key features: cardiomegaly, effusion, and edema. This provides more targeted support for diagnosing and treating heart failure.

From literacy to learning: The sequential mediation of attitudes and enjoyment in AI-assisted EFL education

Artificial intelligence (AI) has become increasingly prominent in English as a Foreign Language (EFL) education. However, research on the factors influencing learners' sustained engagement with AI-assisted learning is scarce. This study bridges this gap by investigating the impact of AI literacy, learner attitudes toward AI-assisted learning, and foreign language enjoyment (FLE) on Chinese EFL learners' continuance intention (CI) for AI-assisted EFL learning. A survey of 417 EFL learners was conducted, and structural equation modeling was utilized to elucidate the relationships between these variables. The findings reveal that AI literacy, learner attitudes, and FLE significantly predict the CI. Moreover, AI literacy indirectly influences CI by mediating learner attitudes and FLE. Finally, FLE mediates the relationship between learner attitudes and CI. The study verifies a sequential mediation model, providing new insights into EFL learners' cognitive and affective traits and extending current models on the continuance intention in AI-assisted EFL learning. The research underscores the necessity for educators and technology developers to consider learners' interests and feedback, which are crucial for the sustainable progress of AI-assisted EFL learning.

Exploring learning characteristics and progression ofGPtrainees based in regional, rural and remote settings:Aqualitative study.

Globally, most doctors train and work in metropolitan areas but a key strategy for developing the rural health workforce is expanding rural training. The aim of this study was to describe the scope and quality of learning along with skill acquisition of GP trainees based in regional, rural, and remote settings. Regional, rural and remote settings in Queensland Australia excluding Brisbane. Thirty-seven general practice trainees who had undertaken their first community placement were recruited from regional, rural and remote learning settings within Queensland. Qualitative descriptive methodology based on constructionist epistemology was employed which allowed for the exploration of current GP trainees' experiences. Trainees were interviewed and data were thematically analysed as to the scope and quality of learning by the setting of training. Learning experiences were then mapped against the Dreyfus and Dreyfus model to explore skill acquisition. In terms of the scope and quality of learning, rural and remote trainees mainly focused on the diverse and unique (sometimes challenging) experiences their setting offered compared with regionally based trainees. Mapping of the trainee comments to the Dreyfus model of skill acquisition demonstrated that regardless of setting, equivalent learning occurred by GP trainees. This exploratory study provides evidence that rural and remote learning may provide a more diverse and challenging experience. It suggests an equivalence of quality of education and skill acquisition across settings rural, regional and remote. This suggests that the training distribution policies may not disadvantage GP trainees and the scope and quality of more remote learning may increase uptake of remote placements.