The accurate diagnosis of early mild cognitive impairment is crucial for timely intervention and treatment of dementia. But it is challenging to distinguish from normal aging due to its complex pathology and mild symptoms. Recently, effective hyper-connectivity identified through directed hypergraph can be considered as an effective analysis approach for early detection of mild cognitive impairment and exploration of its underlying neural mechanisms, because it captures directional higher-order interactions across multiple brain regions. However, current methods face limitations, including inefficiency in high-dimensional spaces, sensitivity to noise, reliance on manually defined structures, lack of global structural information, and static learning mechanisms. To address these issues, we integrate robust dictionary learning with directed hypergraph structure learning within a unified framework. This approach jointly estimates low-dimensional sparse representations and the directed hypergraph. The integration allows both processes to dynamically reinforce each other, leading to the refinement of the directed hypergraph, which improves the estimation of low-dimensional sparse representations and, in turn, enhances the quality of the directed hypergraph estimation. Experimental analyses on simulated data confirm the positive interplay between these processes, demonstrating the effectiveness of the proposed collaborative learning strategy. Furthermore, results on real-world brain signal data show that the proposed method is highly competitive in early detection of mild cognitive impairment, highlighting its ability to identify effective hyper-connectivity networks with significant differences.

Time-series data such as fMRI and MEG carry a wealth of inherent spatio-temporal coupling relationship, and their modeling via deep learning is essential for uncovering biological mechanisms. However, current machine learning models for mining spatio-temporal information usually overlook this intrinsic coupling association, in addition to poor explainability. In this paper, we present an explainable learning framework for spatio-temporal coupling. Specifically, this framework constructs a deep learning network based on spatio-temporal correlation, which can well integrate the time-varying coupled relationships between node representation and inter-node connectivity. Furthermore, it explores spatio-temporal evolution at each time step, providing a better explainability of the analysis results. Finally, we apply the proposed framework to brain dynamic functional connectivity (dFC) analysis. Experimental results demonstrate that it can effectively capture the variations in dFC during brain development and the evolution of spatio-temporal information at the resting state. Two distinct developmental functional connectivity (FC) patterns are identified. Specifically, the connectivity among regions related to emotional regulation decreases, while the connectivity associated with cognitive activities increases. In addition, children and young adults display notable cyclic fluctuations in resting-state brain dFC.

This work-in-progress paper employs a machine learning method for the automated analysis of research interests in Artificial Intelligence in Education (AIED) at scale. We aim to analyze the essential techniques and critical educational problems studied by researchers in five AIED-related conferences and journals between 2010 and 2022. We trained and compared different machine learning models and feature extraction techniques to achieve the research objective. After comparing different models and hyperparameter combinations, our classifier achieves an accuracy of 0.87 and Cohen's kappa of 0.80. Based on the classification results, we identified the top 10 most frequent keywords within each category for every four year period over the past 12 years. Using the classifier, the 10,723 keywords from 2,684 articles were classified into three categories: educational foci, technical solutions, and AIED applications. We find that 'natural language processing' and 'machine learning' are the primary technical keywords in AIED research, and 'deep learning' and 'artificial intelligence' are the trending technical keywords since 2017. Meanwhile, 'massive open online courses', 'self-regulated learning', 'feedback', 'collaborative learning', and 'online learning' are the top educational foci in the field over the last 12 years. 'Intelligent tutoring systems', 'educational data mining', 'knowledge tracing', and 'learning analytics' continue to receive attention as AIED applications of sustained interest. This study helps to understand the educational foci and technical solutions of AIED research at scale and provides insights into the future of AIED research.

Abstract This article mainly focuses on the most common types of high-speed railways malfunctions in overhead contact systems, namely, unstressed droppers, foreign-body invasions, and pole number-plate malfunctions, to establish a deep-network detection model. By fusing the feature maps of the shallow and deep layers in the pretraining network, global and local features of the malfunction area are combined to enhance the network's ability of identifying small objects. Further, in order to share the fully connected layers of the pretraining network and reduce the complexity of the model, Tucker tensor decomposition is used to extract features from the fused-feature map. The operation greatly reduces training time. Through the detection of images collected on the Lanxin railway line, experiments result show that the proposed multiview Faster R-CNN based on tensor decomposition had lower miss probability and higher detection accuracy for the three types faults. Compared with object-detection methods YOLOv3, SSD, and the original Faster R-CNN, the average miss probability of the improved Faster R-CNN model in this paper is decreased by 37.83%, 51.27%, and 43.79%, respectively, and average detection accuracy is increased by 3.6%, 9.75%, and 5.9%, respectively.

Online Q&A has become a very common way for people to find information, and sending instant messages via social networking sites offers new opportunities for companies and individuals to reach out for information at a more private level. Despite the popularity of social instant messaging, little is known about how social instant messaging requests may affect the responses received. This study offers a new contribution by reporting a randomised-controlled study on social instant messaging, in which we phrased a basic request ‘Should I watch the Jurassic World' into 16 variants along four axes: punctuation, number of sentences, scope, and emoji availability, and sent the requests to 160 strangers. The response rate, speed and perceived answer usefulness were measured, indicating that 95% questions received replies, 55% of responses were received within one minute, and 22% were perceived as useful. We also statistically investigated what factors related to the four dimensions might influence the answer usefulness. The results showed that if a request was phrased as an explicit question with ‘?’, addressed to a specific respondent ‘you’, and accompanied with a friendly emoji , it was most likely to receive useful responses. The theoretical and practical implications are discussed.

Abstract Information service providers often require evidence from multiple, heterogeneous information sources to better characterize users and offer personalized service. In many cases, statistic information (for example, users' profiles) and sequentially dynamic information (for example, logs of interaction with information systems) are two prominent sources that can be combined to achieve optimized results. Previous attempts in combining these two sources mainly exploited models designed for either static or sequential information, but not both. This study aims to fill the gap by proposing a novel joint neural network model that can naturally fit both static and sequential user data. To evaluate the effectiveness of the proposed method, this study uses the problem of at‐risk student prediction as an example where both static data (personal profiles) and sequential data (event logs) are involved. A thorough evaluation was conducted on an open data set, with comparisons to a range of existing approaches including both static and sequential models. The results reveal superb performances of the proposed method. Implications of the findings on further research and applications of joint models are discussed.

The low earth orbit (LEO) satellite network has been a valuable architecture due to its characteristics of wide coverage and low transmission delay. Utilizing LEO satellites as edge computing nodes to provide real-time services for access terminals will be the indispensable paradigm of integrated space-air-ground network. However, it is not easy to design resource management strategies in edge computing satellite (ECS), considering different accessing planes and resource requirements of terminals. Moreover, a comprehensive analysis of the network topology, relative motion, and available resources is required to establish ECS collaborative networks. To address these problems, the dynamic resource allocation architecture and advanced K-means algorithm (AKA) in ECSs are proposed. Then, the extended graph model and breadth-first-search-based spanning tree (BFST) algorithm are utilized to guide the inter-satellite link (ISL) construction. As a result, the ECS collaborative network is established with fine-grained resource management. Simulation results show that the proposed fine- grained resource management scheme is feasible and effective.

Gaps between knowledge sources are interesting to various stakeholders: they might indicate potential misconceptions awaiting correction, complex or novel knowledge that requires careful delivery or studying. Motivated by these underlying values, this study explores the knowledge gap phenomenon in the context of student textual responses. In the method proposed in this study, discourses are first mapped into structured knowledge spaces where gaps between correct/incorrect responses and assessed knowledge are measured by network-based metrics. Empirical results demonstrate the effectiveness of the proposed method in measuring gaps in student responses. The networked representation of texts proposed in this study is novel in quantitatively framing gaps of knowledge. It also offers a set of validated metrics for analyzing student responses in research and practice.

Digital platforms enable the observation of learning behaviors through fine-grained log traces, offering more detailed clues for analysis. In addition to previous descriptive and predictive log analysis, this study aims to simultaneously model learner activities, event time spans, and interaction levels using the proposed Hidden Behavior Traits Model (HBTM). We evaluated model performance and explored their capability of clustering learners on a public dataset, and tried to interpret the machine recognized latent behavior patterns. Quantitative and qualitative results demonstrated the promising value of HBTM. Results of this study can contribute to the literature of online learner modeling and learning service planning.

Wikiglass is a learning analytic tool for visualizing collaborative work on Wikis built by groups of secondary or primary school students. This poster presents new features of Wikiglass developed recently based on requests from teachers, including flexible selection of date range, revision network, and thinking order detection. Currently the new features are used and evaluated in two secondary schools in Hong Kong.

Remote measurement of SIF has opened a new perspective to assess plant actual photosynthesis at larger, ecologically relevant scales. However, understanding the underling mechanisms between SIF and GPP remains challenging before SIF used as a robust constraint for estimating GPP. In this study, GOME-2 SIF was found to be consistently related to MODIS GPP. We also noticed the SIF-GPP relationship was ecosystem-specific and influenced by land surface temperature. The former was due to some structural and physiological characteristics related to each ecosystem. The latter can be attributed to the biochemical process influenced by temperature conditions. Model simulations also indicated the SIF-GPP relationship was complex and affected by some factors like chlorophyll content and LAI. Our study contributes to a better understanding of the information inherent in remotely sensed SIF and its functional relationship to GPP.

The scale effects in earth science, which are related to various aspects in remote sensing monitoring, have become an international prosperous research area. As spatial heterogeneity of the earth system limits the transferring between different scale, it is necessary to study these spatial heterogeneity factors, and analyze their impact on NPP scale effect. Then we can introduce an approach to perform spatial scale calibration based on a correction factor for scale effect, and perform it to NPP. This study presented an approach driven by remotely sensed data and meteorological data to estimate GPP and NPP over regional scales. By using multi-scale data and different scaling strategies, NPP of Heihe River Basin in 2012 with various scales were derived. With a focus on differences among land cover types, we introduced and tested a kind of spatial scale calibration method, to get close to the real value of the net primary productivity.

Schizophrenia is one of the most prevalent mental diseases, and is considered to be caused by the interplay of a number of genetic factors. In this paper, by constructing a multilayer restricted Boltzmann machines (RBMs) deep network, we use the genomic data (i.e., SNP data) for unsupervised feature learning and disease diagnosis of schizophrenia. In order to obtain some more accurate diagnosis results by RBMs, firstly, we transform the SNP data into binary sequences, and then by training the multi-layer RBMs deep network on unlabeled data, the multi-level abstract features of the genomic data are obtained and stored in the network. Finally, by adding a linear classifier to the top of the multi-layer RBMs deep network, the classification results on the testing data are gained. The results show that the average performance of this method is better than that of other methods, e.g., SVM (including linear SVM as well as SVM with multilayer perceptron kernel), sparse representations based classifier and k-nearest neighbors method. It is indicated that the multi-layer RBMs deep network can extract deep hierarchical representations of the genomic data, and then promises a more comprehensive approach for the mental disease diagnosis.

The transpiration from shelterbelts and the evapotranspiration (ET) from cropland (maize and vegetables) and orchards (apple) in an irrigated area in the middle reaches of the Heihe River, China, were estimated using a modified Penman-Monteith (P-M) formula and airborne remote sensing data. The results were compared to the shelter transpiration results obtained from measurements of sup flow in tree trunks made with thermal dissipation probes and the latent heat fluxes observed by the eddy covariance technique at flux towers in croplands. The modified P-M formula was found to be an effective means to estimate not only the cropland and orchard ET but also the shelter transpiration. The seasonal variation of shelterbelt transpiration was smaller than those of cropland and orchard ET. Estimates of ET made using the P-M formula along with the remote sensing data showed that 9.9%, 3.1%, and 87.0% of the total ET were allotted to shelterbelts, apple orchards, and cropland, respectively.

In this paper, we propose a new class of parallel branch-and-bound (B&B) schemes. The main idea of the scheme is to focus on the functional parallelism instead of conventional data parallelism, and to support such a heterogeneous and irregular parallelism by using a collection of autonomous agents distributed over the network. After examining several design issues toward the implementation of a prototype of the distributed B&B system, we illustrate the result of our preliminary experiments conducted to estimate the performance of the proposed scheme. The result shows that it could cause a significant performance improvement if each agent autonomously changes its function type according to the change of the underlying environment.