Scale Datasets Research Articles

Predicting Drug-drug Interaction with Graph Mutual Interaction Attention Mechanism

Effective representation of molecules is a crucial step in AI-driven drug design and drug discovery, especially for drug-drug interaction (DDIs) prediction. Previous work usually models the drug information from the drug-related knowledge graph or the single drug molecules, but the interaction information between molecular substructures of drug pair is seldom considered, thus often ignoring the influence of bond information on atom node representation, leading to insufficient drug representation. Moreover, key molecular substructures have significant contribution to the DDIs prediction results. Therefore, in this work, we propose a novel Graph learning framework of Mutual Interaction Attention mechanism (called GMIA) to predict DDIs by effectively representing the drug molecules. Specifically, we build the node-edge message communication encoder to aggregate atom node and the incoming edge information for atom node representation and design the mutual interaction attention decoder to capture the mutual interaction context between molecular graphs of drug pairs. GMIA can bridge the gap between two encoders for the single drug molecules by attention mechanism. We also design a co-attention matrix to analyze the significance of different-size substructures obtained from the encoder-decoder layer and provide interpretability. In comparison with other recent state-of-the-art methods, our GMIA achieves the best results in terms of area under the precision-recall-curve (AUPR), area under the ROC curve (AUC), and F1 score on two different scale datasets. The case study indicates that our GMIA can detect the key substructure for potential DDIs, demonstrating the enhanced performance and interpretation ability of GMIA.

Sign Language Recognition Using Graph and General Deep Neural Network Based on Large Scale Dataset

Sign Language Recognition Using Graph and General Deep Neural Network Based on Large Scale Dataset

Hidden in plain sight for too long: Using text mining techniques to shine a light on workplace sexism and sexual harassment.

Objective: The goal of this study is to understand how people experience sexism and sexual harassment in the workplace by discovering themes in 2,362 experiences posted on the Everyday Sexism Project's website this http URL. Method: This study used both quantitative and qualitative methods. The quantitative method was a computational framework to collect and analyze a large number of workplace sexual harassment experiences. The qualitative method was the analysis of the topics generated by a text mining method. Results: Twenty-three topics were coded and then grouped into three overarching themes from the sex discrimination and sexual harassment literature. The Sex Discrimination theme included experiences in which women were treated unfavorably due to their sex, such as being passed over for promotion, denied opportunities, paid less than men, and ignored or talked over in meetings. The Sex Discrimination and Gender harassment theme included stories about sex discrimination and gender harassment, such as sexist hostility behaviors ranging from insults and jokes invoking misogynistic stereotypes to bullying behavior. The last theme, Unwanted Sexual Attention, contained stories describing sexual comments and behaviors used to degrade women. Unwanted touching was the highest weighted topic, indicating how common it was for website users to endure being touched, hugged or kissed, groped, and grabbed. Conclusions: This study illustrates how researchers can use automatic processes to go beyond the limits of traditional research methods and investigate naturally occurring large scale datasets on the internet to achieve a better understanding of everyday workplace sexism experiences.

Checking Counterfeit Critiques on Commodities using Ensemble Classifiers Enhancing Information Credibility

The conundrum of the ubiquitous deceptive reviews has overruled the online ontology with the obsession of obscure but obligatory posting of product reviews for the customers to believe, behold and beget the online product marketing. This mandates contemporary research in the direction to delve deeper on the application and analysis of deceiving online reviews with matured and advanced AI models functional on large scale datasets to effectively and efficiently demarcate between the genuine and the sham. The research counteracts the counterfeiting product reviews via the applications, assessment and analysis of the befitting AI models - Elastic-net Classifier model based on block coordinate descent with Wordcloud and its further performance enhancement through LightGBM Trees Classifier with Grid Search and Early Stopping support, with Log-Loss as performance metric for experimentation to gain insight into the intricacies of detection, diagnosis and diminution of fake product reviews. The paper also delineates discriminative and affirmative aspects of the dataset quality, statistics, stability and standards inherent and coherent to the creation of the dataset using Large Language Models (LLMs) intrinsic to the zeitgeist juncture of recent times promoting machines to produce large scale, cost effective bogus reviews in lieu of the Amazon Mechanical Turks. The results obtained with the Log-Loss holdout score of 0.1462 conforming the LightGBM classifier proves its performance better than the Elastic-Net classifier, conforming it as better than the ROC-AUC in terms of its proximity to the prediction probability for the matching actual/true value.

Scales of Connectivity within Stream Temperature Networks of the Clackamas River Basin, Oregon

Water quality varies along the stream network; thus, considering the directional, dendritic nature of stream networks with surrounding landscape variables is essential in explaining spatial variations of water quality. Using a spatially extensive stream temperature monitoring effort in the Clackamas River Basin in the United States, we first compare spatial scales of analysis of atmospheric, landscape, and in-stream explanatory variables through their correlation with summer stream temperatures. We then derive a predictive stream temperature model with factors representing the spatial variation of local climate, recent wildfire effects, and discharge. Finally, we compare nonspatial multiple linear regression to a spatial stream network (SSN) model to assess the combined importance of the spatial scale of analysis and flow-connected stream distance in explaining total variation in stream temperatures. Most explanatory variables show the most highly significant relationships to stream temperature when derived as a percentage of the total upstream area above observation sites. Elevation and vegetation cover, however, were most significantly correlated to stream temperature at the riparian buffer area scale and the local reach contributing area scale, respectively. Multiple regression analysis using total upstream burned area, total upstream area with underlying High Cascades geology, and the elevation within the 100-m-wide riparian area explained 81 percent of variation in stream temperature. SSN outperformed this nonspatial statistical model, however, in explaining the total variation in stream temperature. These comparisons of scaled data sets demonstrate both the local and cumulative upstream effects on stream temperature, providing a spatial network-informed framework to those prioritizing watershed restoration and wildfire recovery activities.

Towards accurate knowledge transfer via target-awareness representation disentanglement

Fine-tuning deep neural networks pre-trained on large scale datasets is one of the most practical transfer learning paradigm given limited quantity of training samples. To obtain better generalization, using the starting point as the reference (SPAR), either through weights or features, has been successfully applied to transfer learning as a regularizer. However, due to the domain discrepancy between the source and target task, there exists obvious risk of negative transfer in a straightforward manner of knowledge preserving. In this paper, we propose a novel transfer learning algorithm, introducing the idea of Target-awareness REpresentation Disentanglement (TRED\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{TRED}$$\\end{document}), where the relevant knowledge with respect to the target task is disentangled from the original source model and used as a regularizer during fine-tuning the target model. Two alternative approaches, maximizing Maximum Mean Discrepancy (Max-MMD) and minimizing mutual information (Min-MI) are introduced to achieve the desired disentanglement. Experiments on various real world datasets show that our method stably improves the standard fine-tuning by more than 2% in average. TRED\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{TRED}$$\\end{document} also outperforms related state-of-the-art transfer learning regularizers such as L2-SP\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\mathrm {L^2\ ext {-}SP}$$\\end{document}, AT\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{AT}$$\\end{document}, DELTA\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{DELTA}$$\\end{document}, and BSS\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ extrm{BSS}$$\\end{document}. Moreover, our solution is compatible with different choices of disentangling strategies. While the combination of Max-MMD and Min-MI typically achieves higher accuracy, only using Max-MMD can be a preferred choice in applications with low resource budgets.

A multi-objective comparison of CNN architectures in Arctic human-built infrastructure mapping from sub-meter resolution satellite imagery

ABSTRACTRisk assessment of infrastructure exposed to ice-rich permafrost hazards is essential for climate change adaptation in the Arctic. As this process requires up-to-date, comprehensive, high-resolution maps of human-built infrastructure, gaps in such geospatial information and knowledge of the applications required to produce it must be addressed. Therefore, this study highlights the ongoing development of a deep learning approach to efficiently map the Arctic built environment by detecting nine different types of structures (detached houses, row houses, multi-story blocks, non-residential buildings, roads, runways, gravel pads, pipelines, and storage tanks) from recently-acquired Maxar commercial satellite imagery (<1 m resolution). We conducted a multi-objective comparison, focusing on generalization performance and computational cost, of nine different semantic segmentation architectures. K-fold cross validation was used to estimate the average F1-score of each architecture and the Friedman Aligned Ranks test with the Bergmann-Hommel post-hoc procedure was applied to test for significant differences in generalization performance. ResNet-50-UNet++ performs significantly better than five out of the other eight candidate architectures; no significant difference was found in the pairwise comparisons of ResNet-50-UNet++ to ResNet-50-MANet, ResNet-101-MANet, and ResNet-101-UNet++. We then conducted a high-performance computing scaling experiment to compare the number of service units and runtime required for model inferencing on a hypothetical pan-Arctic scale dataset. We found that the ResNet-50-UNet++ model could save up to ~ 54% on service unit expenditure, or ~ 18% on runtime, when considering operational deployment of our mapping approach. Our results suggest that ResNet-50-UNet++ could be the most suitable architecture (out of the nine that were examined) for deep learning-enabled Arctic infrastructure mapping efforts. Overall, our findings regarding the differences between the examined CNN architectures and our methodological framework for multi-objective architecture comparison can provide a foundation that may propel future pan-Arctic GeoAI mapping efforts of infrastructure.

Distinct biogeographic patterns in Glomeromycotinian and Mucoromycotinian arbuscular mycorrhizal fungi across China: A meta-analysis

Fine root endophytes, recently reclassified as Mucoromycotinian arbuscular mycorrhizal fungi (M-AMF), are now recognized as functionally important as Glomeromycotinian AMF (G-AMF). However, little is known about the biogeography and ecology of M-AMF and G-AMF communities, particularly on a large scale, preventing a systematic assessment of ecosystem diversity and functioning. Here, we investigated the biogeographic assemblies and ecological diversity patterns of both G-AMF and M-AMF, using published 18S rDNA amplicon datasets and associated metadata from 575 soil samples in six ecosystems across China. Contrasting with G-AMF, putative M-AMF were rare in natural/semi-natural sites, where their communities were a subset of those in agricultural sites characterized by intensive disturbances, suggesting different ecological niches that they could occupy. Spatial and environmental factors (e.g., vegetation type) significantly influenced both fungal communities, with soil total‑nitrogen and mean-annual-precipitation being the strongest predictors for G-AMF and M-AMF richness, respectively. Both groups exhibited a strong spatial distance-decay relationship, shaped more by environmental filtering than spatial effects for M-AMF, and the opposite for G-AMF, presumably because stochasticity (e.g., drift) dominantly structured G-AMF communities; while the narrower niche breadth (at community-level) of M-AMF compared to G-AMF suggested its more susceptibility to environmental differences. Furthermore, co-occurrence network links between G-AMF and M-AMF were prevalent across ecosystems, and were predicted to play a key role in stabilizing overall communities harboring both fungi. Based on the macroecological spatial scale datasets, this study provides solid evidence that the two AMF groups have distinct ecological preferences at the continental scale in China, and also highlights the potential impacts of anthropogenic activities on distributions of AMF. These results advance our knowledge of the ecological differences between the two fungal groups in terrestrial ecosystems, suggesting the need for further field-based investigation that may lead to a more sophisticated understanding of ecosystem function and sustainable management.

Gene regulatory network inference based on causal discovery integrating with graph neural network

AbstractGene regulatory network (GRN) inference from gene expression data is a significant approach to understanding aspects of the biological system. Compared with generalized correlation‐based methods, causality‐inspired ones seem more rational to infer regulatory relationships. We propose GRINCD, a novel GRN inference framework empowered by graph representation learning and causal asymmetric learning, considering both linear and non‐linear regulatory relationships. First, high‐quality representation of each gene is generated using graph neural network. Then, we apply the additive noise model to predict the causal regulation of each regulator‐target pair. Additionally, we design two channels and finally assemble them for robust prediction. Through comprehensive comparisons of our framework with state‐of‐the‐art methods based on different principles on numerous datasets of diverse types and scales, the experimental results show that our framework achieves superior or comparable performance under various evaluation metrics. Our work provides a new clue for constructing GRNs, and our proposed framework GRINCD also shows potential in identifying key factors affecting cancer development.

ActiveZero++: Mixed Domain Learning Stereo and Confidence-based Depth Completion with Zero Annotation.

Learning-based stereo methods usually require a large scale dataset with depth, however obtaining accurate depth in the real domain is difficult, but groundtruth depth is readily available in the simulation domain. In this paper we propose a new framework, ActiveZero++, which is a mixed domain learning solution for active stereovision systems that requires no real world depth annotation. In the simulation domain, we use a combination of supervised disparity loss and self-supervised loss on a shape primitives dataset. By contrast, in the real domain, we only use self-supervised loss on a dataset that is out-of-distribution from either training simulation data or test real data. To improve the robustness and accuracy of our reprojection loss in hard-to-perceive regions, our method introduces a novel self-supervised loss called temporal IR reprojection. Further, we propose the confidence-based depth completion module, which uses the confidence from the stereo network to identify and improve erroneous areas in depth prediction through depth-normal consistency. Extensive qualitative and quantitative evaluations on real-world data demonstrate state-of-the-art results that can even outperform a commercial depth sensor. Furthermore, our method can significantly narrow the Sim2Real domain gap of depth maps for state-of-the-art learning based 6D pose estimation algorithms.

Delineating and Characterizing the Metropolitan Fringe Area of Shanghai—A Spatial Morphology Perspective

The metropolitan fringe area is of great significance to a city’s future growth. However, relatively little attention has been paid to delineating and characterizing the metropolitan fringe area from a spatial morphology perspective, which contributes to the planning and design of metropolitan fringe areas. Therefore, the aim of this study is to develop a morphology–based method to delineate the metropolitan fringe area and investigate the characteristics of its spatial morphology. Drawing upon a large–scale dataset on the spatial morphology of Shanghai, this study finds that the metropolitan fringe area is generally circular in shape and is dominated by residential, industrial, agricultural and forestry land. The metropolitan fringe area accounts for 24.65% of the total area of Shanghai and is mainly located between its outer ring and suburban ring areas. The distributions of spatial characteristics of the metropolitan fringe area suggest that the area has a relatively lower level in terms of building height, building density, and development intensity. Furthermore, the metropolitan fringe area of Shanghai contains five key spatial elements, including residential and industrial clusters, shadow spaces accompanying clusters, corridor lines, green wedges, and surfaces. The interaction of the five spatial elements lays the foundation for the prototype of the spatial structure of the metropolitan fringe area of Shanghai, which is of great significance to understanding the heterogeneity within the metropolitan fringe area in terms of the distribution of spatial morphological characteristics. Such heterogeneity also needs to be considered in the planning and design of the metropolitan fringe area.

Do the Words We Choose Matter When Prescribing Medications?

Caregivers are often apprehensive about treating childhood atopic dermatitis (AD) with topical corticosteroids but may find comfort if treatments are presented in a patient-centered manner. We assessed caregivers' willingness to treat AD with either a "topical steroid," "topical medication," or "treatment, similar to the all-natural signals produced by the adrenal glands in the body." A survey randomized 874 caregivers of children with AD to receive a "topical steroid," "topical medication," or "treatment, similar to the all-natural signals produced by the adrenal glands in the body." A scenario-only dataset received these descriptions, while a descriptive heading dataset and expanded scale dataset also received headings of "Topical Steroid Use," "Topical Medication Use," and "All-Natural Treatment Use," respectively. Responses were recorded on a 6-point Likert scale or 0-100 slider scale. Whole and dichotomized responses were evaluated using 2-tailed, independent sample t-tests. For the descriptive heading and expanded scale datasets, those presented with a "topical medication" reported greater willingness to treat than those presented with a "topical steroid" and "all-natural treatment" in the descriptive heading dataset (P&lt;0.05). For the dichotomized scenario-only dataset, those presented with a "treatment, similar to the all-natural signals produced by the adrenal glands in the body," reported greater willingness than those presented with a "topical medication" (P&lt;0.05). Initially presenting caregivers with a "topical medication" rather than a "topical steroid" may improve willingness to treat AD for some caregivers. However, tailoring the discussion to best fit caregivers&rsquo; understanding of treatment may be the most beneficial approach. J Drugs Dermatol. 2023;22(12):1216-1219. doi:10.36849/JDD.5746.

GcForest-based compound-protein interaction prediction model and its application in discovering small-molecule drugs targeting CD47.

Identifying compound-protein interaction plays a vital role in drug discovery. Artificial intelligence (AI), especially machine learning (ML) and deep learning (DL) algorithms, are playing increasingly important roles in compound-protein interaction (CPI) prediction. However, ML relies on learning from large sample data. And the CPI for specific target often has a small amount of data available. To overcome the dilemma, we propose a virtual screening model, in which word2vec is used as an embedding tool to generate low-dimensional vectors of SMILES of compounds and amino acid sequences of proteins, and the modified multi-grained cascade forest based gcForest is used as the classifier. This proposed method is capable of constructing a model from raw data, adjusting model complexity according to the scale of datasets, especially for small scale datasets, and is robust with few hyper-parameters and without over-fitting. We found that the proposed model is superior to other CPI prediction models and performs well on the constructed challenging dataset. We finally predicted 2 new inhibitors for clusters of differentiation 47(CD47) which has few known inhibitors. The IC50s of enzyme activities of these 2 new small molecular inhibitors targeting CD47-SIRPα interaction are 3.57 and 4.79μM respectively. These results fully demonstrate the competence of this concise but efficient tool for CPI prediction.

RAVAR: a curated repository for rare variant-trait associations.

Rare variants contribute significantly to the genetic causes of complex traits, as they can have much larger effects than common variants and account for much of the missing heritability in genome-wide association studies. The emergence of UK Biobank scale datasets and accurate gene-level rare variant-trait association testing methods have dramatically increased the number of rare variant associations that have been detected. However, no systematic collection of these associations has been carried out to date, especially at the gene level. To address the issue, we present the Rare Variant Association Repository (RAVAR), a comprehensive collection of rare variant associations. RAVAR includes 95047 high-quality rare variant associations (76186 gene-level and 18861 variant-level associations) for 4429 reported traits which are manually curated from 245 publications. RAVAR is the first resource to collect and curate published rare variant associations in an interactive web interface with integrated visualization, search, and download features. Detailed gene and SNP information are provided for each association, and users can conveniently search for related studies by exploring the EFO tree structure and interactive Manhattan plots. RAVAR could vastly improve the accessibility of rare variant studies. RAVAR is freely available for all users without login requirement at http://www.ravar.bio.

ILSGRN: inference of large-scale gene regulatory networks based on multi-model fusion.

Gene regulatory networks (GRNs) are a way of describing the interaction between genes, which contribute to revealing the different biological mechanisms in the cell. Reconstructing GRNs based on gene expression data has been a central computational problem in systems biology. However, due to the high dimensionality and non-linearity of large-scale GRNs, accurately and efficiently inferring GRNs is still a challenging task. In this article, we propose a new approach, iLSGRN, to reconstruct large-scale GRNs from steady-state and time-series gene expression data based on non-linear ordinary differential equations. Firstly, the regulatory gene recognition algorithm calculates the Maximal Information Coefficient between genes and excludes redundant regulatory relationships to achieve dimensionality reduction. Then, the feature fusion algorithm constructs a model leveraging the feature importance derived from XGBoost (eXtreme Gradient Boosting) and RF (Random Forest) models, which can effectively train the non-linear ordinary differential equations model of GRNs and improve the accuracy and stability of the inference algorithm. The extensive experiments on different scale datasets show that our method makes sensible improvement compared with the state-of-the-art methods. Furthermore, we perform cross-validation experiments on the real gene datasets to validate the robustness and effectiveness of the proposed method. The proposed method is written in the Python language, and is available at: https://github.com/lab319/iLSGRN.

A Machine Learning Framework Integrating Electronic Health Records for Survival Prognostication in Cervix Cancer

Recent advances in Machine learning (ML) have led to promising clinical applications in oncology, such as improved detection on imaging, predicting emergency department visits, and predicting survival. The hypothesis of this study is a machine learning model using features extracted from electronic health records can be created to improve individual prognostication in patients with cervix cancer managed with radiotherapy in a binary higher and lower risk group model. This study used a single-institution retrospective dataset. Electronic records of patients treated for cervix cancer with definitive radiotherapy from 2003-2014 were queried. Features were selected for model building were based on relevant clinical variables and ease of abstraction from records: age, race, smoking status, parity, self-reported unintentional weight loss, histology, revised 2018 Fédération Internationale de Gynécologie et d'Obstétrique (FIGO) staging, tumor size, number of enlarged lymph nodes, involved node location, presence of metastases, radiotherapy duration, and concurrent chemotherapy. The dataset was split into training and testing cohorts, with multivariable Cox regression with Lasso regularization performed to predict hazard ratios (HR), and internal 5-fold cross-validation (training set) to determine the regularization parameter. For each patient, a partial HR using all variables was predicted. A cut-off value maximally splitting the training dataset into higher and lower risk cohorts was calculated in the training dataset using Kaplan-Meier estimates and log-rank tests, and this predicted cut-off value was evaluated in the test dataset. A total of 226 patients were included in study, with a median follow-up of 55.1 months; 226 patients in training cohort, and 57 in testing cohort. Feature concordance indices of 0.74 and 0.75 were obtained in the testing and training datasets, respectively, with minimal over-fitting. The three variables that contributed most to the model (with Log HR, 95% CI, p-value) were FIGO stage (0.38, 0.1-0.66, p = 0.01), presence of metastases (0.26, -0.02-0.53, p = 0.07), and unintentional weight loss (-0.34, -0.54-0.14, p<0.01), and FIGO IIB maximally divided the cohort into higher and lower risk stage groupings. A partial HR of 1.15 threshold using all features predicted the largest survival separation between higher and lower risk patients, more so than stage alone (p = 1.18 × 10-3 separated by stage alone vs p = 8.97 × 10-7 separated by partial hazard). A machine learning approach was able to improve prognostication of survival in cervix cancer using additional features from electronic health records. Future work should explore prognostication using large scale datasets and pre-treatment variables for potential incorporation into patient discussions and shared decision making.

BK-SAD: A Large Scale Dataset for Student Activity Recognition

Skeleton-based human action recognition has emerged as a prominent research topic in the field of artificial intelligence due to its broad applicability in a wide range of domains, including but not limited to healthcare, security and surveillance, entertainment, and intelligent environments. In this paper, we propose a novel data collection methodology and present BK-Student Activity Dataset (BK-SAD), a new 2D dataset for student activity recognition in smart classrooms that outperforms the existing NTU RGB+D 120 dataset, SBU Kinect Interaction dataset. Our dataset contains three classes: hand raising, doze-off, and normal activities. The dataset was collected using cameras placed in real classroom environments and consisted of video data from multiple viewpoints. The dataset contains over 2700 videos of students raising their hands, over 1700 videos of students dozing off during class, and over 8500 videos of normal activities. In addition, to evaluate the effectiveness of the proposed dataset, we give some baseline performance figures for neural network architectures trained and tested for student activity recognition on BK-SAD dataset. These ConvNet architectures demonstrate significant performance improvement on the proposed dataset. The effectiveness of the proposed novel data collection methodology and BK-SAD dataset in this paper will enable further research and development of activity recognition models for classroom environments, with potential applications in smart education and intelligent classroom management systems.

Reducing Video Coding Complexity Based on CNN-CBAM in HEVC

High-efficiency video coding (HEVC) outperforms H.264 in coding efficiency. However, the rate–distortion optimization (RDO) process in coding tree unit (CTU) partitioning requires an exhaustive exploration of all possible quad-tree partitions, resulting in high encoding complexity. To simplify this process, this paper proposed a convolution neural network (CNN) based optimization algorithm combined with a hybrid attention mechanism module. Firstly, we designed a CNN compatible with the current coding unit (CU) size to accurately predict the CU partitions. In addition, we also designed a convolution block to enhance the information interaction between CU blocks. Then, we introduced the convolution block attention module (CBAM) into CNN, called CNN-CBAM. This module concentrates on important regions in the image and attends to the target object correctly. Finally, we integrated the CNN-CBAM into the HEVC coding framework for CU partition prediction in advance. The proposed network was trained, validated, and tested using a large scale dataset covering various scenes and objects, which provides extensive samples for intra-frame CU partition prediction in HEVC. The experimental findings demonstrate that our scheme can reduce the coding time by up to 64.05% on average compared to a traditional HM16.5 encoder, with only 0.09 dB degradation in BD-PSNR and a 1.94% increase in BD-BR.

Twin support vector quantile regression

We propose a twin support vector quantile regression (TSVQR) to capture the heterogeneous and asymmetric information in modern data. Using a quantile parameter, TSVQR effectively depicts the heterogeneous distribution information with respect to all portions of data points. Correspondingly, TSVQR constructs two smaller sized quadratic programming problems (QPPs) to generate two nonparallel planes to measure the distributional asymmetry between the lower and upper bounds at each quantile level. The QPPs in TSVQR are smaller and easier to solve than those in previous quantile regression methods. Moreover, the dual coordinate descent algorithm for TSVQR also accelerates the training speed. Experimental results on six artificial data sets, five benchmark data sets, two large scale data sets, two time-series data sets, and two imbalanced data sets indicate that the TSVQR outperforms previous quantile regression methods in terms of the effectiveness of completely capturing the heterogeneous and asymmetric information and the efficiency of the learning process.

CEModule: A Computation Efficient Module for Lightweight Convolutional Neural Networks.

Lightweight convolutional neural networks (CNNs) rely heavily on the design of lightweight convolutional modules (LCMs). For an LCM, lightweight design based on repetitive feature maps (LoR) is currently one of the most effective approaches. An LoR mainly involves an extraction of feature maps from convolutional layers (CE) and feature map regeneration through cheap operations (RO). However, existing LoR approaches carry out lightweight improvements only from the aspect of RO but ignore the problems of poor generalization, low stability, and high computation workload incurred in the CE part. To alleviate these problems, this article introduces the concept of key features from a CNN model interpretation perspective. Subsequently, it presents a novel LCM, namely CEModule, focusing on the CE part. CEModule increases the number of key features to maintain a high level of accuracy in classification. In the meantime, CEModule employs a group convolution strategy to reduce floating-point operations (FLOPs) incurred in the training process. Finally, this article brings forth a dynamic adaptation algorithm (α-DAM) to enhance the generalization of CEModule-enabled lightweight CNN models, including the developed CENet in dealing with datasets of different scales. Compared with the state-of-the-art results, CEModule reduces FLOPs by up to 54% on CIFAR-10 while maintaining a similar level of accuracy in classification. On ImageNet, CENet increases accuracy by 1.2% following the same FLOPs and training strategies.