Feature Weights Research Articles

GeometryFormer: Semi-Convolutional Transformer Integrated with Geometric Perception for Depth Completion in Autonomous Driving Scenes

Depth completion is widely employed in Simultaneous Localization and Mapping (SLAM) and Structure from Motion (SfM), which are of great significance to the development of autonomous driving. Recently, the methods based on the fusion of vision transformer (ViT) and convolution have brought the accuracy to a new level. However, there are still two shortcomings that need to be solved. On the one hand, for the poor performance of ViT in details, this paper proposes a semi-convolutional vision transformer to optimize local continuity and designs a geometric perception module to learn the positional correlation and geometric features of sparse points in three-dimensional space to perceive the geometric structures in depth maps for optimizing the recovery of edges and transparent areas. On the other hand, previous methods implement single-stage fusion to directly concatenate or add the outputs of ViT and convolution, resulting in incomplete fusion of the two, especially in complex outdoor scenes, which will generate lots of outliers and ripples. This paper proposes a novel double-stage fusion strategy, applying learnable confidence after self-attention to flexibly learn the weight of local features. Our network achieves state-of-the-art (SoTA) performance with the NYU-Depth-v2 Dataset and the KITTI Depth Completion Dataset. It is worth mentioning that the root mean square error (RMSE) of our model on the NYU-Depth-v2 Dataset is 87.9 mm, which is currently the best among all algorithms. At the end of the article, we also verified the generalization ability in real road scenes.

Multimodal sleep staging network based on obstructive sleep apnea

BackgroundAutomatic sleep staging is essential for assessing sleep quality and diagnosing sleep disorders. While previous research has achieved high classification performance, most current sleep staging networks have only been validated in healthy populations, ignoring the impact of Obstructive Sleep Apnea (OSA) on sleep stage classification. In addition, it remains challenging to effectively improve the fine-grained detection of polysomnography (PSG) and capture multi-scale transitions between sleep stages. Therefore, a more widely applicable network is needed for sleep staging.MethodsThis paper introduces MSDC-SSNet, a novel deep learning network for automatic sleep stage classification. MSDC-SSNet transforms two channels of electroencephalogram (EEG) and one channel of electrooculogram (EOG) signals into time-frequency representations to obtain feature sequences at different temporal and frequency scales. An improved Transformer encoder architecture ensures temporal consistency and effectively captures long-term dependencies in EEG and EOG signals. The Multi-Scale Feature Extraction Module (MFEM) employs convolutional layers with varying dilation rates to capture spatial patterns from fine to coarse granularity. It adaptively fuses the weights of features to enhance the robustness of the model. Finally, multiple channel data are integrated to address the heterogeneity between different modalities effectively and alleviate the impact of OSA on sleep stages.ResultsWe evaluated MSDC-SSNet on three public datasets and our collection of PSG records of 17 OSA patients. It achieved an accuracy of 80.4% on the OSA dataset. It also outperformed the state-of-the-art methods in terms of accuracy, F1 score, and Cohen's Kappa coefficient on the remaining three datasets.ConclusionThe MSDC-SSRNet multi-channel sleep staging architecture proposed in this study enhances widespread system applicability by supplementing inter-channel features. It employs multi-scale attention to extract transition rules between sleep stages and effectively integrates multimodal information. Our method address the limitations of single-channel approaches, enhancing interpretability for clinical applications.

Fast lane detection by fusing multi-scale contour feature extraction and weight mean normalization

Fast lane detection by fusing multi-scale contour feature extraction and weight mean normalization

A Model for Defining the Boundaries of Complex Sentences and Clauses in English Translation Based on the Huffman Tree and Objective Function

To address the issues of high redundancy in sentence features and poor semantic similarity analysis in English translation, a boundary definition model for complex sentence clauses in English translation based on the Huffman tree and objective function is proposed. This model analyzes the boundary structure of complex sentences and clauses in English translation, determines the mathematical expected value of the probability distribution of boundary features, introduces the maximum entropy model to assess the key degree of different features, and utilizes conditional random fields to mine hidden features. Additionally, by employing a hierarchical clustering algorithm to analyze the distance between boundary features of complex sentences and clauses in English translation, similar images are merged based on the minimum distance between data points. Feature redundancy scores are obtained through an attention mechanism, and the weight of boundary features of complex sentences and clauses in English translation is calculated. By using the edit distance in semantic similarity to determine the boundary distance of clauses, and then using cosine similarity to calculate the similarity between the boundary features of complex clauses in English translation, a Huffman tree and objective function are introduced to construct a model for defining the boundary of complex clauses in English translation. Input the boundary feature values of complex sentences and clauses in English translation to complete the final definition. The experimental results show that the proposed method performs well in the task of defining the boundaries of complex sentences and clauses in English translation, with semantic similarity analysis results remaining above 95% and reaching up to 99%, significantly better than the comparative methods. Meanwhile, the recall curve obtained by the proposed method is closest to the ideal curve and has a small fluctuation range, stable between 90% and 98%, further verifying its accuracy and robustness in boundary delineation. In addition, when the sample data size is 1000, the confidence level of the proposed method is as high as 99.6%, which is higher than the 95.6% and 95.1% of the comparison methods. As the sample size increased to 2500, the confidence level of the proposed method remained at a high level of 99.4%, while the confidence level of the comparative method decreased to 94.2% and 94.1%, respectively. These data results fully demonstrate the effectiveness of the proposed method in reducing redundant interference and improving confidence. In summary, the proposed method has improved the performance of defining the boundaries of complex sentences and clauses in English translation.

Personalized College English Learning Experience Assisted by Artificial Intelligence: An Algorithm-Driven Adaptive Learning Approach

Personalized College English Learning offers customized education based on individual necessities and aims. Students enhance their linguistic ability efficiently over engaging sessions, specific curricula, and professional teachers. This study aims to progress a smart Artificial Intelligence (AI)-based adaptive learning method for enhancing personalized college English learning experience. Our proposed model provides an intelligent detecting device-enabled setting for learning English over big data analysis and machine learning based on a DM approach. Intelligent sensing strategies detention significant statistics from college English students and big data analysis examines the data to produce beneficial insights that increase real-time personalizing of English learning experiences. We suggest an innovative Starling Murmuration fine-tuned Dynamic Weighted Random Forest (SM-DWRF) system for classifying the data. Our model influences perceptions from collective behavior in starling flocks to dynamically alter feature weights in DWRF arrangement. It integrates fine-tuning mechanisms to adaptively weight features, developing classification accurateness and strength in adapted college English learning settings. We implemented our suggested model in Python software. In the assessment stage, we accurately measure the effectiveness of our proposed SM-DWRF model in identifying diverse features of English learning across various parameters. Our experimental findings incontestably showcase the greater performance of our model associated with conventional approaches in classifying content from multimodal statistics. Significantly, we perceive prominent enhancements in reliability and robustness, when acclimating to dynamic learning settings.

MSDFNet: multi-scale detail feature fusion encoder–decoder network for self-supervised monocular thermal image depth estimation

Abstract Currently available thermal image depth estimation methods are difficult to efficiently extract fine multi-scale feature information from thermal images and suffer from the problem of blurring details at the edges of the estimated depth map. To address these challenges, this paper proposes MSDFNet, a multi-scale detail feature fusion encoder–decoder network, for self-supervised monocular thermal image depth estimation. The model is based on a channel expansion hourglass residual lightweight feature encoder, which can capture rich and fine-grained multi-scale feature information with low computational effort. MSDFNet utilizes a detail feature weight evaluation decoder to fuse cross-scale features and reevaluate the importance of each feature, thereby emphasizing critical edge information at multiple scales. Additionally, MSDFNet incorporates a depth consistency loss function, which provides self-supervised signals for the detailed features of thermal images and improves the optimization of network performance. The method is applied to the ViViD++ and MS2 datasets and achieves state-of-the-art depth estimation performance compared to existing state-of-the-art algorithms. In the Indoor Dark scenario of the ViViD++ dataset, the Abs Rel, Sq Rel, RMSE, and RMSE log error metric values of MSDFNet are reduced by 6.71%, 11.92%, 9.09%, and 5.73%, respectively, while the accuracy metric values δ < 1.25 i , i = 1 , 2 , 3 were improved by 4.18%, 1.13%, and 0.2%, respectively. In addition, MSDFNet proves its excellent generalization ability on the MS2 dataset. The Abs Rel and RMSE error values in the night scene are reduced by 45.6% and 30.09%, respectively, and the accuracy δ < 1.25 i , i = 1 , 3 is improved by 20.95% and 1.33%, respectively. The Abs Rel and RMSE values in the rainy day scenario are reduced by 1.33% and 1.21%, respectively, and the accuracy δ < 1.25 i , i = 1 , 3 is improved by 0.24% and 0.83%, respectively.

Interrelated dynamic biased feature selection and classification model using enhanced gorilla troops optimizer for intrusion detection

An Intrusion Detection System (IDS) is a valuable tool for network security since it can identify attacks, intrusions, and other types of illegal access. Excessive and irrelevant data slows down the classification process and eventually weakens the system's capacity to make informed decisions when IDS is monitoring a huge volume of network traffic. Innovative approaches are utilized to create large amounts of data and a lot of network traffic in order to test its effectiveness. A vital step in machine learning is feature selection. By determining which features are most essential for describing the dataset and its initial attributes, feature selection seeks to improve intrusion prediction performance while simultaneously delving deeper into the stored data. Making a feature selection is similar to fixing an optimization problem without a clear definition when users don't know where to begin. Finding the fewest characteristics required to describe the dataset, the original features, and to conduct classification is the primary purpose of feature selection, which also aims to improve prediction performance and acquire a deeper understanding of the stored data. As a result, researchers have been focusing on feature-selection issues recently, especially in light of the massive growth in available databases. Metaheuristic algorithms using a learning model have been the subject of studies to optimize feature selection difficulties. This research uses Enhanced Gorilla Troops Optimizer (EGTO), for enhancing the feature selection process and then performing classification. This research presents a Interrelated Dynamic Biased Feature Selection Model using Enhanced Gorilla Troops Optimizer (IDBFS-EGTO) for generation of feature vector set for intrusion detection. Despite its apparent success in handling a wide range of practical problems, it risks getting mired in local optima and premature convergence when faced with more difficult optimization challenges that is overcome with EGTO. The EGTO approach, which uses a collection of operators to strike a more steady equilibrium between exploitation and exploration. The proposed model generates relevant feature subset for machine learning model for accurate detection and classification of intrusions in the network. The proposed model achieved 98.4 % accuracy in intrusion detection and 98.6 % accuracy in EGTO optimization classification. The proposed model is improved by 3.8 % in feature weight allocation accuracy and 1.2 % in detection accuracy levels. The proposed model is compared with the traditional models and the results represent that the proposed model performance is high.

Multi-Layered Algorithm for Chinese Cultural Dissemination Power Based on Feature Selection and Weight Learning

Multi-Layered Algorithm for Chinese Cultural Dissemination Power Based on Feature Selection and Weight Learning

Enhancing Explainable Artificial Intelligence: Using Adaptive Feature Weight Genetic Explanation (AFWGE) with Pearson Correlation to Identify Crucial Feature Groups

The ‘black box’ nature of machine learning (ML) approaches makes it challenging to understand how most artificial intelligence (AI) models make decisions. Explainable AI (XAI) aims to provide analytical techniques to understand the behavior of ML models. XAI utilizes counterfactual explanations that indicate how variations in input features lead to different outputs. However, existing methods must also highlight the importance of features to provide more actionable explanations that would aid in the identification of key drivers behind model decisions—and, hence, more reliable interpretations—ensuring better accuracy. The method we propose utilizes feature weights obtained through adaptive feature weight genetic explanation (AFWGE) with the Pearson correlation coefficient (PCC) to determine the most crucial group of features. The proposed method was tested on four real datasets with nine different classifiers for evaluation against a nonweighted counterfactual explanation method (CERTIFAI) and the original feature values’ correlation. The results show significant enhancements in accuracy, precision, recall, and F1 score for most datasets and classifiers; this indicates the superiority of the feature weights selected via AFWGE with the PCC over CERTIFAI and the original data values in determining the most important group of features. Focusing on important feature groups elaborates the behavior of AI models and enhances decision making, resulting in more reliable AI systems.

MBI-TCMR: A TRADITIONAL CHINESE FORMULA RECOMMENDATION DEEP LEARNING MODEL BASED ON MULTI-TASK META LEARNING AND BIOLOGICAL NETWORK STRUCTURE

The performance of existing traditional Chinese medicine (TCM) recommendation models is generally poor because of their weak generalization ability, overfitting, and inability to use known biological networks. Therefore, building a TCM recommendation model based on artificial intelligence has currently become an important bioinformatics task. This study aimed to design a multitask meta-learning model with good biological interpretation for TCM formula recommendation (MBI-TCMR) for deep learning regularization. This method was based on the known biological network structure to sparse the deep learning network, solve the overfitting problem of the model, and enhance the biological interpretability of the model. Furthermore, a multi-learning framework based on meta-learning was also proposed. The framework allowed the MBI-TCMR model to mine knowledge of TCM formulas and quickly adapt to different types of TCM formula recommendation tasks. Finally, we used a gradient-based deep learning feature backtracking method to calculate the feature weight for each neuron. This weight could provide valuable explanatory information for researchers to study how the model made its medicine recommendations. We designed three independent experiments. The experimental results showed that the hit ratio (HR), AUC, and recall and precision value of the MBI-TCMR model outperformed the existing TCM formula recommendation models. The MBI-TCMR model’s HR of top 1–10 reached 0.15–0.9 (Gynecologic Disease Dataset). HR was 10 for the MBI-TCMR model, which was an improvement of 11.1% compared with the best baseline model. The bio-enrichment analysis showed that the model exhibited good bio-interpretation. In summary, this study proposed a novel TCM formula recommendation model, which expanded the application of the artificial intelligence model and achieved good results.

Consistency Verification Method of Chinese and Russian Traditional Culture Translation Text Corpus Based on CNN-BiGRU

China and Russia have completely different cultural backgrounds, historical traditions and ways of thinking, which makes it difficult to directly correspond or accurately convey many culturally unique concepts, customs and expressions during the translation process. For example, some idioms, sayings or cultural symbols may have completely different meanings or no equivalents in the two cultures. Therefore, there are problems such as poor verification resulting in the consistency verification of Chinese and Russian traditional cultural external publicity translation text corpus. In order to solve this problem, this paper designs a consistency verification method for Chinese and Russian traditional cultural external publicity translation text corpus based on CNN-BiGRU. Through the part-of-speech correspondence of the Chinese and Russian traditional cultural external publicity translation text corpus, it is represented by vectors, the feature weights of the Chinese and Russian traditional cultural external publicity translation text corpus are calculated, the collocation information between the feature items is determined, and the maximum likelihood of the features is calculated using data smoothing technology. The rules are adjusted to extract the characteristics of the Chinese and Russian traditional cultural external publicity translation text corpus. Preprocess the key feature source documents of each Chinese and Russian traditional culture external publicity translation text corpus, determine the source documents and their corresponding key features of the traditional culture external publicity translation text corpus, and use TF-IDF and IDF to calculate the key feature sources of the corpus. Based on the rarity of document data, the BERT model is introduced as an encoder to determine the key features of the Chinese and Russian traditional cultural external publicity translation text corpus, integrating the CNN-BiGRU algorithm to design a consistency verification algorithm for Chinese and Russian traditional cultural external publicity translation text corpus. Set up the convolution layer, weight sharing layer, pooling layer, and BiGRU verification layer to build a CNN-BiGRU model to achieve consistency verification of the Chinese and Russian traditional cultural external publicity translation text corpus. Experimental results show that this method can improve the consistency verification effect of Chinese and Russian traditional cultural external publicity translation text corpus.

Optimization of energy acquisition system in smart grid based on artificial intelligence and digital twin technology

In response to the low operating speed and poor stability of energy harvesting systems in smart grids, an energy harvesting optimization method based on improved convolutional neural networks and digital twin technology is proposed in the experiment. Firstly, a smart grid data transmission framework integrating digital twin technology is proposed. A digital twin mapping method based on time, data, and topology structure is used to realize the digital twin mapping at the device level of power grid. Through data synchronization and interaction between the physical power grid and the digital twin model, the operational efficiency and reliability of the power grid are improved. Then, the classical convolutional neural network and attention mechanism are used to comprehensively analyze the physical topology data in the smart grid energy acquisition system. The improved lightweight target detection model is combined to monitor the equipment status of the smart grid and extract key features. Simultaneously utilizing convolutional attention mechanism to dynamically adjust the feature weights of channels or spaces, completing the preprocessing of energy harvesting data. Finally, combined with energy harvesting and power grid switching system, the process of energy harvesting and power grid operation are optimized together. On the training and validation sets, when the channels exceeded 60, the proposed method achieved a system energy efficiency of 55% during operation. The system energy efficiency of the other three comparative algorithms was all less than 40%. In practical applications, as the energy transfer loss increased to 1.0, the system throughput increased to 50 bits. The electricity needs of different users were met, and the difference between power allocation and optimal power allocation was small, which was very reasonable. This proves that the research has effectively optimized the energy harvesting system in the smart grid, improving the efficiency and reliability of the system in practical applications of the smart grid. At the same time, in the increasingly severe energy problem, this system can further provide technical references for the utilization of renewable energy and help achieve the goal of sustainable energy.

Unsupervised person re‐identification based on adaptive information supplementation and foreground enhancement

AbstractUnsupervised person re‐identification has attracted vital interest because of its ability to protect privacy, significantly lower the expense of manual annotation, and eliminate the need for data labels. General unsupervised methods train the network only through global features, which causes the fine‐grained information contained in local features to be ignored in the recognition process, resulting in large amounts of label noise and affecting the recognition accuracy. Moreover, more robust pedestrian features can also improve the accuracy of clustering and enable unsupervised person re‐identification to obtain better results. To address these issues, first, a dual‐branch structure was proposed, which separately obtains the global features of the pedestrian and the local features by dividing the global features into a few equal sections. Then, an adaptive information supplementation (AIS) method based on the k‐nearest neighbor algorithm is designed to ascertain each local feature's relevance to the global features, calculating adaptive weight scores for information supplementation. Finally, these weight scores are used to reallocate the weights of the global features in each part, acquiring features that contain more pedestrian information during the representation learning process. These better features are used to reduce label noise to obtain more accurate pseudo‐labels. Second, an adaptive foreground enhancement module (AFEM) was proposed and inserted before clustering to increase the robustness of pedestrian features, which increases the precision of the pseudo‐labels that are produced after clustering. Experiments on Market‐1501, DukeMTMC‐reID, and MSMT17 demonstrate that the proposed method achieves better results than state‐of‐the‐art methods in fully unsupervised person re‐identification tasks.

Aquatic plants detection in crab ponds using UAV hyperspectral imagery combined with transformer-based semantic segmentation model

Aquatic plants provide habitat and food for Chinese mitten crab growth, the identification of aquatic plant species and monitoring of their coverage can provide basic information for the management of aquatic plants, which can help to improve the efficiency of aquaculture. In this study, to address the time-consuming and labour-intensive nature of traditional aquatic plant monitoring in crab ponds relying on manual observation, a classification method for aquatic plant species using unmanned aerial vehicle and hyperspectral imagery (UAV-HSI) technology, combined with an improved semantic segmentation model named SpectralUFormer was reported for the first time. The UAV-HSI data provides a high-quality data source for automatic aquatic plants detection, and the proposed SpectralUFormer integrates hybrid attention block and hybrid cascaded upsampler. Specifically, the hybrid attention block aggregates abundant spectral features in the encoder. In the decoder part, the hybrid cascaded upsampler is designed by incorporating PixelShuffle and G-L MLP block, which together perform the importance calculation and alignment of feature weights. Experimental results show that the SpectralUFormer achieves high-precision classification of aquatic plant species, with an overall accuracy of 93.15% and a Kappa coefficient of 89.14%. This study offers a feasible approach for the automatic identification of aquatic plant species in crab ponds and the estimation of their coverage.

An indoor thermal comfort model for group thermal comfort prediction based on K-means++ algorithm

Predicting indoor thermal comfort plays an essential role in controlling energy consumption in buildings. Existing studies have used supervised machine learning to predict thermal comfort, which were more accurate than traditional models. However, these models required occupants’ subjective feedback for model training, which reduced the accuracy of the model. In this study, a prediction model that didn’t require feedback was proposed for the first time using the K-means++ algorithm based on the ASHRAE Global Thermal Comfort Database II. Firstly, the data quality was improved through feature selection, dimensional processing, and feature weighting. Then the influence of different outlier judgment methods, feature weight and data set size on model accuracy were compared. Finally, the K-means++ algorithm was applied for thermal comfort clustering analysis. The result showed that the model with an accuracy higher than 90 % could be constructed using only three factors (CLO, TA, RH), and the proposed model could predict indoor group thermal comfort reliably, and provide a foundation for the indoor thermal sensation evaluation.

Compact object populations over cosmic time II. Compact object merger rates and masses over redshift from varying initial conditions

ABSTRACT We perform a first study of the impact of varying two components of the initial conditions in binary population synthesis of compact binary mergers – the initial mass function, which is made metallicity- and star formation rate-dependent, and the orbital parameter (orbital period, mass ratio, and eccentricity) distributions, which are assumed to be correlated – within a larger grid of initial condition models also including alternatives for the primary mass-dependent binary fraction and the metallicity-specific cosmic star formation history. We generate the initial populations with the sampling code bossa and evolve them with the rapid population synthesis code compas. We find strong suggestions that the main role of initial conditions models is to set the relative weights of key features defined by the evolution models. In the two models we compare, black hole–black hole (BHBH) mergers are the most strongly affected, which we connect to a shift from the common envelope to the stable Roche lobe overflow formation channels with decreasing redshift. We also characterize variations in the black hole–neutron star (BHNS) and neutron star–neutron star (NSNS) final parameter distributions. We obtain the merger rate evolution for BHBH, BHNS, and NSNS mergers up to $z=10$, and find a variation by a factor of $\sim 50\textnormal {--}60$ in the local BHBH and BHNS merger rates, suggesting a more important contribution from initial conditions than previously thought, and calling for a complete exploration of the initial conditions model permutations.

Research on an Eye Control Method Based on the Fusion of Facial Expression and Gaze Intention Recognition

With the deep integration of psychology and artificial intelligence technology and other related technologies, eye control technology has achieved certain results at the practical application level. However, it is found that the accuracy of the current single-modal eye control technology is still not high, which is mainly caused by the inaccurate eye movement detection caused by the high randomness of eye movements in the process of human–computer interaction. Therefore, this study will propose an intent recognition method that fuses facial expressions and eye movement information and expects to complete an eye control method based on the fusion of facial expression and eye movement information based on the multimodal intent recognition dataset, including facial expressions and eye movement information constructed in this study. Based on the self-attention fusion strategy, the fused features are calculated, and the multi-layer perceptron is used to classify the fused features, so as to realize the mutual attention between different features, and improve the accuracy of intention recognition by enhancing the weight of effective features in a targeted manner. In order to solve the problem of inaccurate eye movement detection, an improved YOLOv5 model was proposed, and the accuracy of the model detection was improved by adding two strategies: a small target layer and a CA attention mechanism. At the same time, the corresponding eye movement behavior discrimination algorithm was combined for each eye movement action to realize the output of eye behavior instructions. Finally, the experimental verification of the eye–computer interaction scheme combining the intention recognition model and the eye movement detection model showed that the accuracy of the eye-controlled manipulator to perform various tasks could reach more than 95 percent based on this scheme.

Luminance decomposition and Transformer based no-reference tone-mapped image quality assessment

Tone-Mapping Operators (TMOs) play a crucial role in converting High Dynamic Range (HDR) images into Tone-Mapped Images (TMIs) with standard dynamic range for optimal display on standard monitors. Nevertheless, TMIs generated by distinct TMOs may exhibit diverse visual artifacts, highlighting the significance of TMI Quality Assessment (TMIQA) methods in predicting perceptual quality and guiding advancements in TMOs. Inspired by luminance decomposition and Transformer, a new no-reference TMIQA method based on deep learning is proposed in this paper, named LDT-TMIQA. Specifically, a TMI will change under the influence of different TMOs, potentially resulting in either over-exposure or under-exposure, leading to structure distortion and changes in texture details. Therefore, we first decompose the luminance channel of a TMI into a base layer and a detail layer that capture structure information and texture information, respectively. Then, they are employed with the TMI collectively as inputs to the Feature Extraction Module (FEM) to enhance the availability of prior information on luminance, structure, and texture. Additionally, the FEM incorporates the Cross Attention Prior Module (CAPM) to model the interdependencies among the base layer, detail layer, and TMI while employing the Iterative Attention Prior Module (IAPM) to extract multi-scale and multi-level visual features. Finally, a Feature Selection Fusion Module (FSFM) is proposed to obtain final effective features for predicting the quality scores of TMIs by reducing the weight of unnecessary features and fusing the features of different levels with equal importance. Extensive experiments on the publicly available TMI benchmark database indicate that the proposed LDT-TMIQA reaches the state-of-the-art level.

Early stroke behavior detection based on improved video masked autoencoders for potential patients

Stroke is the prevalent cerebrovascular disease characterized by significant incidence and disability rates. To enhance the early perceive and detection of potential stroke patients, the early stroke behavior detection based on improved Video Masked Autoencoders (VideoMAE) for potential patients (EPBR-PS) is proposed. The proposed method begins with novel time interval-based sampling strategy, capturing video frame sequences enriched with sparse motion features. On the basis of establishing the masking mechanism for adjacent frames and pixel blocks within these sequences, The EPBR-PS employes pipeline mask strategy to extract spatiotemporal features effectively. Then, the local convolution attention mechanism is designed to capture local dynamic feature information, and central to the EPBR-PS is the integration of local convolutional attention mechanism with VideoMAE's multi-head attention mechanism. This integration facilitates the simultaneous leveraging of global high-level semantics and local dynamic feature information. Dual attention mechanism-based method for the fusion of these global and local features is proposed. After that, the optimal parameters of EPBR-PS were determined through the experiment of learning rate and fusion weights of different features. On the NTU-ST dataset, comparative analysis with eight models demonstrated the superiority of EPBR-PS, evidenced by the average recognition accuracy of 89.61%, surpassing that 1.67% over the benchmark VideoMAE. On the HMDB51 dataset, EPBR-PS has Top1 of 71.31%, which is 0.73% higher than that of the VideoMAE, providing the viable behavior detection for perception early signs of potential stroke in the home environment. This code is available at https://github.com/wang-325/EPBR-PS/.

Graph Representation Learning Enhanced Semi-Supervised Feature Selection

Feature selection is a key step in machine learning by eliminating features that are not related to the modeling target to create reliable and interpretable models. By exploring the potential complex correlations among features of unlabeled data, recently introduced self-supervision-enhanced feature selection greatly reduces the reliance on the labeled samples. However, they are generally based on the autoencoder with sample-wise self-supervision, which can hardly exploit the relations among samples. To address this limitation, this article proposes graph representation learning enhanced semi-supervised feature selection (G-FS) which performs feature selection based on the discovery and exploitation of the non-Euclidean relations among features and samples by translating unlabeled “plain” tabular data into a bipartite graph. A self-supervised edge prediction task is designed to distill rich information on the graph into low-dimensional embeddings, which remove redundant features and noise. Guided by the condensed graph representation, we propose a batch attention feature weight generation mechanism that generates more robust weights according to batch-based selection patterns rather than individual samples. The results show that G-FS achieves significant performance edges in 14 datasets compared to twelve state-of-the-art baselines, including two recent self-supervised baselines. The source code is public available at https://github.com/Icannotnamemyselff/G-FS_Graph_enhacned_feature_selection .