检索结果-内蒙古大学图书馆

Exploiting Multi-Level Data Uncertainty for Japanese-Chinese Neural Machine Translation

IEICE Transactions on Information and Systems 2025年第5期E108.D卷 440-443页

作者： Li, Zezhong Ma, Jianjun Ren, Fuji Faculty of Computer Science Zhejiang University of Water Resources and Electric Power China Faculty of Foreign Languages Dalian University of Technology China College of Computer Science and Engineering The Shenzhen Institute for Advanced Study University of Electronic Science and Technology of China China

The performance of Neural Machine Translation (NMT) heavily depends on the severity of data uncertainty existing in the training examples. In terms of its causes, data uncertainty can be categorized into intrinsic and extrinsic uncertainty, both of which can increase the learning difficulty in NMT, and lead to degradation in translation performance. To cope with this challenge, we propose a simple yet effective method to estimate the data uncertainty and incorporate it into the adaptive training of NMT, which can mitigate the hurt brought by the uncertain data. Our method consists of two modules: 1) we propose a mBERT based model for estimating token-level and sentence-level data uncertainties jointly, which is trained with multi-task learning;2) we propose heterogeneous ways of incorporating the derived multi-level uncertainties into the NMT training by using soft-masked embedding and weighted loss. Extensive experiments on Japanese↔Chinese translation show that our proposed methods substantially outperform the strong baselines in terms of BLEU scores, and verify the effectiveness of modeling data uncertainty in NMT. Copyright © 2025 The Institute of Electronics, Information and Communication Engineers.

关键词： Neural machine translation

来源：评论

学校读者我要写书评

暂无评论

Multisource Data Fusion Using MLP for Human Activity Recognition

引用

computers, Materials & Continua 2025年第2期82卷 2109-2136页

作者： Sujittra Sarakon Wansuree Massagram Kreangsak Tamee Department of Computer Science and Information Technology Faculty of ScienceNaresuan UniversityPhitsanulok65000Thailand Center of Excellence for Innovation and Technology for Detection and Advanced Materials(ITDAM) Faculty of ScienceNaresuan UniversityPhitsanulok65000Thailand Center of Excellence in Nonlinear Analysis and Optimizations Faculty of ScienceNaresuan UniversityPhitsanulok65000Thailand

This research investigates the application of multisource data fusion using a Multi-Layer Perceptron (MLP) for Human Activity Recognition (HAR). The study integrates four distinct open-source datasets—WISDM, DaLiAc, MotionSense, and PAMAP2—to develop a generalized MLP model for classifying six human activities. Performance analysis of the fused model for each dataset reveals accuracy rates of 95.83 for WISDM, 97 for DaLiAc, 94.65 for MotionSense, and 98.54 for PAMAP2. A comparative evaluation was conducted between the fused MLP model and the individual dataset models, with the latter tested on separate validation sets. The results indicate that the MLP model, trained on the fused dataset, exhibits superior performance relative to the models trained on individual datasets. This finding suggests that multisource data fusion significantly enhances the generalization and accuracy of HAR systems. The improved performance underscores the potential of integrating diverse data sources to create more robust and comprehensive models for activity recognition.

关键词： Multisource data fusion human activity recognition multi-layer perceptron(MLP) artificial intelligent

来源：评论

学校读者我要写书评

暂无评论

Weighted rank aggregation based on ranker accuracies for feature selection

引用

Soft Computing 2025年第4期29卷 1981-2001页

作者： Abdolrazzagh-Nezhad, Majid Kherad, Mahdi Faculty of Engineering Department of Computer Engineering Bozorgmehr University of Qaenat Qaen Iran Department of Computer Science Faculty of Computer and Industrial Engineering Birjand University of Technology Birjand Iran Faculty of Engineering Department of Computer Engineering University of Qom Qom Iran

Rank aggregation is the combination of several ranked lists from a set of candidates to achieve a better ranking by combining information from different sources. In feature selection problem, due to the heterogeneity of methods, there are some base rankers (Filter-based methods) that are of diverse quality and usually the ground truth of ratings is not available. Existing rank aggregation methods that take the diverse quality of base rankers into account do not have any explicit approach for appropriate weighting, require prior assumptions, and suffers from high computational complexity. In this paper, to overcome these challenges, an efficient unsupervised method is introduced for estimating the base rankers’ qualities and aggregating the rankers based on the estimated weights. We first compute the ratio of disagreement between base rankers in ordering different element pairs and then estimate the accuracies in a way that to minimize the discrepancy between these computed ratios and their analytical counterparts. We use the weighted majority voting method for obtaining the aggregated results. To resolve the probable inconsistencies in the final aggregation, the result is formed as a graph, and a greedy algorithm is used to find an acyclic subgraph with the highest weigh. To demonstrate the performance of the proposed method, nine standard UCI datasets are used. The obtained results by the proposed method have higher values of classifier measures than the existing baseline Feature Selection methods and rank aggregation-based multi-filter methods in the most datasets. The experiments show that rank aggregation-based Feature Selection methods outperform individual methods. The proposed method also shows the weight of each Filter-based Feature Selection method, in which the MRMR method has a higher weight than other methods. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Wiener filtering

来源：评论

学校读者我要写书评

暂无评论

CNN-HOG based hybrid feature mining for classification of coffee bean varieties using image processing

引用

Multimedia Tools and Applications 2025年第2期84卷 749-764页

作者： Molla, Yirga Kene Mitiku, Emebet Abeje Department of Computer Science Gafat Institute of Technology Debre Tabor University Debre Tabor Ethiopia Department of Computer Science Faculty of Natural and Computational Science Debark University Debark Ethiopia

Ethiopia, known as the birthplace of coffee, relies on coffee exports as a major source of foreign currency. This research paper focuses on developing a hybrid feature mining technique to automatically classify Ethiopian coffee beans based on their provenance: Harrar, Jimma, Limu, Sidama, and Wellega, which correspond to their botanical origins. A dataset of coffee bean images is collected from various regions through the Ethiopian Commodity Exchange (ECX) in Addis Ababa. The proposed system incorporates preprocessing phases including image resizing, filtering, contrast enhancement, noise removal, grayscale conversion, and segmentation using a combined thresholding and K-means approach for grayscale and RGB images, respectively. Classification is performed using a radial basis function (RBF) kernel function of support vector machine (SVM). To address the color-feature similarity challenge, the study explores merging color and texture features using the histogram of oriented gradients (HOG) local feature descriptor. Performance evaluation is conducted for HOG feature extraction, CNN feature extraction, and a hybrid feature vector (HOG-CNN) using a multi-class SVM classifier, achieving accuracies of 74.17%, 85.83%, and 97.5%, respectively. The deep-shallow-based feature (CNN-HOG) combination demonstrates the highest accuracy of 97.5% in this study. The findings highlight the effectiveness of the proposed hybrid feature mining approach in automatically classifying Ethiopian coffee bean varieties, with potential applications in quality control and traceability within the coffee industry. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Robust security risk estimation for android apps using nearest neighbor approach and hamming distance

引用

Soft Computing 2025年第2期29卷 593-611页

作者： Deypir, Mahmood Zoughi, Toktam Faculty of Computer Engineering Shahid Sattari Aeronautical University of Science and Technology Tehran Iran Tehran Iran

Nowadays, Android-based devices such as smart phones, tablets, smart watches, and virtual reality headsets have found increasing use in our daily lives. Along with the development of various applications for these devices, new malicious apps are released by intruders, which are more difficult to identify and deal with because they exploit more sophisticated techniques. Although methods have been provided to calculate the security risk and identify malicious apps in Android operating system, but with the expansion of the level and depth of the threats, the need for more effective methods in this context is still required. In this paper, we have devised a new algorithm to calculate the security risk of Android apps, which can be used to identify malicious apps from benign ones. In this algorithm, to estimate the security risk of an unknown input app, its nearest neighbors to malicious apps and its nearest neighbors to normal apps are computed separately using Hamming distance. Then, the security risk of the input app can be computed using a simple formulation. After implementing this algorithm, its parameter for the number of neighbors using various real datasets is adjusted. The extensive experiments conducted on these data show the superiority of the proposed method over the previously proposed ones in terms of detection rate. Our additional experimentations show the robustness of the proposed algorithm in adversarial situations. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Android malware

来源：评论

学校读者我要写书评

暂无评论

VoteDroid: a new ensemble voting classifier for malware detection based on fine-tuned deep learning models

引用

Multimedia Tools and Applications 2025年第12期84卷 10923-10944页

作者： Bakır, Halit Faculty of Engineering and Natural Sciences Department of Computer Engineering Sivas University of Science and Technology Sivas Turkey

In this work, VoteDroid a novel fine-tuned deep learning models-based ensemble voting classifier has been proposed for detecting malicious behavior in Android applications. To this end, we proposed adopting the random search optimization algorithm for deciding the structure of the models used as voter classifiers in the ensemble classifier. We specified the potential components that can be used in each model and left the random search algorithm taking a decision about the structure of the model including the number of each component that should be used and its location in the structure. This optimization method has been used to build three different deep learning models namely CNN-ANN, pure CNN, and pure ANN. After selecting the best structure for each DL model, the selected three models have been trained and tested using the constructed image dataset. Afterward, we suggested hybridizing the fine-tuned three deep-learning models to form one ensemble voting classifier with two different working modes namely MMR (Malware Minority Rule) and LMR (Label Majority Rule). To our knowledge, this is the first time that an ensemble classifier has been fine-tuned and hybridized in this way for malware detection. The results showed that the proposed models were promising, where the classification accuracy exceeded 97% in all experiments. © The Author(s) 2024.

关键词： Android malware

来源：评论

学校读者我要写书评

暂无评论

Research Progress in Solar Flare Prediction Methods

引用

Research in Astronomy and Astrophysics 2025年第3期25卷 280-309页

作者： Ke Han Zhen Liu Xian-Yi Zhao Yi-Fei Li De-Quan Zheng Jie Wan School of Computer and Information Engineering Harbin University of Commerce Faculty of Computing Harbin Institute of Technology School of Energy Science and Engineering Harbin Institute of Technology

Solar flares are one of the strongest outbursts of solar activity,posing a serious threat to Earth’s critical infrastructure,such as communications,navigation,power,and ***,it is essential to accurately predict solar flares in order to ensure the safety of human ***,the research focuses on two directions:first,identifying predictors with more physical information and higher prediction accuracy,and second,building flare prediction models that can effectively handle complex observational *** terms of flare observability and predictability,this paper analyses multiple dimensions of solar flare observability and evaluates the potential of observational parameters in *** flare prediction models,the paper focuses on data-driven models and physical models,with an emphasis on the advantages of deep learning techniques in dealing with complex and high-dimensional *** reviewing existing traditional machine learning,deep learning,and fusion methods,the key roles of these techniques in improving prediction accuracy and efficiency are *** prevailing challenges,this study discusses the main challenges currently faced in solar flare prediction,such as the complexity of flare samples,the multimodality of observational data,and the interpretability of *** conclusion summarizes these findings and proposes future research directions and potential technology advancement.

关键词： (Sun:) sunspots magnetohydrodynamics (MHD) Sun: activity Sun: flares Sun: magnetic fields

来源：评论

学校读者我要写书评

暂无评论

Advanced Machine Learning Approaches for Breast Cancer Detection with Neutrosophic Sets

引用

Neutrosophic Sets and Systems 2025年 81卷 273-284页

作者： Elbehiery, Hussam Fathi, Hanaa Eassa, Mohamed Abdelhafeez, Ahmed Abdellah, Mohamed Refaat Mahmoud, Hadeer Computer Science Department Faculty of Information System and Computer Science October 6 University Giza12585 Egypt College of Information Technology Amman Arab University Amman Jordan Applied Science Research Center Applied Science Private University Amman Jordan The Department of Computer Science College of Information Technology Misr University for Science and Technology Cairo Egypt Department of Artificial Intelligence Faculty of Computers and Artificial Intelligence Modern University for Technology & Information Cairo Egypt

Breast Cancer (BC) remains a significant health challenge for women and is one of the leading causes of mortality worldwide. Accurate diagnosis is critical for successful therapy and increased survival rates. Recent advances in medical imaging and computational technologies have enabled more precise methods of detecting and evaluating breast cancer. Accurate analysis and diagnosis utilizing medical imaging have developed as essential research topics, providing important help in clinical decision-making for various illnesses, including breast cancer. Machine learning (ML) can accurately predict breast cancer. But the breast cancer data has vague and uncertainty information. So, the neutrosophic sets (NSs) are used in this study to deal with uncertainty data. We convert the original dataset into neutrosophic data with three components such as truth, indeterminacy, and falsity values. Then we applied four ML models with N-data such as logistic regression, gradient boosting (GB), k-nearest neighbor (KNN), and support vector machines (SVM), to improve diagnostic accuracy. Then we compared the ML models with and without using N-data. The results show the logistic regression has higher accuracy with 98.6% with the N-data and 95.80% without N-data. So, the NSs can improve the accuracy of ML models. © 2025, University of New Mexico. All rights reserved.

关键词： Lung cancer

来源：评论

学校读者我要写书评

暂无评论

An Efficient Two-State Feature Attention-Based GRU for Text Classification

Informatica (Slovenia)

引用

Informatica (Slovenia) 2025年第15期49卷 95-112页

作者： Mohammed, Dhurgham Ali Patel, Kalyani A. Faculty of Computer Science Gujarat University Ahmedabad India Department of Computer Science Faculty of Education for Girl University of Kufa Iraq K.S. School of Business Management and Information Technology Gujarat University India

Text classification has become crucial for mechanically sorting documents into specific categories. The goal of classification is to assign a predefined group or class to an instance based on its characteristics. To attain precise text categorization, a feature selection scheme is employed to categorize significant features and eliminate irrelevant, undesirable, and noisy ones, thereby reducing the dimensionality of the feature space. Many advanced deep learning algorithms have been developed to handle text classification drawbacks. Recurrent neural networks (RNNs) are broadly employed in text classification tasks. In this paper, we referred to a novel Two-state GRU based on a Feature Attention strategy, known as Two-State Feature Attention GRU (TS-FA-GRU). The proposed framework identifies and categorizes word polarity through consecutive mechanisms and word-feature capture. Furthermore, the developed study incorporates a pre-feature attention TS-FA-GRU to capture essential features at an early stage, followed by a post-feature attention GRU that mimics the decoder’s function to refine the extracted features. To enhance computational performance, the reset gate in the ordinary GRU is replaced with an update gate, which helps to reduce redundancy and complexity. The effectiveness of the developed model was tested on five benchmark text datasets and compared with five well-established traditional text classification methods. The proposed TS-FA-GRU model demonstrated superior performance over several traditional approaches regarding convergence rate and accuracy. Experimental outcomes revealed that the TS-FA-GRU model achieved excellent text classification accuracies of 93.86%, 92.69%, 94.73%, 92.46%, and 88.23 on the 20NG, R21578, AG News, IMDB, and Amazon review dataset respectively. Moreover, the results indicated that the proposed model effectively minimized the loss function and captured long-term dependencies, leading to exceptional outcomes when compared to the

关键词： Recurrent neural networks

来源：评论

学校读者我要写书评

暂无评论

An Improved Hybrid Deep Learning Approach for Security Requirements Classification

引用

computers, Materials & Continua 2025年第3期82卷 4041-4067页

作者： Shoaib Hassan QianmuLi Muhammad Zubair Rakan AAlsowail Muhammad Umair School of Computer Science and Engineering Nanjing University of Science and TechnologyNanjing210094China Faculty of Information Technology and Computer Science University of Central PunjabLahore5400Pakistan Computer Skills Self-Development Skills DevelopmentDeanship of Common First YearKing Saud UniversityRiyadh11362Saudia Arabia

As the trend to use the latestmachine learning models to automate requirements engineering processes continues,security requirements classification is tuning into the most researched field in the software engineering *** literature studies have proposed numerousmodels for the classification of security ***,adopting those models is constrained due to the lack of essential datasets permitting the repetition and generalization of studies employing more advanced machine learning ***,most of the researchers focus only on the classification of requirements with security *** did not consider other nonfunctional requirements(NFR)directly or indirectly related to *** has been identified as a significant research gap in security requirements *** major objective of this study is to propose a security requirements classification model that categorizes security and other relevant security *** use PROMISE_exp and DOSSPRE,the two most commonly used datasets in the software engineering *** proposed methodology consists of two *** the first step,we analyze all the nonfunctional requirements and their relation with security *** found 10 NFRs that have a strong relationship with security *** the second step,we categorize those NFRs in the security requirements *** proposedmethodology is a hybridmodel based on the ConvolutionalNeural Network(CNN)and Extreme Gradient Boosting(XGBoost)***,we evaluate the model by updating the requirement type column with a binary classification column in the dataset to classify the requirements into security and non-security *** performance is evaluated using four metrics:recall,precision,accuracy,and F1 Score with 20 and 28 epochs number and batch size of 32 for PROMISE_exp and DOSSPRE datasets and achieved 87.3%and 85.3%accuracy,*** proposed study shows an enhancement in metrics

关键词： Requirements engineering security requirements deep learning CNN XGBoost classification

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：