检索结果-内蒙古大学图书馆

46th Annual International Conference of the IEEE engineering in Medicine and Biology Society, EMBC 2024

作者： Tan, Jieyuan Wang, Yiwen Hong Kong Univesrsity of Science and Technology Department of Electronics and Computer Engineering Hong Kong Hong Kong Univesrsity of Science and Technology Department of Electronics and Computer Engineering Department of Chemical and Biological Engineering Hong Kong

ISBN: (纸本)9798350371499

Reinforcement learning (RL)-based brain machine interfaces (BMIs) provide a promising solution for paralyzed people. Enhancing the decoding performance of RL-based BMIs relies on the design of effective reward signals. Inverse reinforcement learning (IRL) offers an approach to infer subjects' own evaluation from the observed behavior. However, applying IRL to extract reward information in complex BMI tasks requires consideration of the dynamics of subjects' goal during the control process. This dynamic nature of subjects' evaluation requires the IRL method to be able to estimate a time varying reward function. Previous IRL methods applied in BMI systems only estimated a static reward function. Existing IRL algorithms for dynamic reward estimation employ optimization methods to approximate the reward map for each state at each time, which demands substantial amounts of data to achieve convergence. In this paper, we propose a dynamic IRL method to estimate the feedback-driven reward of subjects during BMI tasks. We utilize a state-observation model to continuously infer the reward value for each state, with sensory feedback serving as the external input to model the transition process of the reward. We evaluate our proposed method on a simulated BMI fetch task, which is a multistep task with a time varying reward function. Our method demonstrates improved reward estimation close to the ground truth value, and it significantly outperforms the existing dynamic IRL method when the map size exceeds 25(p © 2024 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Analysis and Classification of Fake News Using Sequential Pattern Mining

引用

Big Data Mining and Analytics 2024年第3期7卷 942-963页

作者： M.Zohaib Nawaz M.Saqib Nawaz Philippe Fournier-Viger Yulin He College of Computer Science and Software Engineering Shenzhen UniversityShenzhen 518060 Guangdong Laboratory of Artificial Intelligence and Digital Economy(SZ) Shenzhen 518107China Department of Computer Science Faculty of Computing and Information TechnologyUnivesity of SargodhaSargodha 40100Pakistan

Disinformation,often known as fake news,is a major issue that has received a lot of attention *** researchers have proposed effective means of detecting and addressing *** machine and deep learning based methodologies for classification/detection of fake news are content-based,network(propagation)based,or multimodal methods that combine both textual and visual *** introduce here a framework,called FNACSPM,based on sequential pattern mining(SPM),for fake news analysis and *** this framework,six publicly available datasets,containing a diverse range of fake and real news,and their combination,are first transformed into a proper ***,algorithms for SPM are applied to the transformed datasets to extract frequent patterns(and rules)of words,phrases,or linguistic *** obtained patterns capture distinctive characteristics associated with fake or real news content,providing valuable insights into the underlying structures and commonalities of ***,the discovered frequent patterns are used as features for fake news *** framework is evaluated with eight classifiers,and their performance is assessed with various *** experiments were performed and obtained results show that FNACSPM outperformed other state-of-the-art approaches for fake news classification,and that it expedites the classification task with high accuracy.

关键词： disinformation fake news sequential pattern mining(SPM) frequent patterns classification

来源：评论

学校读者我要写书评

暂无评论

Enhancing Codebook Utilization in VQ Models via E-Greedy Strategy and Balance Loss 21

Enhancing Codebook Utilization in VQ Models via E-Greedy Str...

引用

21st International computer Conference on Wavelet Active Media technology and Information Processing, ICCWAMTIP 2024

作者： Bo, Chen Jianzhi, Liu School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu611731 China

ISBN: (纸本)9798331519254

Recently, multimodal processing has made significant progress through unified representations, with one of the most common approaches being the quantization of different modalities into unified tokens via Vector Quantification (VQ) models. However, our experiments reveal that due to design flaws inherent to VQ models, they fail to fully utilize all quantized codes during the quantization process. To this end, we propose two novel strategies to address the issue of low and imbalanced codebook utilization. 1) The ϵ-Greedy Strategy enhances codebook exploration by introducing randomness in code selection, while 2) the Balance Loss promotes more even distribution of code usage. Experiments prove that our methods achieve a codebook utilization rate of over 95%, significantly improving image reconstruction accuracy and generation diversity by up to 30% compared to baseline models. These strategies are generic and can be applied across different VQ-based architectures. Moreover, they deliver these improvements while maintaining low computational costs, making them highly efficient and impactful for large-scale image generation tasks. © 2024 IEEE.

关键词： Vector quantization

来源：评论

学校读者我要写书评

暂无评论

A Deep-Siamese CNN-based Approach for Identifying Mycosphaerella Coffeeicola Disease in Coffee Foliage 8

A Deep-Siamese CNN-based Approach for Identifying Mycosphaer...

引用

8th International Conference on Electronics, Communication and Aerospace technology, ICECA 2024

作者： Selvanarayanan, Raveena Rajendran, Surendran Jayaraman, Venkatesh Sathish Kumar, P.J. Saveetha Institute of Medical and Technical Sciences Saveetha School of Engineering Department of Computer Science and Engineering Chennai India Chennai Institute of Technology Department of Computer Science and Engineering Chennai India Panimalar Engineering College Department of Computer Science and Engineering Chennai India

ISBN: (纸本)9798350367904

Mycosphaerella Coffeeicola (Black Spot Disease) is the most destructive disease in coffee leaves and affects global production. Diseased coffee leaves can adversely affect coffee quality, discolor, distort, or rot the coffee leaves, and impair the coffee beans' cherry, taste, and aroma. The disease can also affect the quality of the coffee beans, leading to lower market value. The management of the Coffee Production Department of India controls the disease through a combination of cultural practices, chemical control, and the use of resistant coffee varieties. In the existing work, using single-layer perceptron was trained and validated with 100 epochs to obtain an accuracy of 9 5. 7 1 %. In the piece above of work, an infected leaf is accurately diagnosed using deep learning-based models by employing the Deep-Siamese-based CNN Approach, recognized by a deep Siamese Neural network architecture which is also called a Twin Neural network with high illness recognition precision and simplified hyper-parameters to maximize description capacity while avoiding overfitting concerns, using a network with high sickness detection precision and minimal hyper-parameters. The prototype was tested in the lab on natural plants for known sickness instances, and the results matched the model validation datasets. When compared to several deep-learning models, the suggested structure attained an accuracy of 9 6. 8 9 %. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Improved Alzheimer's Disease Early Prediction using Machine Learning Method : A Comprehensive Diagnostic Approach with Random Forest Classifier 4

Improved Alzheimer's Disease Early Prediction using Machine ...

引用

4th International Conference on Soft Computing for Security Applications, ICSCSA 2024

作者： Munawar, Madiha Monika Singh, T. Kishor Kumar Reddy, C. Hanafiah, Marlia Mohd Stanley College of Engineering and Technology for Women Department of Computer Science and Engineering Hyderabad India Universiti Kebangsaan Malaysia Faculty of Science and Technology Malaysia

ISBN: (纸本)9798331515720

Alzheimer's Disease (AD) often affects the elder persons and is the prevalent kind of Dementia. AD has a huge expense, especially when it comes to the treatment. AD is a main reason the older generations die whereas when Alzheimer Disease is detected early there will be minimal deaths and accurate diagnosing likely occurs in most cases. Using anonymized Oasis longitudinal MRI data, the study trains several machine learning algorithms to identify AD patients. Analyzing the test result includes utilizing different learning strategies or algorithms e.g., Ensemble Random Forest (RF), Hierarchical Decision Tree (DT), Logistic Regression (LR), Convolutional Neural Network (CNN), Deep Neural Network binary classifier (DNN), soft voting classifier (SVC), transfer learning approaches (TLA), Support Vector Machine (SVM) and boosting AdaBoost in reference of their Accuracy, Error Rate, AUC (area under the curve), Time Taken and Recall for each method. The obtained results concluded that the proposed Random Forest classifier achieves higher accuracy (92.84%), Error Rate (7.16%), AUC (95%) (Area Under Curve) and Recall (89%). The Random Forest classifier also achieves higher Time Taken that is 483.654 sec. © 2024 IEEE.

关键词： Alzheimer's Disease Decision Trees Deep Learning Dementia Learning Approaches Machine Learning Random Forest

来源：评论

学校读者我要写书评

暂无评论

A two-dimensional MoS_(2) array based on artificial neural network learning for high-quality imaging

引用

Nano Research 2023年第7期16卷 10139-10147页

作者： Long Chen Siyuan Chen Jinchao Wu Luhua Chen Shuai Yang Jian Chu Chengming Jiang Sheng Bi Jinhui Song School of Mechanical Engineering Dalian University of TechnologyDalian 116024China School of Computer Science Southwest Petroleum UniversityChengdu 610500China

As the basis of machine vision,the biomimetic image sensing devices are the eyes of artificial *** recent years,with the development of two-dimensional(2D)materials,many new optoelectronic devices are developed for their outstanding ***,there are still little sensing arrays based on 2D materials with high imaging quality,due to the poor uniformity of pixels caused by material defects and fabrication ***,we propose a 2D MoS_(2)sensing array based on artificial neural network(ANN)*** equipping the MoS_(2)sensing array with a“brain”(ANN),the imaging quality can be effectively *** the test,the relative standard deviation(RSD)between pixels decreased from about 34.3%to 6.2%and 5.49%after adjustment by the back propagation(BP)and Elman neural networks,*** peak signal to noise ratio(PSNR)and structural similarity(SSIM)of the image are improved by about 2.5 times,which realizes the re-recognition of the distorted *** provides a feasible approach for the application of 2D sensing array by integrating ANN to achieve high quality imaging.

关键词： two-dimensional MoS_(2) sensing array artificial neural network individual difference imaging quality

来源：评论

学校读者我要写书评

暂无评论

Automated Underwater Fish Species Recognition using Deep Learning-based Techniques 11th

Automated Underwater Fish Species Recognition using Deep Lea...

引用

11th International Conference on Intelligent Computing and Applications, ICRTC 2023

作者： Anantha Prabha, P. Sachin, S. Srinithish, U. Deva Priya, M. Karthick, S. Department of Computer Science and Engineering Sri Krishna College of Technology Tamil Nadu Coimbatore India Department of Computer Science and Engineering Sri Eshwar College of Engineering Tamil Nadu Coimbatore India Department of Computer Science and Engineering SRM Institute of Science and Technology Delhi-NCR Campus Uttar Pradesh Ghaziabad India

ISBN: (纸本)9789819717231

Knowledge about fish species with continuous monitoring play a dominant role in determining short and long-term effects on ecosystems and generating ways to manage the problem through specific treatment. Categorization and identification of fish species from underwater film are critical for maintaining resources of fisheries and environmental balance. It is difficult to get an accurate result from actual underwater films owing to poor resolution and object motion. Deep Learning (DL) technology is effectively used in a variety of sectors including aquaculture. In order to identify and recognize fishes in the presence of marine background noise, this research offers architecture for fish detection system that includes an updated training model and a multi-scale fish detection module using enhanced Convolutional Neural Network (CNN). The training technique entails gradual training of enhanced CNN by initially focusing on learning tough species followed by gradual learning of new species by incrementally employing knowledge distillation loss while maintaining high performance of previously learned species. This detection model employs a DL strategy based on CNN to improve detection performance on small-sized objects. As a consequence, the issues prevalent in identifying objects underwater are reduced, and performance of the framework is improved. In contrast to current identification methods, the proposed techniques offer better accuracy. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

Abstraction of Design Information from Legacy C++ Program 3

Abstraction of Design Information from Legacy C++ Program

引用

3rd IEEE International Conference on Distributed Computing and Electrical Circuits and Electronics, ICDCECE 2024

作者： Edara, Venkata Sandeep Kulkarni, R.N. Sasi Institute Of Technology And Engineering Affiliated To Vtu Belagavi Department Of Computer Science And Engineering Andhra Pradesh Tadepalligudem India Ballari Institute Of Technology And Management Department Of Computer Science And Engineering Karnataka Ballari India

ISBN: (纸本)9798350318609

During the last three decades, the importance of the application software is tremendously increased. Every organization has automated all their day-to-day activities. For automation purpose they used the procedural or object-oriented programming languages. During the 1990's to 2010 many organizations used C++ programming language for the development of application software. Today because of the changes in Hardware and Software technology the C++ programming is not much widely used. But these applications require trained experts to maintain the system otherwise it becomes legacy system. These Legacy systems are at the heart of many outstanding companies. Although they are still essential to the general operation of the business, their solutions are out-of-date and, in certain circumstances, no longer even supported by the original software developer or vendor. Abstracting any reusable components from these systems is difficult. These systems can be modernized as to stay competitive by adding current sustainable features. In this paper, a methodology is proposed to abstract the design information which is a reusable component from the given input legacy C++ program in form of Control Flow and Data Flow Tables. © 2024 IEEE.

关键词： Application programs

来源：评论

学校读者我要写书评

暂无评论

Sufficient learning: mining denser high-quality pixel-level labels for edge detection

引用

Neural Computing and Applications 2025年第14期37卷 8245-8260页

作者： Chen, Xiaodiao Yang, Wenya Wu, Wen Tao, Xiuting Mao, Xiaoyang School of Computer Science and Technology Hangzhou Dianzi University Hangzhou310018 China School of Media and Design Hangzhou Dianzi University Hangzhou310018 China Department of Computer Science and Engineering University of Yamanashi Kofu400-8511 Japan

The success of deep neural networks can largely be attributed to large-scale datasets with accurate annotations. In many practical applications, labels are annotated by multiple annotators, resulting in ambiguous labels. To mitigate the impact of such labels, existing deep edge detectors tend only to exploit sparsely reliable samples with high-confident labels by simple voting mechanism. However, lacking sufficient pixel-level learnable samples would severely degrade detection performance. To combat this issue, this work designs a sufficient learning framework for edge detection (SLED) to sufficiently explore reliable samples for model learning, which consists of three stages: sample selection, label refinement, and retraining. Firstly, we propose a sample selection scheme founded on history prediction, which aims to identify more reliable samples and locate unreliable regions containing potentially reliable samples. Secondly, we design an efficient label refinement strategy based on a graph convolutional network (GCN) to propagate reliable label information to unreliable nodes. This strategy can generate "new training targets" with denser, reliable labels for all images. Finally, we retrain the network with these new targets, obtaining a more robust model. Experimental results on three widely used benchmarks show that the proposed detector outperforms all compared SOTA methods. Our code will be released at https://***/wenya1994/SLED. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Pixels

来源：评论

学校读者我要写书评

暂无评论

Dynamic domain testing with multi-agent Markov chain Monte Carlo method

引用

Soft Computing 2024年第13-14期28卷 8293-8317页

作者： Golmohammadi, Roshan Parsa, Saeed Zakeri-Nasrabadi, Morteza School of Computer Engineering Iran University of Science and Technology Hengam St. Resalat Sq. Tehran16846-13114 Iran

Path testing is one of the most efficient approaches for covering a program during the test. However, executing a path with a single or limited number of test data does not guarantee that the path is fault-free, specifically in the fault-prone paths. A common solution in these cases is to extract the corresponding domain of the path constraint that covers the path and select the arbitrary numbers of test data from the extracted domain. Many approaches have been proposed to extract and partition the program input domain statically without executing the program under test. However, extracting domain based on the path constraints is inaccurate and time-consuming, specifically in high dimensional inputs. This paper presents a dynamic approach for extracting the input domain of a given path based on the modified Markov chain Monte Carlo method, multi-agent Metropolis algorithm. The algorithm uses sampling and randomly walks around symmetric proposal distribution to find a finite number of points that approximate the target domain. Our approach chooses the highest fault-prone path from the minimum-cost test paths for domain extraction to maximize the fault detection capabilities. Our experiments with ten Python programs reveal a 14% improvement in the accuracy of the extracted domains compared to the state-of-the-art domain extraction approaches. In addition, the mutation score of automatically generated test cases has improved by an average of 17.95% compared to branch and prime-path testing approaches. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

关键词： Extraction

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：