检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Chen, Mingyu Chen, Yiding Sun, Wen Zhang, Xuezhou Department of Electrical & Computer Engineering Boston University United States Department of Computer Science Cornell University United States Faculty of Computing & Data Sciences Boston University United States

Reinforcement Learning from Human Feedback (RLHF) has emerged as a pivotal technique for large language model (LLM) alignment. This paper studies the setting of online RLHF and focus on improving sample efficiency. All existing algorithms in online RLHF, whether doing passive exploration or active exploration, suffer from a sample complexity that scales exponentially with the scale of the reward function. This fundamental limitation hinders their effectiveness in scenarios with heavily skewed preferences, e.g. questions with a unique correct solution. To address this, we introduce Self-Exploring Preference-Incentive Online Preference Optimization (SE-POPO), an online RLHF algorithm that for the first time achieves a sample complexity that scales polynomially with the reward scale, answering an open problem raised by Xie et al. (2024). Theoretically, we demonstrate that the sample complexity of SE-POPO dominates that of existing exploration algorithms. Empirically, our systematic evaluation confirms that SE-POPO is more sample-efficient than both exploratory and non-exploratory baselines, in two primary application scenarios of RLHF as well as on public benchmarks, marking a significant step forward in RLHF algorithm design. The code is available at https://***/MYC000801/SE-POPO. © 2025, CC BY.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Meta Learning based Pneumonia detection using stacking ensemble of CNN Snapshots

Meta Learning based Pneumonia detection using stacking ensem...

引用

2022 IEEE North Karnataka Subsection Flagship International Conference, NKCon 2022

作者： Nithyapriya, S. Keerthika, J. Tamizh Kuzhali, R. Saranya, M.D. Revathi, P. Daniya, T. Bannari Amman Institute of Technology Department of Artificial Intelligence and Data Science Sathyamangalam India Sri Eshwar College of Engineering Department of Computer Science and Engineering Coimbatore India Sree Sastha Institution of Engineering and Technology Department of Computer Science and Engineering Chennai India KPR Institute of Engineering and Technology Department of Electronics and Communication Engineering Coimbatore India Hindusthan College of Engineering and Technology Department of Information Technology Coimbatore India GMR Institute of Technology Department of Information Technology Andhra Pradesh Rajam India

ISBN: (纸本)9781665453424

Pneumonia is a respiratory infection caused by bacteria, virus or fungi. This infection affects the lungs by filling the air sacs(alveoli) with fluid or pus. Chest X-ray(CXR) imaging is the preferred diagnostic examination to locate the inflammation in the lungs. Examining the chest X-rays with high diagnostic accuracy is very difficult for the medical practitioners. To counteract this, computer-aided diagnosis system is used in this study for classifying the Chest X-ray (CXR) images into pneumonia and normal lungs. This research uses transfer learning based stacked ensemble CNN snapshots for the pneumonia prediction from the chest X-ray images. The DenseNet-121 architecture is used as the base learner. To reduce the computational time, the model is trained once to yield several CNN snapshots. The created snapshots give various details regarding the features taken from the CXR. The decision scores of multiple snapshots are combined using the stacking ensemble technique through a metalearner(Random Forest). The proposed ensemble model obtains an accuracy score of 98.36% on the publicly available Kermany dataset. © 2022 IEEE.

关键词： Air

来源：评论

学校读者我要写书评

暂无评论

Machine Learning and Multiple Regression-Based Study of the Medicinal Properties of Food

Machine Learning and Multiple Regression-Based Study of the ...

引用

作者： Jiasheng Zhou Zhihong Fan Wenxin Zhang Fuzhou University College of Computer and Data Science/College of Software Fuzhou University College of Physics and Information Engineering

We study the relationship between food ingredients and hot and cold properties based on the idea that "medicine and food have the same origin". Firstly, we classify foods with known hot and cold properties as typical representatives of flat, warm and cold properties, and use various machine learning algorithms to classify the typical food representatives. In order to further improve the reliability and accuracy of the classification, we applied Bayesian optimization to the SVM and the SVM was able to classify the typical food representatives again with an accuracy of 96.53%. Based on the above findings, we further analysed which chemical components played a key role in the hot and cold properties. We then applied multivariate logistic regression for quantitative analysis, using stepwise forward regression to minimise complete multicollinearity and OLS + robust standard errors to eliminate the effect of heteroscedasticity. Based on the analysis of the coefficients of the regression equations and the significance test results, the conclusions reached were consistent with the qualitative analysis, with the four main components of energy, water, minerals and fat playing a major role in the chilling and heating properties of food. In conjunction with the analysis in the full text, we make recommendations for the development of functional foods where heat and cold are the principles.

关键词： Factor analysis Clustering algorithms Food classification

来源：评论

学校读者我要写书评

暂无评论

From data to Harvest through Machine Learning based Crop Yield Forecasting

From Data to Harvest through Machine Learning based Crop Yie...

引用

Advancements in Smart, Secure and Intelligent Computing (ASSIC), International Conference on

作者： Aadithya Vikram Budarapu John Hyde Gaddam Ramu Boyedi Thoshan Kumer Muthyala Nikhila Kathirisetty Dept. of Information Technology Vardhaman College of Engineering (Affiliated by JNTUH) Hyderabad India Dept. of Computer Science and Data Science Vardhaman College of Engineerin (Affiliated by JNTUH) Hyderabad India

The maintenance and enhancement of dynamic soil characteristics are the primary focus of soil management in agriculture to increase crop productivity. Higher productivity may result from efficient soil control of resources and corrective micronutrient treatments. Using CNN and ’KNN’ algorithms, the “soil land classification and crop prediction system” application was created. In this study, two datasets are used: one to obtain crop prediction and the other for soil land categorization. The kind of soils are trained using CNN (“VGG- 19”) algorithm, and the accuracy of the model is calculated. The trained model is then utilized in the Flask web app to forecast the type of soil. Another data set with nitrogen, phosphorus, potassium, pH, and temperature as features and the class type of crop as a label is used to forecast crop production. These two algorithms are used to create the flask website, which accepts inputs such as soil picture, soil type prediction, and land parameter inputs for crop prediction.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Providing and evaluating a model for big data anonymization streams by using in-memory processing

引用

Knowledge and Information Systems 2025年

作者： Shamsinejad, Elham Banirostam, Hamid BaniRostam, Touraj Pedram, Mir Mohsen Rahmani, Amir Masoud Computer Engineering (Software) Specializing in Data Science Tehran Iran Master of Data Analytics University of Niagara Falls Niagara Falls ON Canada Electrical and Computer Engineering Department Faculty of Engineering Kharazmi University Tehran Iran Future Technology Research Center National Yunlin University of Science and Technology Yunlin Taiwan

Extracting valuable information from vast sources of social networks while protecting confidentiality and preventing data disclosure is a significant challenge in big data environments. Traditional anonymization methods often fall short in handling the volume, variety, and velocity of big data, leading to high data loss and inefficiency. This article addresses these challenges by proposing a novel anonymization method based on K-means clustering within the Spark framework, leveraging its in-memory processing capabilities. Our model uses K-means clustering to determine optimal cluster heads, significantly reducing data loss and identity disclosure risks. By utilizing Spark's RDD abilities and the MLlib component, our method achieves faster processing times compared to traditional methods that rely on non-in-memory big data tools. Performance evaluation demonstrates that at k = 9, the cost factor is minimized to 0.20, indicating the efficiency and effectiveness of our approach. The proposed method not only enhances processing speed but also ensures minimal data loss, making it suitable for real-time anonymization of big data streams. This work provides a balanced solution that addresses the critical need for high-speed data anonymization while maintaining data privacy and utility. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Anonymity confidentiality Big data data disclosure Privacy

来源：评论

学校读者我要写书评

暂无评论

Cloud-magnetic resonance imaging system:In the era of 6G and artificial intelligence

引用

Magnetic Resonance Letters 2025年第1期5卷 52-63页

作者： Yirong Zhou Yanhuang Wu Yuhan Su Jing Li Jianyu Cai Yongfu You Jianjun Zhou Di Guo Xiaobo Qu Department of Electronic Science Intelligent Medical Imaging R&D CenterFujian Provincial Key Laboratory of Plasma and Magnetic ResonanceNational Institute for Data Science in Health and MedicineXiamen UniversityXiamen361104China Department of Electronic Science Key Laboratory of Digital Fujian on IoT CommunicationXiamen UniversityXiamen361104China Shanghai Electric Group Co. LtdShanghai200002China China Telecom Group Quanzhou362018China China Mobile Group Xiamen361009China Department of Radiology Zhongshan Hospital(Xiamen)Fudan UniversityXiamen Municipal Clinical Research Center for Medical ImagingFujian Province Key Clinical Specialty Construction Project(Medical Imaging Department)Xiamen Key Laboratory of Clinical Transformation of Imaging Big Data and Artificial IntelligenceXiamen361006China School of Computer and Information Engineering Xiamen University of TechnologyXiamen361024China

Magnetic resonance imaging(MRI)plays an important role in medical diagnosis,generating petabytes of image data annually in large *** voluminous data stream requires a significant amount of network bandwidth and extensive storage ***,local data processing demands substantial manpower and hardware *** isolation across different healthcare institutions hinders crossinstitutional collaboration in clinics and *** this work,we anticipate an innovative MRI system and its four generations that integrate emerging distributed cloud computing,6G bandwidth,edge computing,federated learning,and blockchain *** system is called Cloud-MRI,aiming at solving the problems of MRI data storage security,transmission speed,artificial intelligence(AI)algorithm maintenance,hardware upgrading,and collaborative *** workflow commences with the transformation of k-space raw data into the standardized Imaging Society for Magnetic Resonance in Medicine Raw data(ISMRMRD)***,the data are uploaded to the cloud or edge nodes for fast image reconstruction,neural network training,and automatic ***,the outcomes are seamlessly transmitted to clinics or research institutes for diagnosis and other *** Cloud-MRI system will save the raw imaging data,reduce the risk of data loss,facilitate inter-institutional medical collaboration,and finally improve diagnostic accuracy and work efficiency.

关键词： Magnetic resonance imaging Cloud computing 6G bandwidth Artificial intelligence Edge computing Federated learning Blockchain

来源：评论

学校读者我要写书评

暂无评论

Multiscale and Auto-Tuned Semi-Supervised Deep Subspace Clustering and Its Application in Brain Tumor Clustering

引用

computers, Materials & Continua 2024年第6期79卷 4741-4762页

作者： Zhenyu Qian Yizhang Jiang Zhou Hong Lijun Huang Fengda Li Khin Wee Lai Kaijian Xia School of Artificial Intelligence and Computer Science Jiangnan UniversityWuxi214122China Department of Medical Imaging The Changshu Affiliated Hospital of Soochow UniversitySuzhou215500China Department of Neurosurgery The Changshu Affiliated Hospital of Soochow UniversityChangshu215500China Department of Scientific Research The Changshu Affiliated Hospital of Soochow UniversitySuzhou215500China Changshu Key Laboratory of Medical Artificial Intelligence and Big Data Suzhou215500China Department of Biomedical Engineering Faculty of EngineeringUniversity of MalayaKuala Lumpur50603Malaysia

In this paper,we introduce a novel Multi-scale and Auto-tuned Semi-supervised Deep Subspace Clustering(MAS-DSC)algorithm,aimed at addressing the challenges of deep subspace clustering in high-dimensional real-world data,particularly in the field of medical *** deep subspace clustering algorithms,which are mostly unsupervised,are limited in their ability to effectively utilize the inherent prior knowledge in medical *** MAS-DSC algorithm incorporates a semi-supervised learning framework that uses a small amount of labeled data to guide the clustering process,thereby enhancing the discriminative power of the feature ***,the multi-scale feature extraction mechanism is designed to adapt to the complexity of medical imaging data,resulting in more accurate clustering *** address the difficulty of hyperparameter selection in deep subspace clustering,this paper employs a Bayesian optimization algorithm for adaptive tuning of hyperparameters related to subspace clustering,prior knowledge constraints,and model loss *** experiments on standard clustering datasets,including ORL,Coil20,and Coil100,validate the effectiveness of the MAS-DSC *** results show that with its multi-scale network structure and Bayesian hyperparameter optimization,MAS-DSC achieves excellent clustering results on these ***,tests on a brain tumor dataset demonstrate the robustness of the algorithm and its ability to leverage prior knowledge for efficient feature extraction and enhanced clustering performance within a semi-supervised learning framework.

关键词： Deep subspace clustering multiscale network structure automatic hyperparameter tuning semi-supervised medical image clustering

来源：评论

学校读者我要写书评

暂无评论

Mapping the Landscape of Personalization: A Comprehensive Review of Prediction and Trends in Recommendation Systems

引用

Wiley Interdisciplinary Reviews: data Mining and Knowledge Discovery 2025年第2期15卷

作者： Sachdeva, Tamanna Goyal, Lalit Mohan Mittal, Mamta Department of Computer Engineering J.C. Bose University of Science and Technology YMCA Faridabad India Department of Data Analytics Delhi Skill and Entrepreneurship University New Delhi India

Recommendation systems (RSs) have become indispensable features in nearly all web applications. Sifting through data and alleviating information overload, these systems offer more streamlined and personalized recommendations. E-commerce giants such as Amazon, Netflix, and YouTube are offering recommendations to users based on their interests, past experiences, demographic information, etc. hence, increasing the user's engagement on these applications. This study offers a comprehensive review of recommendation systems, covering their types, fundamental techniques, and emerging trends, with a focus on the predictive models and algorithms that power personalization. This study shows how in comparison to traditional collaborative and content-based recommendation systems-building techniques, the novel approaches of deep learning, graph-based techniques, meta-learning, few-shot learning, exploration, and federated learning offer promising prospects to improve recommendation systems' scalability, privacy-preserving abilities, and accuracy. These advanced methods deliver more diverse, context-aware, and personalized recommendations by leveraging large-scale data and complex predictive algorithms. Furthermore, this paper depicts forthcoming trajectories in the field of recommendation systems, including the adoption of graph-based approaches, federated learning, and the exploration of ethical considerations. By mapping the current landscape of prediction-driven personalization and identifying emerging trends, this review serves as a valuable resource for scholars and practitioners seeking to deepen their understanding of the field and drive innovation in recommendation systems. Readers can expect to gain insights into both foundational and cutting-edge techniques and how these can shape the future of personalized recommendations. © 2025 Wiley Periodicals LLC.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

A Survey of Research on Few-Shot Relation Classification

SSRN

引用

SSRN 2024年

作者： Zhang, YaChuan Guo, Yi Department of Computer Science and Engineering East China University of Science and Technology Shanghai200237 China Shanghai Engineering Research Center of Big Data Internet Audience Shanghai China Business Intelligence and Visualization Research Center National Engineering Laboratory for Big Data Distribution and Exchange Technologies Shanghai China

Relation classification is a crucial sub-task of information extraction, playing a significant role in extracting structured information from vast texts and constructing knowledge graphs. However, in practical scenarios, the availability of labeled data is often limited, and some datasets exhibit a long-tail distribution. To address these challenges, few-shot learning relation classification (FSRC) has emerged as a promising approach. The purpose of FSRC is to learn high-quality features with insufficient data for distinguishing the relation between given entities in a sentence. This task presents greater challenges compared to traditional relation classification tasks. Recently, FSRC has gained a lot of attention, prompting numerous researchers to provide innovative methods, datasets, and ideas to tackle this challenge. This paper focuses on relation classification based on few-shot learning and provides a comprehensive review of state-of-the-art techniques. Based on distinct research characteristics, we classified FSRC methods into five categories: metric-based method, optimization-based method, pre-trained and LLMs based method, prompt-based method, and lower source situation. We analyze the advantages and disadvantages of each category, provide a summary of the datasets, metrics, and results obtained by representative models. Finally, the paper concludes the current limitations of the few-shot relation classification and proposes future research directions. © 2024, The Authors. All rights reserved.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Face Mask Detection and Recognition System 1

Face Mask Detection and Recognition System

引用

1st IEEE International Conference on Blockchain and Distributed Systems Security, ICBDS 2022

作者： Gaiki, Sonali Deshmukh, Hrucha Gore, Sanjana Maddipati, Ramya Kamble, Shailesh Yeshwantrao Chavan College of Engineering Department of Computer Technology Nagpur India Indira Gandhi Delhi Technical University for Women Department of Artificial Intelligence and Data Science Delhi India

ISBN: (数字)9781665428323

ISBN: (纸本)9781665428323

As today a disease called COVID-19 is causing health crisis and deaths, it became most essential to wear a mask for protecting ourselves from Corona virus. Even in public areas, where is more rush we should wear mask as no virus can spread from person to person if any one of person from public is Corona positive. This paper introduces face mask detection that can be used by the authorities to make mitigation, evaluation, prevention, and action planning against COVID19. So basically in this project we are going to use Python, Keras, OpenCV alongwith MobileNet for this Face Mask Detection System. This includes some steps like data preprocessing, training and testing the model, run and view the accuracy and applying model in the camera. The inputs has provided here are 1000+ images of people with mask and without mask. First the data get processed and then by checking features of each image it will train all the models and the persons with and without mask get separated to two categories: with mask and without mask. If person is wearing mask with 90 or more percent of accuracy, then he will get added to with mask category and person not wearing mask get added to without mask category, so that we can permit with mask person to public areas. © 2022 IEEE.

关键词： Convolutional neural networks

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：