检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Klein, Lukas Lüth, Carsten Schlegel, Udo Bungert, Till El-Assady, Mennatallah Jäger, Paul Interactive Machine Learning Group Germany ETH Zürich Department of Computer Science Switzerland Germany Heidelberg University Department of Computer Science Germany University of Konstanz Department of Computer Science Germany

Explainable AI (XAI) is a rapidly growing domain with a myriad of proposed methods as well as metrics aiming to evaluate their efficacy. However, current studies are often of limited scope, examining only a handful of XAI methods and ignoring underlying design parameters for performance, such as the model architecture or the nature of input data. Moreover, they often rely on one or a few metrics and neglect thorough validation, increasing the risk of selection bias and ignoring discrepancies among metrics. These shortcomings leave practitioners confused about which method to choose for their problem. In response, we introduce LATEC, a large-scale benchmark that critically evaluates 17 prominent XAI methods using 20 distinct metrics. We systematically incorporate vital design parameters like varied architectures and diverse input modalities, resulting in 7,560 examined combinations. Through LATEC, we showcase the high risk of conflicting metrics leading to unreliable rankings and consequently propose a more robust evaluation scheme. Further, we comprehensively evaluate various XAI methods to assist practitioners in selecting appropriate methods aligning with their needs. Curiously, the emerging top-performing method, Expected Gradients, is not examined in any relevant related study. LATEC reinforces its role in future XAI research by publicly releasing all 326k saliency maps and 378k metric scores as a (meta-)evaluation dataset. The benchmark is hosted at: https://***/IML-DKFZ/latec. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Computationally Efficient machine learning Methodology for Indian Nobel Laureate Classification 6

Computationally Efficient Machine Learning Methodology for I...

引用

6th IEEE International Conference on Recent Advances in Intelligent Computational Systems, RAICS 2024

作者： Banerjee, Swastik Joy, Helen. K. Artificial Intelligence & Machine Learning Christ Deemed to be University Department of Computer Science Bengaluru India Christ Deemed to be University Department of Computer Science Bengaluru India

ISBN: (纸本)9798350381689

A computationally efficient methodology for Indian Nobel Laureate classification is proposed in this study, emphasizing the optimization of image categorization through supervised learning techniques. Leveraging advancements in Convolutional Neural Networks (CNNs), the research aims to enhance the efficiency and precision of image classification tasks. The study utilizes Logistic Regression for dataset analysis, initially employing browser extensions for mass downloading categorized image data. Haar cascade classifiers are then used for data wrangling, focusing on facial, nose, and mouth recognition. Following this, feature engineering through wavelet transformation reduces image dimensionality, preparing the dataset for the chosen ML model, Logistic Regression. The primary focus is to simplify technology for improved image categorization. Support Vector machines (SVM), Random Forest, and Logistic Regression are examined, with Logistic Regression emerging as the most effective model, achieving an accuracy rate of 87.5%. A thorough evaluation using Confusion Matrices reveals Logistic Regression's superior performance in classifying images of Indian Nobel laureates. A strategic up-sampling approach is implemented to address dataset inconsistencies, ensuring balanced representation across classes. The Haar wavelet transform is then applied for feature extraction, optimizing the dataset for ML models. The dataset is split into training and testing sets (80-20), and the three models are trained and evaluated for accuracy. Logistic Regression proves to be the best performer, offering insights into prominent leaders' identification. The research offers a detailed pipeline for data preprocessing, feature engineering, and model assessment, culminating in a robust image categorization system. Logistic Regression emerges as a reliable method for biographical picture identification, demonstrating superior accuracy over SVM and Random Forest. This research underscores the import

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Leveraging machine learning for intelligent agriculture

引用

Discover Internet of Things 2025年第1期5卷 1-21页

作者： Sowmya, B.J. Meeradevi, A.K. Supreeth, S. Pradeep Kumar, D. Ravi Kumar, B.N. Rohith, S. Mishra, Divyansh Koushik, Abhishek Patil, Ankit U. Department of Artificial Intelligence and Data Science Ramaiah Institute of Technology Bengaluru560054 India Department of Artificial Intelligence and Machine Learning M S Ramaiah Institute of Technology Bengaluru560054 India School of Computer Science and Engineering REVA University Bengaluru560064 India Department of Computer Science Ramaiah Institute of Technology Bengaluru560054 India Department of Information Science and Engineering BMS Institute of Technology & Management Bengaluru560119 India Department of Electronics and Engineering Nagarjuna College of Engineering and Technology Bengaluru562164 India

Agriculture, the backbone of many economies, faces challenges like lack of information, outdated practices, and limited access to technology, hindering farmer productivity. This work proposes a user-friendly, multilingual platform leveraging Generative AI to address farmers' diverse needs. The platform encompasses various features to enhance agricultural practices. An LLM-powered Government Scheme Advisor functions as a multilingual chatbot offering intelligent guidance on government agricultural schemes and subsidies. The Disease Detection module utilizes AI technology for real-time identification and treatment recommendations, minimizing crop diseases and yield losses. The Soil Testing Centre feature locates nearby soil testing centers, providing essential information based on geographical data to assist farmers in optimizing soil quality. A Crop Recommendation feature employs machine learning algorithms to offer personalized crop recommendations, considering various factors and aiding informed decision-making. The Crop Planning Tool, with its intuitive user interface, simplifies planning planting schedules and managing resources. Additionally, the platform includes an MSP Center Locator to find nearby Minimum Support Price (MSP) centers based on location. By integrating these innovative solutions, this platform bridges the gap between conventional agricultural techniques and contemporary technology, equipping farmers with the resources and expertise essential for advancing productivity and sustainability. Multilingual support ensures accessibility for a wider audience, breaking down language barriers and promoting inclusivity in the agricultural sector. This work proposes an innovative, multilingual platform powered by Generative AI to address these issues. Key features include an LLM-driven chatbot for government scheme guidance, AI-based real-time disease detection, and location-based tools for soil testing and MSP center identification. Additionally, the platf

关键词： Smart agriculture

来源：评论

学校读者我要写书评

暂无评论

Understanding and Minimising Outlier Features in Transformer Training 38

Understanding and Minimising Outlier Features in Transformer...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： He, Bobby Noci, Lorenzo Paliotta, Daniele Schlag, Imanol Hofmann, Thomas Department of Computer Science ETH Zürich Switzerland Machine Learning Group University of Geneva Switzerland

Outlier Features (OFs) are neurons whose activation magnitudes significantly exceed the average over a neural network's (NN) width. They are well known to emerge during standard transformer training and have the undesirable effect of hindering quantisation in afflicted models. Despite their practical importance, little is known behind why OFs emerge during training, nor how one can minimise them. Our work focuses on the above questions, first identifying several quantitative metrics, such as the kurtosis over neuron activation norms, to measure OFs. With these metrics, we study how architectural and optimisation choices influence OFs, and provide practical insights to minimise OFs during training. As highlights, we introduce a novel unnormalised transformer block, the Outlier Protected block, and present a previously unknown benefit of non-diagonal preconditioning optimisers, finding both approaches to significantly reduce OFs and improve quantisation without compromising convergence speed, at scales of up to 7B parameters. Notably, our combination of OP block and non-diagonal preconditioner (SOAP) achieves 14.87 weight-and-activation int8 perplexity (from 14.71 in standard precision), compared to 63.4 int8 perplexity (from 16.00) with a default OF-prone combination of Pre-Norm model and Adam, when quantising OPT-125m models post-training. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

SecureFedPROM: A Zero-Trust Federated learning Approach with Multi-Criteria Client Selection

引用

IEEE Journal on Selected Areas in Communications 2025年第6期43卷 2025-2041页

作者： Tahir, Mehreen Mawla, Tanjila Awaysheh, Feras Alawadi, Sadi Gupta, Maanak Ali, Muhammad Intizar Dublin City University SFI Centre for Research Training in Machine Learning Dublin Ireland Tennessee Tech University Department of Computer Science CookevilleTN United States Umeå University Department of Computer Science Umeå Sweden Blekinge Tekniska Högskola School of Computer Science Karlskrona Sweden Dublin City University School of Electronic and Computer Engineering Dublin Ireland

Federated learning (FL) enables decentralized learning while preserving data privacy. However, ensuring security and optimizing resource utilization in FL remains challenging, particularly in untrusted environments. To address this, we propose SecureFedPROM, a novel zero-trust FL framework that integrates Attribute-Based Access Control (ABAC) for secure client authorization and Preference Ranking Organization Method for Enrichment of Evaluations (PROMETHEE) for dynamic, multi-criteria client selection. Unlike traditional FL client selection methods that prioritize security or efficiency, SecureFedPROM optimizes trustworthiness, computational efficiency, and performance, ensuring robust participation in each training round. We evaluate SecureFedPROM across multiple real-world datasets, demonstrating its superiority over state-of-the-art client selection protocols. Our results show that SecureFedPROM achieves a 7.19% improvement in model accuracy, accelerates convergence, and reduces the number of training rounds. Additionally, it minimizes wall-clock time and computational overhead, making it highly scalable for edge AI environments. These findings highlight the importance of integrating zero-trust security principles with multi-criteria decision-making to enhance security and efficiency in FL. © 1983-2012 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

De-Styling Bias in #Climatical Tweets: A Neural Language Style Transfer Approach 3

De-Styling Bias in #Climatical Tweets: A Neural Language Sty...

引用

3rd International Conference on Intelligent and Innovative Technologies in Computing, Electrical and Electronics, IITCEE 2025

作者： George, Elizabeth Leah Parthasarathy, Subashini Avinashilingam Institute for Home Science and Higher Education for Women Department of Computer Science Coimbatore India Centre for Machine Learning and Intelligence Avinashilingam Institute for Home Science and Higher Education for Women Department of Computer Science Coimbatore India

ISBN: (纸本)9798331515911

The spread of biased and misleading opinions on social media regarding climate change necessitates robust solutions to counteract misinformation and promote balanced discourse. In this study, we introduce the Semantic Neural Style Transfer Bias (SNLT) Framework, an AI-driven solution designed to identify and mitigate bias in climate-related tweets. Leveraging a curated dataset of climate change tweets collected from Twitter, we employ the all-MiniLM-L6-v2 sentence transformer for feature extraction and BERT-based embeddings to capture semantic nuances. Our methodology integrates semantic clustering models to evaluate clustering performance using silhouette scores, including KMeans, Gaussian KMeans, keyNMF clustering, and an Autoencoder-based model. DistilBERT is further employed to classify tweets as biased or unbiased, while a Named Entity Recognition (NER) model trained on RobERTa transformers identifies word-level bias indicators. The fundamental basis of our framework is a transformer-based style transfer model, such as GPT-2, which adjusts the stylistic tone of biased text while preserving its semantic meaning. This approach employs a style-specific decoder to ensure grammatical accuracy and contextual relevance, producing text with a neutral tone through refined phrasing and word choice. Evaluation results demonstrate a substantial reduction in bias without compromising fluency or naturalness, validated through statistical metrics and human assessments. Additionally, Explainable AI techniques are integrated to enhance the transparency of model decisions, offering insights into bias mitigation processes. Our findings highlight the framework's efficacy in addressing climate misinformation and fostering constructive conversations on social media, underscoring its potential as a responsible AI tool for digital discourse. © 2025 IEEE.

关键词： Tweets

来源：评论

学校读者我要写书评

暂无评论

Enhancement of Infant Health Assessment: Predicting Body Mass Index (BMI) from Real-Time Facial Images Using machine learning Techniques 23th

Enhancement of Infant Health Assessment: Predicting Body Mas...

引用

23rd International Conference on Intelligent Systems Design and Applications, ISDA 2023

作者： Krishnaveni, M. Subashini, P. Janani, R. Jeeva, N. Department of Computer Science Centre for Machine Learning and Intelligence Avinashilingam Institute for Home Science and Higher Education for Women Coimbatore India Centre for Machine Learning and Intelligence Avinashilingam Institute for Home Science and Higher Education for Women Coimbatore India Department of Computer Science Avinashilingam Institute for Home Science and Higher Education for Women Coimbatore India

ISBN: (纸本)9783031648120

Body Mass Index in infants is a valuable indicator for assessing their growth and nutritional state, helping to identify any potential issues in the early stages. It establishes whether infants are underweight, within a healthy weight range, or overweight, which can specify their overall health. The capability to accurately determine an infant’s BMI can provide significant insight into their general health and helps parents to adopt smart feeding and healthcare decisions. In today’s world, where parents often have limited time and wealth, supporting the proper growth of their infant is vital. The fundamental objective of this research is to propose a user-friendly mobile application which is constructed for predicting infant BMI. This proposed application provides a simple and easy way for parents to track and understand their infant’s nutritionary condition. The proposed application requires only a facial image of an infant, eliminating the necessity for addition parameters such as height, weight. For the prediction of infant BMI, regression algorithms, which includes Linear, Polynomial, Random Forest, and Kernal Ridge, have been employed. The performance of these models has been assessed using several evaluation metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R2) score, and the results clearly show their effectiveness in this significant task. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： machine learning

来源：评论

学校读者我要写书评

暂无评论

Hybrid Design for Privacy Preserved Image Representation in a Cloud Environment 12th

Hybrid Design for Privacy Preserved Image Representation in ...

引用

12th International Conference on Recent Trends in Computing, ICRTC 2024

作者： Vijay, K. Sorna Shanthi, D. Jaeyalakshmi, M. Vignesh, P. Yuvaraja, M. Department of Artificial Intelligence and Machine Learning Rajalakshmi Engineering College Chennai India Department of Artificial Intelligence and Data Science Rajalakshmi Engineering College Chennai India Department of Computer Science and Engineering Rajalakshmi Engineering College Chennai India

ISBN: (纸本)9789819789450

The digital era has made seamless sharing and keeping of media such as images on cloud platforms an integral part of our lives. Still, there is a big issue about user privacy and data security in these repositories. We have thus created a very strong system that is able to do image encryption and decryption so as to protect the privacy of users while at the same time making their pictures accessible. Our system uses hybrid encryption where we try to prevent even cloud service providers from accessing the original files hence ensuring that only the owner who has been authenticated through password based authentication can decrypt or access them. In this way, it allows maintaining high privacy standards without sacrificing user convenience. It starts with preparing photos for encryption so that they are kept safe during transmission and storage. When needed by a user, images are decrypted by our system and become available only to authorized users. Consequently, users can comfortably leave their private photo albums on cloud storage systems knowing too well that their friends cannot access them thereby increasing trust in using cloud services for storing sensitive information. The aim of this work is to increase user privacy and data security in cloud repositories through employing this encryption technique. The users are guaranteed that their private details are not exposed to others enabling them to use cloud storage efficiently without interfering with their privacy at all. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Cloud platforms

来源：评论

学校读者我要写书评

暂无评论

Impact of Immersive Technology on Physiological Parameters in Psychiatric Disorders 13

Impact of Immersive Technology on Physiological Parameters i...

引用

13th International Conference on System Modeling and Advancement in Research Trends, SMART 2024

作者： Kataria, Anjali Singh, Gurjinder Sandhu, Jasminder Chitkara Institute of Engineering and Technology Punjab India Iilm University Department of Machine Learning and Data Science Noida India

ISBN: (纸本)9798350380583

The brain is an essential component that regulates the general functioning of the body. The brain consists of millions of neurons that govern human behavior in response to sensory stimuli. To understand the mental functioning of the brain, it is important to assess its neural signals. Presently, the exponential expansion of prescription medications and substances, including alcohol and drugs, is exerting a significant influence on human psychology. The rising prevalence of different health disorders can be attributed to the increasing number of cases. The sensorimotor cortex, which in humans is in charge of coordinating both physical and cognitive activities, can be used to visualize brain signals. The present research describes an empirical investigation that employs Virtual Reality therapy to examine the affective condition of individuals suffering from drug addiction. The study utilized the Emotiv Insight 5-channel EEG headset to collect EEG data from 50 participants, and the Emotiv Pro software was employed for data analysis. The study findings indicate that Virtual Reality treatment has a substantial positive impact on the stress levels and concentration of persons suffering from drug addiction. © 2024 IEEE.

关键词： Electroencephalography

来源：评论

学校读者我要写书评

暂无评论

Functional linear non-Gaussian acyclic model for causal discovery

引用

Behaviormetrika 2024年第2期51卷 567-588页

作者： Yang, Tian-Le Lee, Kuang-Yao Zhang, Kun Suzuki, Joe Graduate School of Engineering Science Osaka University Osaka Japan Department of Statistics Operations and Data Science Temple University Philadelphia United States Machine Learning Department Carnegie Mellon University Pittsburgh United States Machine Learning Department MBZUAI Abu Dhabi United Arab Emirates

In causal discovery, non-Gaussianity has been used to characterize the complete configuration of a linear non-Gaussian acyclic model (LiNGAM), encompassing both the causal ordering of variables and their respective connection strengths. However, LiNGAM can only deal with the finite-dimensional case. To expand this concept, we extend the notion of variables to encompass vectors and even functions, leading to the functional linear non-Gaussian acyclic model (Func-LiNGAM). Our motivation stems from the desire to identify causal relationships in brain-effective connectivity tasks involving, for example, fMRI and EEG datasets. We demonstrate why the original LiNGAM fails to handle these inherently infinite-dimensional datasets and explain the availability of functional data analysis from both empirical and theoretical perspectives. We establish theoretical guarantees of the identifiability of the causal relationship among non-Gaussian random vectors and even random functions in infinite-dimensional Hilbert spaces. To address the issue of sparsity in discrete time points within intrinsic infinite-dimensional functional data, we propose optimizing the coordinates of the vectors using functional principal component analysis. Experimental results on synthetic data verify the ability of the proposed framework to identify causal relationships among multivariate functions using the observed samples. For real data, we focus on analyzing the brain connectivity patterns derived from fMRI data. © The Behaviormetric Society 2024.

关键词： Causal discovery Darmois–Skitovich theorem Functional data Gaussian process LiNGAM Non-Gaussian

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：