检索结果-内蒙古大学图书馆

MITIGATING REWARD OVER-OPTIMIZATION IN RLHF VIA BEHAVIOR-SUPPORTED REGULARIZATION

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Dai, Juntao Chen, Taiye Yang, Yaodong Zheng, Qian Pan, Gang College of Computer Science and Technology Zhejiang University China The State Key Lab of Brain-Machine Intelligence Zhejiang University China LLM Safety Centre Beijing Academy of Artificial Intelligence China Center for AI Safety and Governance Peking University China

Reinforcement learning from human feedback (RLHF) is an effective method for aligning large language models (LLMs) with human values. However, reward over-optimization remains an open challenge leading to discrepancies between the performance of LLMs under the reward model and the true human objectives. A primary contributor to reward over-optimization is the extrapolation error that arises when the reward model evaluates out-of-distribution (OOD) responses. However, current methods still fail to prevent the increasing frequency of OOD response generation during the reinforcement learning (RL) process and are not effective at handling extrapolation errors from OOD responses. In this work, we propose the Behavior-Supported Policy Optimization (BSPO) method to mitigate the reward over-optimization issue. Specifically, we define behavior policy as the next token distribution of the reward training dataset to model the in-distribution (ID) region of the reward model. Building on this, we introduce the behavior-supported Bellman operator to regularize the value function, penalizing all OOD values without impacting the ID ones. Consequently, BSPO reduces the generation of OOD responses during the RL process, thereby avoiding overestimation caused by the reward model’s extrapolation errors. Theoretically, we prove that BSPO guarantees a monotonic improvement of the supported policy until convergence to the optimal behavior-supported policy. Empirical results from extensive experiments show that BSPO outperforms baselines in preventing reward overoptimization due to OOD evaluation and finding the optimal ID policy. Copyright © 2025, The Authors. All rights reserved.

关键词： Extrapolation

Machine Learning in Fraud Detection for Financial Sustainability in Credit Card 2nd

学校读者我要写书评

暂无评论

Machine Learning in Fraud Detection for Financial Sustainabi...

2nd International Conference on Sustainable Development Using Machine Learning, artificial intelligence, and IoT, ICSD 2024

作者： Kaur, Harleen Chauhan, Ritu Shamsi, Yezdanul Haque Alankar, Bhavya Artificial Intelligence and IoT Lab Center for Computational Biology and Bioinformatics Amity University Noida India Department of Computer Science and Engineering School of Engineering Sciences and Technology Jamia Hamdard New Delhi India

ISBN: (纸本)9783031717284

In today’s financial world, various technologies have been emerged and with the evolution in the technologies it has led to the increase in the fraud cases. The sustainable growth of businesses and the financial markets are severely harmed by financial statement fraud. Consequently, it is crucial and vital to build an efficient methodology to identify financial statement fraud in businesses. Usage of credit card are very frequent these days and due to loopholes in the system fraud cases in credit card is rising which causes huge financial losses. In this paper the focus is on credit card fraud detection in real world with the help of machine learning. In this case Logistic Regression, Decision tree method and SVM (support vector machine) are applied on credit card fraud detection for knowledge discovery among the data. The applied methods can detect the fraudulent transactions with highest accuracy. However, the results are predicted using evaluation metrics such as accuracy, error rate, specificity, sensitivity, and ROC curves are calculated to assess each model's performance. Hence, the predicted results suggested that logistic regression is the most accurate model for prediction of fraud datasets in financial sustainability. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Decentralized finance

Explainable Machine Learning for Radio Environment Mapping: An Intelligent System for Electric Field Strength Monitoring.

学校读者我要写书评

暂无评论

IEEE Access 2025年 13卷 75104-75122页

作者： Kiouvrekis, Yiannis Panagiotakopoulos, Theodor Nousi, Efthymia Filippopoulos, Ioannis Ploussi, Agapi Spyratou, Ellas Efstathopoulos, Efstathios P. University of Thessaly Mathematics Computer Science and Artificial Intelligence Lab Faculty of Public and One Health Karditsa43100 Greece University of Limassol Department of Information Technologies Limassol Cyprus University of Nicosia Business School Cyprus University of Patras Department of Management Science and Technology Patras26334 Greece Infralabs Ltd Agiou Pavlou 61 Nicosia Cyprus University of Limassol Limassol Cyprus National and Kapodistrian University of Athens Medical School Department of Applied Medical Physics Athens11527 Greece

The accurate characterization of signal propagation is critical for optimizing wireless network performance and supporting applications such as electromagnetic field (EMF) exposure assessment and the development of Radio Environmental Maps (REMs). This study proposes a novel, explainable machine learning system to predict electric field strength across diverse urban, semi-urban, and rural environments in Cyprus. The system is trained on a rich dataset comprising 6,543 EMF measurements collected in 2023 at mobile phone and digital TV stations, following CEPT/ECC/REC/(02)04 recommendations. The dataset includes geospatial and environmental features such as antenna distance, population density, urbanization level, and detailed built environment characteristics (e.g., volume, surface, and height). We evaluate multiple machine learning models—kNN, neural networks, decision trees, random forests, XGBoost, and LightGBM—using a two-semester split for training and assessment. Best performance was achieved with the Random Forest model, which yielded the lowest RMSE among all models. Gradient boosting models (XGBoost and LightGBM) also performed well, with RMSE values slightly higher than RF while offering flexible and scalable configurations. In contrast, k-NN and neural networks showed higher RMSE values, indicating they were less effective for this specific task. Across all models, confidence intervals were narrow, demonstrating stable and reliable predictions. Explainable AI techniques revealed that antenna distance, building volume, and population density are the most influential predictors of EMF intensity. Our approach outperforms traditional signal models by incorporating urban morphology and demographic context. As part of this system, we also create a Geographic Information System (GIS) that displays electromagnetic field strength maps derived from our explainable machine learning models. This contributes a scalable, interpretable framework for EMF exposure mapping t

关键词： Machine learning Predictive models Radio frequency Accuracy Electric fields Random forests Population density Explainable AI Estimation Buildings

artificial intelligence Detects Antidepressant Use From Nocturnal Breathing

学校读者我要写书评

暂无评论

Biological Psychiatry 2025年第9期97卷 S22-S22页

作者： Dina Katabi MIT Computer Science & Artificial Intelligence Lab

Neuronal Activation States as Sample Embeddings for Data Selection in Task-Specific Instruction Tuning

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Ma, Da Shang, Gonghu Chen, Zhi Qin, Libo Luo, Yijie Pan, Lei Fan, Shuai Chen, Lu Yu, Kai X-LANCE Lab Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence SJTU AI Institute Shanghai Jiao Tong University Shanghai China AISpeech Co. Ltd. Suzhou China ByteDance China School of Computer Science and Engineering Central South University China

Task-specific instruction tuning enhances the performance of large language models (LLMs) on specialized tasks, yet efficiently selecting relevant data for this purpose remains a challenge. Inspired by neural coactivation in the human brain, we propose a novel data selection method called NAS, which leverages neuronal activation states as embeddings for samples in the feature space. Extensive experiments show that NAS outperforms classical data selection methods in terms of both effectiveness and robustness across different models, datasets, and selection ratios. © 2025, CC BY.

关键词： Embeddings

Non-Equilibrium MAV-Capture-MAV via Time-Optimal Planning and Reinforcement Learning

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Zheng, Canlun Guo, Zhanyu Yin, Zikang Wang, Chunyu Wang, Zhikun Zhao, Shiyu College of Computer Science and Technology Zhejiang University Hangzhou China WINDY Lab Department of Artificial Intelligence Westlake University Hangzhou China Department of Electrical Engineering California Institute of Technology Pasadena United States

The capture of flying MAVs (micro aerial vehicles) has garnered increasing research attention due to its intriguing challenges and promising applications. Despite recent advancements, a key limitation of existing work is that capture strategies are often relatively simple and constrained by platform performance. This paper addresses control strategies capable of capturing high-maneuverability targets. The unique challenge of achieving target capture under unstable conditions distinguishes this task from traditional pursuit-evasion and guidance problems. In this study, we transition from larger MAV platforms to a specially designed, compact capture MAV equipped with a custom launching device while maintaining high maneuverability. We explore both time-optimal planning (TOP) and reinforcement learning (RL) methods. Simulations demonstrate that TOP offers highly maneuverable and shorter trajectories, while RL excels in real-time adaptability and stability. Moreover, the RL method has been tested in real-world scenarios, successfully achieving target capture even in unstable states. Copyright © 2025, The Authors. All rights reserved.

关键词： Micro air vehicle (MAV)

Deep Learning and Shape-Driven Combined Approach for Breast Cancer Tumor Segmentation

学校读者我要写书评

暂无评论

Deep Learning and Shape-Driven Combined Approach for Breast ...

International Conference on Advancements in Computational sciences (ICACS)

作者： Mudassar Ali Haoji Hu Tufail Muhammad Muhammad Ahsan Qureshi Tariq Mahmood College of Information Science and Electronic Engineering Zhejiang University Hangzhou China Department of Computer Science Air University Islamabad Pakistan Artificial Intelligence and Data Analytics Lab CCIS Prince Sultan University Riyadh Kingdom of Saudi Arabia.

ISBN: (数字)9798331533038

ISBN: (纸本)9798331533045

Our aim is to enhance the performance of segmenting breast cancer from medical images by overcoming major challenges, including data insufficiency and complexity when training a model. Using the INbreast dataset, it proposes a hybrid method for improving the segmentation accuracy and computational efficiency by combining M3D-Neural Cellular Automata with Shape-Guided Segmentation. This work aims at optimal performance on complex tasks of tumor segmentation by leveraging shape priors and reducing the need for extensive training data. Results confirm that significant improvements in both accuracy and robustness are achieved, which reduces computation time and improves tumor boundary delineation. The hybrid approach provides a time- and cost-effective method for precise detection in breast cancer diagnosis and treatment.

关键词： Deep learning Training Image segmentation Accuracy Shape Training data Breast cancer Data models Robustness Real-time systems

Applicability of the Minimal Dominating Set for Influence Maximisation in Multilayer Networks

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Czuba, Michal Jia, Mingshan Bródka, Piotr Musial, Katarzyna Department of Artificial Intelligence Wroclaw University of Science and Technology 27 wybrzeze Wyspiańskiego st Wroclaw50-370 Poland Complex Adaptive Systems Lab Data Science Institute School of Computer Science University of Technology Sydney UltimoNSW2007 Australia

The minimal dominating set (MDS) is a well-established concept in network controllability and has been successfully applied in various domains, including sensor placement, network resilience, and epidemic containment. In this study, we adapt the local-improvement MDS routine and explore its potential for enhancing seed selection for influence maximisation in multilayer networks (MLN). We employ the Linear Threshold Model (LTM), which offers an intuitive representation of influence spread or opinion dynamics by accounting for peer influence accumulation. To ensure interpretability, we utilise rank-refining seed selection methods, with the results further filtered with MDS. Our findings reveal that incorporating MDS into the seed selection process improves spread only within a specific range of situations. Notably, the improvement is observed for larger seed set budgets, lower activation thresholds, and when an"AND" strategy is used to aggregate influence across network layers. This scenario reflects situations where an individual does not require the majority of their acquaintances to hold a target opinion, but must be influenced across all social circles. Copyright © 2025, The Authors. All rights reserved.

关键词： Multilayers

Benchmarking Graph Representations and Graph Neural Networks for Multivariate Time Series Classification

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Yang, Wennuo Wu, Shiling Zhou, Yuzhi Luo, Cheng He, Xilin Xie, Weicheng Shen, Linlin Song, Siyang Computer Vision Institute School of Computer Science & Software Engineering Shenzhen University China Shenzhen Institute of Artificial Intelligence and Robotics for Society China Guangdong Provincial Key Laboratory of Intelligent Information Processing China HBUG Lab University of Exeter United Kingdom

Multivariate Time Series Classification (MTSC) enables the analysis if complex temporal data, and thus serves as a cornerstone in various real-world applications, ranging from healthcare to finance. Since the relationship among variables in MTS usually contain crucial cues, a large number of graph-based MTSC approaches have been proposed, as the graph topology and edges can explicitly represent relationships among variables (channels), where not only various MTS graph representation learning strategies but also different Graph Neural Networks (GNNs) have been explored. Despite such progresses, there is no comprehensive study that fairly benchmarks and investigates the performances of existing widely-used graph representation learning strategies/GNN classifiers in the application of different MTSC tasks. In this paper, we present the first benchmark which systematically investigates the effectiveness of the widely-used three node feature definition strategies, four edge feature learning strategies and five GNN architecture, resulting in 60 different variants for graph-based MTSC. These variants are developed and evaluated with a standardized data pipeline and training/validation/testing strategy on 26 widely-used suspensor MTSC datasets. Our experiments highlight that node features significantly influence MTSC performance, while the visualization of edge features illustrates why adaptive edge learning outperforms other edge feature learning methods. The code of the proposed benchmark is publicly available at https://***/CVI-yangwn/*** Codes 68T10 © 2025, CC BY.

关键词： Graph neural networks