检索结果-内蒙古大学图书馆

SSRN 2022年

作者： Lee, Jae Eun Kim, Young-Bong Kim, Jongnam Division of Computer Engineering and AI Pukyong National University Busan48513 Korea Republic of

3D film images appear three-dimensional based on the angle, and viewing position of incident light rays. However, if the pixel contrast of the image is low or the patterns are cloudy, images do not appear three-dimensional, and it is difficult to perform quality inspection because detection is not straightforward. To address this issue, we propose a method to calculate the width of pixels for a specific height from the histogram of a 3D film image and classify it based on a threshold. The proposed algorithm is based on the feature that the widths of pixels by height in the histogram of a realistic 3D film image is wider than the histogram of an unrealistic 3D film image. In the experiment, it was confirmed that the position of the height section of the histogram has the highest classification accuracy. Through comparison tests with conventional algorithms, our approach demonstrated excellent classification accuracy for 3D film image classification and was proven to have high accuracy even for images with low contrast and indiscernible patterns. © 2022, The Authors. All rights reserved.

关键词： Pixels

来源：评论

学校读者我要写书评

暂无评论

Towards Reasonable Budget Allocation in Untargeted Graph Structure Attacks via Gradient Debias

arXiv

引用

arXiv 2023年

作者： Liu, Zihan Luo, Yun Wu, Lirong Liu, Zicheng Li, Stan Z. AI Division School of Engineering Westlake University Hangzhou310030 China College of Computer Science and Technology Zhejiang University Hangzhou310058 China

It has become cognitive inertia to employ cross-entropy loss function in classification related tasks. In the untargeted attacks on graph structure, the gradients derived from the attack objective are the attacker's basis for evaluating a perturbation scheme. Previous methods use negative cross-entropy loss as the attack objective in attacking node-level classification models. However, the suitability of the cross-entropy function for constructing the untargeted attack objective has yet been discussed in previous works. This paper argues about the previous unreasonable attack objective from the perspective of budget allocation. We demonstrate theoretically and empirically that negative cross-entropy tends to produce more significant gradients from nodes with lower confidence in the labeled classes, even if the predicted classes of these nodes have been misled. To free up these inefficient attack budgets, we propose a simple attack model for untargeted attacks on graph structure based on a novel attack objective which generates unweighted gradients on graph structures that are not affected by the node confidence. By conducting experiments in gray-box poisoning attack scenarios, we demonstrate that a reasonable budget allocation can significantly improve the effectiveness of gradient-based edge perturbations without any extra hyper-parameter. Copyright © 2023, The Authors. All rights reserved.

关键词： Entropy

来源：评论

学校读者我要写书评

暂无评论

SmartFit: ai-Powered Training Recommendations with Biometric Integration 5

SmartFit: AI-Powered Training Recommendations with Biometric...

引用

5th IEEE Global Conference for Advancement in Technology, GCAT 2024

作者： Parwej, Firoj Suganya, A. Naresh, Burri Goud, Veer Sudheer Kaliappan, S. Manikandan, G. Mandsaur University Department of Computer Science and Engineering Daulatpura Madhya Pradesh Mandsaur India Karpagam College of Engineering Department of Information Technology Tamilnadu Coimbatore India Cvr College of Engineering Department of Cse -CS Telangana Hyderabad India St. Martin's Engineering College Deptartment of Ai&ds Secunderabad India Lovely Professional University Division of Research and Development Delhi G.T.Road Punjab Phagwara India Saveetha University Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences Department of Electronics and Communication Engineering Chennai India

ISBN: (纸本)9798350376685

This research introduces SmartFit, a state-of-the-art technology that seamlessly integrates biometric data and ai-powered training recommendations. SmartFit monitors the user's pulse rate, oxygen saturation, and muscle activity in real time. Subsequently, it implements machine learning algorithms to develop customized training regimens. This ongoing adjustment to the user's evolving fitness levels and objectives guarantees that the workout regimens are both highly effective and customized to their unique requirements. The proposed system is designed to optimize exercise efficacy, promote extended engagement in fitness activities, and improve user satisfaction. The novel method significantly enhanced the fitness ratings and overall satisfaction of participants in comparison to more conventional forms of exercise, as evidenced by a three-month pilot study. The results demonstrate the potential for SmartFit's data-driven, individualized training recommendations to transform the fitness industry. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Learning How to Propagate Messages in Graph Neural Networks

arXiv

引用

arXiv 2023年

作者： Xiao, Teng Chen, Zhengyu Wang, Donglin Wang, Suhang AI Division School of Engineering Westlake University China College of Computer Science & Technology Zhejiang University China The Pennsylvania State University United States

This paper studies the problem of learning message propagation strategies for graph neural networks (GNNs). One of the challenges for graph neural networks is that of defining the propagation strategy. For instance, the choices of propagation steps are often specialized to a single graph and are not personalized to different nodes. To compensate for this, in this paper, we present learning to propagate, a general learning framework that not only learns the GNN parameters for prediction but more importantly, can explicitly learn the interpretable and personalized propagate strategies for different nodes and various types of graphs. We introduce the optimal propagation steps as latent variables to help find the maximum-likelihood estimation of the GNN parameters in a variational Expectation-Maximization (VEM) framework. Extensive experiments on various types of graph benchmarks demonstrate that our proposed framework can significantly achieve better performance compared with the state-of-the-art methods, and can effectively learn personalized and interpretable propagate strategies of messages in GNNs. Code is available at https://***/tengxiao1/L2P. Copyright © 2023, The Authors. All rights reserved.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Secure Federated Learning with Blockchain and Homomorphic Encryption for Healthcare Data Sharing

Secure Federated Learning with Blockchain and Homomorphic En...

引用

International Conference on Cyberworlds

作者： Muhammad Firdaus Kyung-Hyune Rhee Industrial Science Technology Research Center Pukyong National University Republic of Korea School of Computing Telkom University Indonesia Division of Computer Engineering and AI Pukyong National University Republic of Korea

ISBN: (数字)9798331527174

ISBN: (纸本)9798331527181

Healthcare data is often fragmented across various institutions due to its highly sensitive and private nature. In this sense, hospitals and clinics maintain electronic health records (EHRs) independently; hence, valuable data is siloed within individual organizations, preventing comprehensive analysis that could benefit from diverse data sources. Federated learning (FL) addresses these challenges by enabling the training of a shared global model using data distributed across multiple institutions without moving the data from its source. By leveraging FL, healthcare institutions can combine their data assets to improve predictive analytics, personalized medicine, and overall healthcare outcomes, ultimately benefiting patients and the healthcare system. However, the current FL model with a central server presents several challenges within healthcare, including the risk of malicious attacks, regulatory compliance, and privacy vulnerabilities. To overcome these issues, this paper introduces the FL framework with blockchain and homomorphic encryption (HE). Our framework aims to minimize the role of the central server, enable collaborative model training across healthcare organizations, and enhance data security and privacy. In this sense, blockchain ensures the integrity and transparency of the process, while homomorphic encryption ensures that the data remains private. This framework can potentially enable institutions to enrich medical knowledge while securely keeping patient data collaboratively and facilitating healthcare analytics in practical settings.

关键词： Training Data privacy Privacy Federated learning Collaboration Medical services Data models Blockchains Servers Homomorphic encryption

来源：评论

学校读者我要写书评

暂无评论

Electron charge qubit with 0.1 millisecond coherence time

引用

Nature Physics 2024年第1期20卷 116-122页

作者： Zhou, Xianjing Li, Xinhao Chen, Qianfan Koolstra, Gerwin Yang, Ge Dizdar, Brennan Huang, Yizhong Wang, Christopher S. Han, Xu Zhang, Xufeng Schuster, David I. Jin, Dafei Center for Nanoscale Materials Argonne National Laboratory LemontIL United States Pritzker School of Molecular Engineering University of Chicago ChicagoIL United States Computational Research Division Lawrence Berkeley National Laboratory BerkeleyCA United States The NSF AI Institute for Artificial Intelligence and Fundamental Interactions CambridgeMA United States Computer Science and Artificial Intelligence Laboratory Massachusetts Institute of Technology CambridgeMA United States James Franck Institute and Department of Physics University of Chicago ChicagoIL United States Department of Electrical and Computer Engineering Northeastern University BostonMA United States Department of Applied Physics Stanford University StanfordCA United States Department of Physics and Astronomy University of Notre Dame Notre DameIN United States

Electron charge qubits are compelling candidates for solid-state quantum computing because of their inherent simplicity in qubit design, fabrication, control and readout. However, electron charge qubits built on conventional semiconductors and superconductors suffer from severe charge noise that limits their coherence time to the order of one microsecond. Here we report electron charge qubits that exceed this limit, based on isolated single electrons trapped on an ultraclean solid neon surface in a vacuum. Quantum information is encoded in the motional states of an electron that is strongly coupled with microwave photons in an on-chip superconducting resonator. The measured relaxation and coherence times are both on the order of 0.1 ms, surpassing all existing charge qubits and rivalling state-of-the-art superconducting transmon qubits. The simultaneous strong coupling of two qubits with a common resonator is also demonstrated, as the first step towards two-qubit entangling gates for universal quantum computing. © 2023, The Author(s), under exclusive licence to Springer Nature Limited.

关键词： Superconducting resonators

来源：评论

学校读者我要写书评

暂无评论

Calibration of Ordinal Regression Networks

arXiv

引用

arXiv 2024年

作者： Kim, Daehwan Chung, Haejun Jang, Ikbeom Department of Artificial Intelligence Semiconductor Engineering Hanyang University Seoul Korea Republic of Department of Artificial Intelligence Hanyang University Seoul Korea Republic of Department of Electronic Engineering Hanyang University Seoul Korea Republic of Division of Computer Engineering Hankuk University of Foreign Studies Yongin Korea Republic of Division of AI Data Convergence Hankuk University of Foreign Studies Yongin Korea Republic of Division of Language & AI Hankuk University of Foreign Studies Yongin Korea Republic of

Recent studies have shown that deep neural networks are not well-calibrated and often produce over-confident predictions. The miscalibration issue primarily stems from using cross-entropy in classifications, which aims to align predicted softmax probabilities with one-hot labels. In ordinal regression tasks, this problem is compounded by an additional challenge: the expectation that softmax probabilities should exhibit unimodal distribution is not met with cross-entropy. The ordinal regression literature has focused on learning orders and overlooked calibration. To address both issues, we propose a novel loss function that introduces ordinal-aware calibration, ensuring that prediction confidence adheres to ordinal relationships between classes. It incorporates soft ordinal encoding and ordinal-aware regularization to enforce both calibration and unimodality. Extensive experiments across four popular ordinal regression benchmarks demonstrate that our approach achieves state-of-the-art calibration without compromising classification accuracy. © 2024, CC BY-NC-SA.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Optimal Prescribed Performance Control for Euler-Lagrange Systems

Optimal Prescribed Performance Control for Euler-Lagrange Sy...

引用

International Conference on Systems and Control (ICSC)

作者： Ioanna Malli Christos Vlachos Charalampos P. Bechlioulis Kostas J. Kyriakopoulos School of Mechanical Engineering National Technical University of Athens Greece Department of Electrical and Computer Engineering University of Patras Greece Engineering Division and Center for AI and Robotics New York University Abu Dhabi UAE

This work introduces an optimal control policy aiming to stabilize an Euler-Lagrange system of unknown dynamics, while simultaneously satisfying predefined response criteria. The proposed methodology unfolds in two stages. In the initial phase, a neural network is trained online via an iterative process to capture the system's dynamics. Subsequently, in the second phase, by utilizing the acquired dynamics, a successive approximation algorithm is employed to derive a nearly optimal control law while taking into consideration prescribed performance specifications such as convergence speed and steady state error. Validation of our method is achieved through a simulated paradigm, confirming the success of both the identification process and the minimization of the adopted integral cost.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Structural health monitoring of steel moment frame buildings via sequence-based recurrent neural networks

Structural health monitoring of steel moment frame buildings...

引用

Artificial Intelligence (Cai), IEEE Conference on

作者： Khashayar Heydarpour Doeun Choe Kyungyong Chung Department of Civil Engineering New Mexico State University Las Cruces USA Division of AI Computer Science&Eng Kyonggi University Suwon South Korea

Signal-based damage detection has gained extensive attention in recent years due to its capability in improving the deficiencies of previous structural health monitoring methods. Deep learning, with high capabilities in feature learning, has emerged as a powerful tool for sequence classification. In this paper, sequence-based deep learning models using long short-term memory (LSTM) and gated recurrent units (GRU) networks are used to detect structural damages and damage locations applied to steel building structures. To propose an appropriate deep-learning method for structural health monitoring, sets of monitoring data from the IASC-ASCE benchmark building were used. The data was collected from 15 sensors to collect accelerations attached to a 4-story steel moment frame building. The data has been properly pre-processed, denoised, sliced, and normalized. K-fold cross-validation validation is performed. The networks are designed using various combinations of recurrent neural networks, such as LSTM and GRU. It is concluded that stacked multilayer bidirectional long short-term memory networks, with an accuracy of 98%, have a superior performance in detecting the presence and location of structural damage.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Adaptive sampling for efficient softmax approximation 24

Adaptive sampling for efficient softmax approximation

引用

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Tavor Z. Baharav Ryan Kang Colin Sullivan Mo Tiwari Eric Luxenberg David Tse Mert Pilanci Eric and Wendy Schmidt Center Broad Institute Cambridge MA Department of Computer Science Stanford University Stanford CA AI Division Software Engineering Institute Pittsburgh PA Department of Computer Science Stanford University Stanford CA Gridmatic Cupertino CA Department of Electrical Engineering Stanford University Stanford CA

ISBN: (纸本)9798331314385

The softmax function is ubiquitous in machine learning and optimization applications. Computing the full softmax evaluation of a matrix-vector product can be computationally expensive in high-dimensional settings. In many applications, however, it is sufficient to calculate only the top few outputs of the softmax function. In this work, we present an algorithm, dubbed AdaptiveSoftmax, that adaptively computes the top k softmax values more efficiently than the full softmax computation, with probabilistic guarantees. We demonstrate the sample efficiency improvements afforded by AdaptiveSoftmax on real and synthetic data to corroborate our theoretical results. AdaptiveSoftmax yields > 10x gain over full softmax computation on most datasets, yielding up to 30x improvement for Mistral7B evaluated on the Wikitext dataset. The adaptive method we propose for estimating the partition function (the softmax denominator) is of independent interest and can be used in other applications such as kernel density estimation.

关键词：

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：