检索结果-内蒙古大学图书馆

International Conference on High Performance computing

作者： Abrar Hossain Abdel-Hameed A. Badawy Mohammad A. Islam Tapasya Patki Kishwar Ahmed Department of Electrical Engineering and Computer Science The University of Toledo Department of Electrical and Computer Engineering New Mexico State University Department of Computer Science and Engineering The University of Texas at Arlington Lawrence Livermore National Laboratory Center for Applied Scientific Computing

ISBN: (数字)9798331509095

ISBN: (纸本)9798331509101

The growing necessity for enhanced processing capabilities in edge devices with limited resources has led us to develop effective methods for improving high-performance computing (HPC) applications. In this paper, we introduce LASP (Lightweight Autotuning of scientific Application Parameters), a novel strategy designed to address the parameter search space challenge in edge devices. Our strategy employs a multi-armed bandit (MAB) technique focused on online exploration and exploitation. Notably, LASP takes a dynamic approach, adapting seamlessly to changing environments. We tested LASP with four HPC applications: Lulesh, Kripke, Clomp, and Hypre. Its lightweight nature makes it particularly well-suited for resource-constrained edge devices. By employing the MAB framework to efficiently navigate the search space, we achieved significant performance improvements while adhering to the stringent computational limits of edge devices. Our experimental results demonstrate the effectiveness of LASP in optimizing parameter search on edge devices.

关键词： Performance evaluation Analytical models Navigation High performance computing Computational modeling Data models Computational efficiency

来源：评论

学校读者我要写书评

暂无评论

Unknown-Aware Bilateral Dependency Optimization for Defending Against Model Inversion Attacks

引用

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年 PP卷 PP页

作者： Peng, Xiong Liu, Feng Wang, Nannan Lan, Long Liu, Tongliang Cheung, Yiu-Ming Han, Bo Hong Kong Baptist University Department of Computer Science Hong Kong The University of Melbourne School of Computing and Information Systems Australia Xidian University State Key Laboratory of Integrated Services Networks School of Telecommunications Engineering Shaanxi Xi'an710071 China National University of Defense Technology State Key Laboratory of High Performance Computing School of Computer Hunan Changsha410000 China The University of Sydney Sydney AI Centre School of Computer Science Faculty of Engineering Australia

By abusing access to a well-trained classifier, model inversion (MI) attacks pose a significant threat as they can recover the original training data, leading to privacy leakage. Previous studies mitigated MI attacks by imposing regularization to reduce the dependency between input features and outputs during classifier training, a strategy known as unilateral dependency optimization. However, this strategy contradicts the objective of minimizing the supervised classification loss, which inherently seeks to maximize the dependency between input features and outputs. Consequently, there is a trade-off between improving the model's robustness against MI attacks and maintaining its classification performance. To address this issue, we propose the bilateral dependency optimization strategy (BiDO), a dual-objective approach that minimizes the dependency between input features and latent representations, while simultaneously maximizing the dependency between latent representations and labels. BiDO is remarkable for its privacy-preserving capabilities. However, models trained with BiDO exhibit diminished capabilities in out-of-distribution (OOD) detection compared to models trained with standard classification supervision. Given the open-world nature of deep learning systems, this limitation could lead to significant security risks, as encountering OOD inputs—whose label spaces do not overlap with the in-distribution (ID) data used during training-is inevitable. To address this, we leverage readily available auxiliary OOD data to enhance the OOD detection performance of models trained with BiDO. This leads to the introduction of an upgraded framework, unknown-aware BiDO (BiDO+), which mitigates both privacy and security concerns. As a highlight, with comparable model utility, BiDO-HSIC+ reduces the FPR95 by 55.02% and enhances the AUCROC by 9.52% compared to BiDO-HSIC, while also providing superior MI robustness. © 1979-2012 IEEE.

关键词： Differential privacy

来源：评论

学校读者我要写书评

暂无评论

End-to-end infrared imaging system design with diffractive optical element 14

End-to-end infrared imaging system design with diffractive o...

引用

Optical Design and Testing XIV 2024

作者： Shi, Runzhou Zhou, Yan Zhang, Tian Shao, Yuqi Zhang, Hanwen Wei, Ruoyan Bai, Jian College of Optical Science and Engineering Zhejiang University Hangzhou310027 China ZJU-Hangzhou Global Scientific and Technological Innovation Center Zhejiang University Hangzhou311215 China State Key Laboratory of Extreme Photonics and Instrumentation Zhejiang University Hangzhou310027 China Tianjin Jinhang Institute of Technical Physics Tianjin300308 China

ISBN: (纸本)9781510682023

The hybrid refractive-diffractive optics system exhibits strong capabilities in achromatic and non-thermalized design, as well as information encoding. This paper introduces an innovative end-to-end design scheme for refractive-diffractive hybrid imaging optical systems, which optimizes both optical and neural network parameters simultaneously. An all-ray differentiable ray-tracing model is proposed to integrate lens and diffractive optical elements into a unified design framework, thereby maintaining precision by avoiding wave-to-ray conversion losses. The neural network is constructed based on the imaging characteristics of infrared hybrid optical systems and enhances aberration correction. The proposed method is applied to the design of a single-lens short-wave infrared imaging system, outperforming traditional discrete designs and demonstrating significant potential for infrared optical system applications. © 2024 SPIE.

关键词： Ray tracing

来源：评论

学校读者我要写书评

暂无评论

Microstructural impacts on ionic conductivity of oxide solid electrolytes from a combined atomistic-mesoscale approach

引用

npj Computational Materials 2021年第1期7卷 1959-1973页

作者： Tae Wook Heo Andrew Grieder Bo Wang Marissa Wood Tim Hsu Sneha A.Akhade Liwen F.Wan Long-Qing Chen Nicole Adelstein Brandon C.Wood Materials Science Division Lawrence Livermore National LaboratoryLivermoreCA 94550USA Department of Chemistry and Biochemistry San Francisco State UniversitySan FranciscoCA 94132USA Department of Materials Science and Engineering The Pennsylvania State UniversityUniversity ParkPA 16802USA Center for Applied Scientific Computing Lawrence Livermore National LaboratoryLivermoreCA 94550USA

Although multiple oxide-based solid electrolyte materials with intrinsically high ionic conductivities have emerged,practical processing and synthesis routes introduce grain boundaries and other interfaces that can perturb primary conduction *** directly probe these effects,we demonstrate an efficient and general mesoscopic computational method capable of predicting effective ionic conductivity through a complex polycrystalline oxide-based solid electrolyte microstructure without relying on simplified equivalent circuit *** parameterize the framework for Li_(7-x)La_(3)Zr_(2)0_(12)(LLZO)gamet solid electrolyte by combining synthetic microstructures from phase-field simulations with diffusivities from molecular dynamics simulations of ordered and disordered *** designed simulations reveal an interdependence between atomistic and mesoscopic microstructural impacts on the effective ionic conductivity of polycrystalline LLZO,quantified by newly defined metrics that characterize the com plex ionic transport *** results provide fundamental understanding of the physical origins of the reported variability in ionic conductivities based on an extensive analysis of literature data,while simultaneously outlining practical design guidance for achieving desired ionic transport properties based on conditions for which sensitivity to microstructural features is *** implications of our results are discussed,including a possible connection between ion conduction behavior and dendrite formation.

关键词： microstructure solid conductivity

来源：评论

学校读者我要写书评

暂无评论

Reinforcement Learning-Based Optimal Control for Continuous-Time Quantized Systems via Adaptive Dynamic Programming

Reinforcement Learning-Based Optimal Control for Continuous-...

引用

IEEE SSD International Multi-Conference on Systems, Signals and Devices

作者： Omar Qasem Weinan Gao Omar R. Daoud Electrical and Computer Engineering Department School of Engineering and Computing American International University Al-Jahra Kuwait State Key Laboratory of Synthetical Automation for Process Industries Northeastern University Shenyang Lianning China

ISBN: (数字)9798331542726

ISBN: (纸本)9798331542733

Ahstract- This paper presents a comprehensive investigation into the stability analysis and design conditions for quantizers to ensure the closed-loop stability of continuous-time linear quantized systems, under the conditions derived by the small-gain theorem. The study begins by deriving explicit bounds on the quantizer parameters required for maintaining system stability. Building on this foundation, an optimal controller is designed using the linear quadratic regulator (LQR) framework, providing an efficient data-driven control strategy. To further enhance the system's performance, an adaptive dynamic programming (ADP) algorithm, referred to as the hybrid iteration (HI) method, is developed. This algorithm effectively learns the optimal control policy by leveraging the trajectories of the quantized states and inputs, thereby addressing the challenges posed by quantization constraints. The proposed HI approach combines the advantages of adaptive learning and optimization, making it well-suited for continuous-time systems with limited information. The simulation results confirm that the ADP approach with the provided conditions not only stabilizes the quantized system but also achieves optimal control performance under the specified quantization conditions. This study offers valuable insights and a robust methodological framework for addressing stability and control challenges, with insights to be expanded to continuous-time nonlinear quantized systems, with potential applications in various engineering domains, such as networked systems, robotics and autonomous systems.

关键词： Quantization (signal) Adaptive systems Regulators Simulation Heuristic algorithms Optimal control Stability analysis Dynamic programming Trajectory Thermal stability

来源：评论

学校读者我要写书评

暂无评论

FIRE: a dataset for feedback integration and refinement evaluation of multimodal models 24

FIRE: a dataset for feedback integration and refinement eval...

引用

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

作者： Pengxiang Li Zhi Gao Bofei Zhang Tao Yuan Yuwei Wu Mehrtash Harandi Yunde Jia Song-Chun Zhu Qing Li Beijing Key Laboratory of Intelligent Information Technology School of Computer Science & Technology Beijing Institute of Technology and State Key Laboratory of General Artificial Intelligence BIGAI State Key Laboratory of General Artificial Intelligence BIGAI and State Key Laboratory of General Artificial Intelligence Peking University State Key Laboratory of General Artificial Intelligence BIGAI Beijing Key Laboratory of Intelligent Information Technology School of Computer Science & Technology Beijing Institute of Technology and Guangdong Laboratory of Machine Perception and Intelligent Computing Shenzhen MSU-BIT University Department of Electrical and Computer System Engineering Monash University Guangdong Laboratory of Machine Perception and Intelligent Computing Shenzhen MSU-BIT University and Beijing Key Laboratory of Intelligent Information Technology School of Computer Science & Technology Beijing Institute of Technology State Key Laboratory of General Artificial Intelligence BIGAI and State Key Laboratory of General Artificial Intelligence Peking University and Department of Automation Tsinghua University

ISBN: (纸本)9798331314385

Vision language models (VLMs) have achieved impressive progress in diverse applications, becoming a prevalent research direction. In this paper, we build FIRE, a feedback-refinement dataset, consisting of 1.1M multi-turn conversations that are derived from 27 source datasets, empowering VLMs to spontaneously refine their responses based on user feedback across diverse tasks. To scale up the data collection, FIRE is collected in two components: FIRE-100K and FIRE-1M, where FIRE-100K is generated by GPT-4V, and FIRE-1M is freely generated via models trained on FIRE-100K. Then, we build FIRE-Bench, a benchmark to comprehensively evaluate the feedback-refining capability of VLMs, which contains 11K feedback-refinement conversations as the test data, two evaluation settings, and a model to provide feedback for VLMs. We develop the FIRE-LLaVA model by fine-tuning LLaVA on FIRE-100K and FIRE-1M, which shows remarkable feedback-refining capability on FIRE-Bench and outperforms untrained VLMs by 50%, making more efficient user-agent interactions and underscoring the significance of the FIRE dataset.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Computational diffraction-limited imaging with a non-imaging spatial information transfer lens

引用

Photonics Research 2024年第10期12卷 2265-2278页

作者： YINGLI LIU ZHIHUI DING FENGXU GUO CHENG GUO SHENGNAN JIA HAIFENG LI RENGMAO WU XU LIU State Key Laboratory of Extreme Photonics and Instrumentation College of Optical Science and EngineeringZhejiang UniversityHangzhou 310027China Faculty of Computing Harbin Institute of TechnologyHarbin 150001China Department of General Surgery Sir Run Run Shaw HospitalSchool of MedicineZhejiang UniversityHangzhou 310000China

The demand for low-cost,high-performance miniaturized optical imaging systems requires creating a new imaging *** this paper,we propose an imaging paradigm that achieves diffraction-limited imaging with a non-imaging spatial information transfer *** spatial information transfer lens realizes a perfect match between the space–bandwidth product(SBP)of the lens and that of the image sensor so that the collected spatial information from the object can be totally recorded and fully resolved by the image sensor.A backward wave propagation model is developed to reconstruct the object by propagating the light wave modulated by the information transfer lens back from the image space to object *** proposed imaging paradigm breaks the point-to-point imaging structure and removes the focusing-distance constraint,allowing a flexible arrangement of the object and the image sensor along the optical axis with a compact form factor of the optical *** experimentally demonstrate the versatility and effectiveness of the proposed imaging *** proposed imaging paradigm is low-cost,simple in configuration,flexible in arrangement,and diffraction limited with great potential applications in biomedical imaging.

关键词： transfer arrangement image

来源：评论

学校读者我要写书评

暂无评论

SUMI-IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints 39

SUMI-IFL: An Information-Theoretic Framework for Image Forge...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Sheng, Ziqi Lu, Wei Luo, Xiangyang Zhou, Jiantao Cao, Xiaochun School of Computer Science and Engineering MoE Key Laboratory of Information Technology Guangdong Province Key Laboratory of Information Security Technology Sun Yat-sen University China State Key Laboratory of Mathematical Engineering and Advanced Computing China State Key Laboratory of Internet of Things for Smart City Department of Computer and Information Science University of Macau China School of Cyber Science and Technology Shenzhen Campus of Sun Yat-sen University China

ISBN: (纸本)157735897X

Image forgery localization (IFL) is a crucial technique for preventing tampered image misuse and protecting social safety. However, due to the rapid development of image tampering technologies, extracting more comprehensive and accurate forgery clues remains an urgent challenge. To address these challenges, we introduce a novel information-theoretic IFL framework named SUMI-IFL that imposes sufficiency-view and minimality-view constraints on forgery feature representation. First, grounded in the theoretical analysis of mutual information, the sufficiency-view constraint is enforced on the feature extraction network to ensure that the latent forgery feature contains comprehensive forgery clues. Considering that forgery clues obtained from a single aspect alone may be incomplete, we construct the latent forgery feature by integrating several individual forgery features from multiple perspectives. Second, based on the information bottleneck, the minimality-view constraint is imposed on the feature reasoning network to achieve an accurate and concise forgery feature representation that counters the interference of task-unrelated features. Extensive experiments show the superior performance of SUMI-IFL to existing state-of-the-art methods, not only on in-dataset comparisons but also on cross-dataset comparisons. © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Correlation theory

来源：评论

学校读者我要写书评

暂无评论

Effects of Tooth Directional Modification Error on Internal Excitation and Vibration Characteristics of Gears with Early-Stage Crack Fault 1

引用

UNIfied Conference of International Workshop on Defence Applications of Multi-Agent Systems, DAMAS 2023, International Conference on Maintenance engineering, IncoME-V 2023, International conference on the Efficiency and Performance engineering Network, TEPEN 2023

作者： Yang, Lantao Yu, Wennian Zou, Desheng Wang, Liming Gu, Fengshou Mba, David State Key Laboratory of Mechanical Transmission Chongqing University Chongqing400044 China Centre for Efficiency and Performance Engineering University of Huddersfield HuddersfieldHD1 3DH United Kingdom Creative Computing Institute University of the Arts London LondonWC1V 7EY United Kingdom

ISBN: (数字)9783031494215

ISBN: (纸本)9783031494208

Tooth directional modification (TDM) is a widely used method to eliminate the stress concentration at the edge of gear teeth, and plays a key role in vibration and noise reduction and life extension of gears. The existence of TDM and early-stage crack (ESC) will both change the tooth contact characteristic and then cause the variation of internal excitation and vibration characteristics of the gear system. However, the coupling effect of the ESC and TDM on the tooth contact characteristic is usually ignored in the existing internal excitation modelling and vibration characteristic simulation of the gear system with ESC, causing the effect of TDM on the ESC excitation characteristics is still unclear. To address this shortcoming, an internal excitation analytical model of the gear system with ESC is established considering this coupling. On the basis of verifying the correctness and accuracy of the proposed model, the influence mechanism of TDM on the internal excitation and vibration characteristics of the gear with ESC is investigated. The results show that the effects of the gear contact characteristic changes caused by the TDM on the gear internal excitation are much greater than that resulting from the ESC, and the vibration characteristics of the ESC can be easily distracted by the TDM errors. Therefore, it is suggested that the TDM errors should be taken into account in the early-stage fault simulation or diagnosis method establishment to avoid the simulation results deviating from the actual situation. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

关键词： Errors

来源：评论

学校读者我要写书评

暂无评论

Encoding Temporal Change of Longitudinal EBV DNA for Dynamic Risk Prediction of Nasopharyngeal Carcinoma 7

Encoding Temporal Change of Longitudinal EBV DNA for Dynamic...

引用

7th International Conference on Biological Information and Biomedical engineering, BIBE 2024

作者： Yang, Fangfang Zhen, Zicheng Lin, Li Liu, Yuyu Gao, Rui Zhang, Taoyu Lu, Lijun Lv, Wenbing School of Information Yunnan University Yunnan Kunming650504 China Yunnan Key Laboratory of Intelligent Systems and Computing Yunnan University Yunnan Kunming650504 China Department of Radiation Oncology Sun Yat-sen University Cancer Center State Key Laboratory of Oncology in South China Collaborative Innovation Center for Cancer Medicine Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy Guangzhou510060 China School of Software Yunnan University Yunnan Kunming650504 China School of Biomedical Engineering Southern Medical University 1023 Shatai Road Guangzhou510515 China

ISBN: (纸本)9783800762910

—To accurately predict the risk of recurrence and metastasis of nasopharyngeal carcinoma (NPC) patients, an image-encoded risk prediction model was proposed. 752 patients (training vs. testing: 527 vs. 225) were retrospectively collected, and the plasma EBV DNA copies were tested 13 times during treatment. Based on 13 times plasma EBV DNA copies, three kinds of image encoding methods were employed to characterize the linear and nonlinear dynamic change of longitudinal EBV DNA, i.e. Gramian Angular Summation Field (GASF), Gramian Angular Difference Field (GADF), and Recurrence Plot (RP). The Cox proportional hazard regression model was constructed by incorporating both image and clinical features as covariates to determine the risk factors and prediction models for disease-free survival (DFS). On the test set, the proposed model combining image and clinical features achieved C-index and AUC of 0.777 and 0.776, respectively outperforming the model relying solely on clinical features (C-index = 0.667, AUC = 0.675). In conclusion, the dynamic risk prediction model with image encoding can monitor changes in plasma EBV DNA copies during treatment, aiding doctors in assessing whole-body tumor burden, accurately predicting the dynamic risk of recurrence and metastasis, and planning individualized treatments in NPC patients. © VDE VERLAG GMBH.

关键词： Gene encoding

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：