检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Oh, Yeongtak Lee, Saehyung Hwang, Uiwon Yoon, Sungroh Department of Electrical and Computer Engineering Seoul National University Korea Republic of Interdisciplinary Program in Artificial Intelligence Seoul National University Korea Republic of Division of Digital Healthcare Yonsei University Korea Republic of

Large-scale language-vision pre-training models, such as CLIP, have achieved remarkable text-guided image morphing results by leveraging several unconditional generative models. However, existing CLIP-guided image morphing methods encounter difficulties when morphing photorealistic images. Specifically, existing guidance fails to provide detailed explanations of the morphing regions within the image, leading to misguidance. In this paper, we observed that such misguidance could be effectively mitigated by simply using a proper regularization loss. Our approach comprises two key components: 1) a geodesic cosine similarity loss that minimizes inter-modality features (i.e., image and text) on a projected subspace of CLIP space, and 2) a latent regularization loss that minimizes intra-modality features (i.e., image and image) on the image manifold. By replacing the naïve directional CLIP loss in a drop-in replacement manner, our method achieves superior morphing results on both images and videos for various benchmarks, including CLIP-inversion. © 2024, CC BY.

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

FedClassAvg: Local Representation Learning for Personalized Federated Learning on Heterogeneous Neural Networks 22

FedClassAvg: Local Representation Learning for Personalized ...

引用

Proceedings of the 51st International Conference on Parallel Processing

作者： Jaehee Jang Heoneok Ha Dahuin Jung Sungroh Yoon Department of Electrical and Computer Engineering Seoul National University Republic of Korea Department of Electrical and Computer Engineering and Interdisciplinary Program in Artificial Intelligence Seoul National University Republic of Korea and AIIS ASRI and INMC Seoul National University Republic of Korea

ISBN: (纸本)9781450397339

Personalized federated learning is aimed at allowing numerous clients to train personalized models while participating in collaborative training in a communication-efficient manner without exchanging private data. However, many personalized federated learning algorithms assume that clients have the same neural network architecture, and those for heterogeneous models remain understudied. In this study, we propose a novel personalized federated learning method called federated classifier averaging (FedClassAvg). Deep neural networks for supervised learning tasks consist of feature extractor and classifier layers. FedClassAvg aggregates classifier weights as an agreement on decision boundaries on feature spaces so that clients with not independently and identically distributed (non-iid) data can learn about scarce labels. In addition, local feature representation learning is applied to stabilize the decision boundaries and improve the local feature extraction capabilities for clients. While the existing methods require the collection of auxiliary data or model weights to generate a counterpart, FedClassAvg only requires clients to communicate with a couple of fully connected layers, which is highly communication-efficient. Moreover, FedClassAvg does not require extra optimization problems such as knowledge transfer, which requires intensive computation overhead. We evaluated FedClassAvg through extensive experiments and demonstrated it outperforms the current state-of-the-art algorithms on heterogeneous personalized federated learning tasks.

关键词： Communication Efficient Resource Constraint Representation Learning Neural Networks Model Heterogeneity Federated Learning

来源：评论

学校读者我要写书评

暂无评论

SF(DA)2: SOURCE-FREE DOMAIN ADAPTATION THROUGH THE LENS OF DATA AUGMENTATION

arXiv

引用

arXiv 2024年

作者： Hwang, Uiwon Lee, Jonghyun Shin, Juhyeon Yoon, Sungroh Division of Digital Healthcare Yonsei University Korea Republic of Department of Electrical and Computer Engineering Seoul National University Korea Republic of Interdisciplinary Program in Artificial Intelligence Seoul National University Korea Republic of

In the face of the deep learning model’s vulnerability to domain shift, source-free domain adaptation (SFDA) methods have been proposed to adapt models to new, unseen target domains without requiring access to source domain data. Although the potential benefits of applying data augmentation to SFDA are attractive, several challenges arise such as the dependence on prior knowledge of class-preserving transformations and the increase in memory and computational requirements. In this paper, we propose Source-free Domain Adaptation Through the Lens of Data Augmentation (SF(DA)2), a novel approach that leverages the benefits of data augmentation without suffering from these challenges. We construct an augmentation graph in the feature space of the pretrained model using the neighbor relationships between target features and propose spectral neighborhood clustering to identify partitions in the prediction space. Furthermore, we propose implicit feature augmentation and feature disentanglement as regularization loss functions that effectively utilize class semantic information within the feature space. These regularizers simulate the inclusion of an unlimited number of augmented target features into the augmentation graph while minimizing computational and memory demands. Our method shows superior adaptation performance in SFDA scenarios, including 2D image and 3D point cloud datasets and a highly imbalanced dataset. Copyright © 2024, The Authors. All rights reserved.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Modeling and Designing Non-Pharmaceutical Interventions in Epidemics: A Submodular Approach

arXiv

引用

arXiv 2024年

作者： Cheng, Shiyu Niu, Luyao Ramasubramanian, Bhaskar Clark, Andrew Poovendran, Radha Department of Electrical and Systems Engineering Washington University in St. Louis St. LouisMO United States Department of Electrical and Computer Engineering University of Washington SeattleWA United States Network Security Lab Department of Electrical and Computer Engineering University of Washington SeattleWA United States Electrical and Computer Engineering Program Western Washington University BellinghamWA United States

This paper considers the problem of designing non-pharmaceutical intervention (NPI) strategies, such as masking and social distancing, to slow the spread of a viral epidemic. We formulate the problem of jointly minimizing the infection probabilities of a population and the cost of NPIs based on a Susceptible-Infected-Susceptible (SIS) propagation model. To mitigate the complexity of the problem, we consider a steady-state approximation based on the quasi-stationary (endemic) distribution of the epidemic, and prove that the problem of selecting a minimum-cost strategy to satisfy a given bound on the quasi-stationary infection probabilities can be cast as a submodular optimization problem, which can be solved in polynomial time using the greedy algorithm. We carry out experiments to examine effects of implementing our NPI strategy on propagation and control of epidemics on a Watts-Strogatz small-world graph network. We find the NPI strategy reduces the steady state of infection probabilities of members of the population below a desired threshold value. © 2024, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Coupled adsorption-photocatalysis process for the removal of diclofenac using magnetite/reduced graphene oxide nanocomposite

引用

Chemosphere 2024年 349卷 140788页

作者： Lee, Jooyoung Jo, Seong-Jun Yoon, Soyeong Ko, Mingi Jang, Taesoon Kim, Hyun-Kyung Park, Jeong-Ann Department of Environmental Engineering Kangwon National University Chuncheon24341 Korea Republic of Department of Battery Convergence Engineering Kangwon National University Chuncheon24341 Korea Republic of Interdisciplinary Program in Advanced Functional Materials and Devices Development Kangwon National University Chuncheon24341 Korea Republic of Department of Integrated Energy and Infra System Kangwon National University Chuncheon24341 Korea Republic of

Diclofenac (DCF) is frequently detected in water bodies (ng/L to g/L) as it is not completely removed by conventional wastewater treatment plants. Adsorption and photocatalysis have been studied as promising methods for treating DCF;however, both processes have limitations. Thus, in this study, the removal efficiency of DCF is evaluated using a magnetite/reduced graphene oxide (Fe3O4/RGO) nanocomposite via a coupled adsorption-catalysis process. The Fe3O4/RGO nanocomposite was successfully synthesized using a microwave-assisted solvothermal method and exhibited a bandgap of 2.60 eV. The kinetic data best fitted the Elovich model (R2 = 0.994, χ2 = 0.29), indicating rapid adsorption. The maximum DCF adsorption capacity calculated using the Langmuir model was 80.33 mg/g. An ultraviolet C (UVC) light source and 0.1 g/L of Fe3O4/RGO nanocomposite were the optimum conditions for the removal of DCF (C0 = 30 mM) by a coupled adsorption-photocatalysis process (first-order rate constant (k) = 0.088/min), which was greater than the single adsorption (k = 0.029/min) and pre-adsorption and post-photocatalysis (k = 0.053/min) processes. This indicates that the adsorbed DCF did not hamper the photocatalytic reaction of the Fe3O4/RGO nanocomposite, but rather enhanced the coupled adsorption-photocatalytic reaction. DCF removal efficiency was higher at acidic conditions (pH 4.3–5.0), because high H+ promotes the generation of certain reactive oxygen species (ROS) and increases of electrostatic interaction. The presence of NaCl and CaCl2 (10 mM) did not notably affect the total DCF removal efficiency;however, Ca2+ affected the initial DCF adsorption affinity. Scavenger experiments demonstrated O2∙− and h+ play a key ROS than ·OH to degrade DCF. The acute toxicity of DCF towards Aliivibrio fischeri gradually decreased with increasing treatment time. © 2023 Elsevier Ltd

关键词： Magnetite

来源：评论

学校读者我要写书评

暂无评论

Variance-stationary Differentiable NAS 32

Variance-stationary Differentiable NAS

引用

32nd British Machine Vision Conference, BMVC 2021

作者： Choe, Hyeokjun Na, Byunggook Mok, Jisoo Yoon, Sungroh Department of Electrical and Computer Engineering Seoul National University Seoul Korea Republic of Interdisciplinary Program in AI Seoul National University Seoul Korea Republic of

Differentiable architecture search (DARTS) has become the popular method of neural architecture search (NAS) due to its adaptability and low computational cost. However, following the publication of DARTS, it has been found that DARTS often yields a suboptimal neural architecture because architecture parameters do not accurately represent operation strengths. Through extensive theoretical analysis and empirical observations, we reveal that this issue occurs as a result of the existence of unnormalized operations. Based on our finding, we propose a novel variance-stationary differentiable architecture search (VS-DARTS), which consists of node normalization, local adaptive learning rate, and pβ-continuous relaxation. Comprehensively, VS-DARTS makes the architecture parameters a more reliable metric for deriving a desirable architecture without increasing the search cost. In addition to the theoretical motivation behind all components of VS-DARTS, we provide strong experimental results to demonstrate that they synergize to significantly improve the search performance. The architecture searched by VS-DARTS achieves the test error of 2.50% on CIFAR-10 and 24.7% on ImageNet. © 2021. The copyright of this document resides with its authors.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

Land Surface Temperature Anomalies as Indicator of Land Cover Change: Case Study Over Chennai City

Land Surface Temperature Anomalies as Indicator of Land Cove...

引用

2021 IEEE India Geoscience and Remote Sensing Symposium, InGARSS 2021

作者： Roy, Anusha Harod, Rahul Rajasekaran, Eswar Indian Institute of Technology Bombay Department of Civil Engineering Mumbai India Indian Institute of Technology Bombay Interdisciplinary Program in Climate Studies Mumbai India

ISBN: (纸本)9781665442497

Rapid urbanization over the years has led to the loss of natural land cover, thereby affecting the Land Surface Temperature (LST) distribution in urban areas. The aim of this study is to analyze LST anomalies (calculated as the deviation from long term mean) over Chennai and check if critical land cover changes can be identified from them. LST from Landsat thermal infrared (TIR) images acquired in the months of April and May from the year 1988 onwards were used to estimate the LST anomalies. Positive LST anomalies were observed over areas that have been converted from a natural vegetated state to impervious surface conditions. There has been an overall shift towards positive anomalies over Chennai city,indicating a clear sign of urbanization. © 2021 IEEE.

关键词： Atmospheric temperature

来源：评论

学校读者我要写书评

暂无评论

Porous silicon on paper sensors: Factors affecting biosensing performance 17

Porous silicon on paper sensors: Factors affecting biosensin...

引用

Frontiers in Biological Detection: From Nanosensors to systems XVII 2025

作者： An, Huijin Laibinis, Paul E. Locke, Andrea K. Weiss, Sharon M. Interdisciplinary Material Science Program Vanderbilt University NashvilleTN37235 United States Department of Electrical and Computer Engineering Vanderbilt University NashvilleTN37235 United States Department of Chemical and Biomolecular Engineering Vanderbilt University NashvilleTN37235 United States Department of Chemistry Vanderbilt University NashvilleTN37235 United States Department of Biomedical Engineering Vanderbilt University NashvilleTN37235 United States

ISBN: (纸本)9781510684249

A paper-based biosensor integrating a functionalized porous silicon (PSi) membrane as the active sensing element for quantifiable protein detection has been developed. For similar short-time exposures to an analyte, improved molecular transport in PSi membranes when on paper leads to larger signal changes compared to traditional PSi films that remain on a silicon substrate. In this work, we discuss controlling the incubation time of the analyte and the overall testing time of the sensor by incorporating different combinations of wicking and absorbent paper beneath the PSi membrane. With this control, the PSi-on-paper sensor platform has the potential to serve as an effective low-cost rapid diagnostic test with highly sensitive, quantitative readout for a wide range of analytes. © 2025 SPIE.

关键词： Porous silicon

来源：评论

学校读者我要写书评

暂无评论

Bilateral-Head Region-Based Convolutional Neural Networks: A Unified Approach for Incremental Few-Shot Object Detection

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第9期5卷 4376-4390页

作者： Li, Yiting Zhu, Haiyue Tian, Sichao Ma, Jun Xiang, Cheng Vadakkepat, Prahlad National University of Singapore Department of Electrical and Computer Engineering 117583 Singapore Singapore Institute of Manufacturing Technology Agency for Science Technology and Research 138634 Singapore Peking Union Medical College Institute of Medicinal Plant Development Chinese Academy of Medical Sciences Beijing100193 China Robotics and Autonomous Systems Thrust Guangzhou511453 China The Hong Kong University of Science and Technology Division of Emerging Interdisciplinary Areas Hong Kong SAR Hong Kong Hkust Shenzhen-Hong Kong Collaborative Innovation Research Institute Shenzhen Futian518045 China

Practical object detection systems are highly desired to be open-ended for learning on frequently evolved datasets. Moreover, learning with little supervision further adds flexibility for real-world applications such as autonomous driving and robotics, where large-scale datasets could be prohibitive or expensive to obtain. However, continual adaption with small training examples often results in catastrophic forgetting and dramatic overfitting. To address such issues, a compositional learning system is proposed to enable effective incremental object detection from nonstationary and few-shot data streams. First of all, a novel bilateral-head framework is proposed to decouple the representation learning of base (pretrained) and novel (few-shot) classes into separate embedding spaces, which takes care of novel concept integration and base knowledge retention simultaneously. Moreover, to enhance learning stability, a robust parameter updating rule, i.e., recall and progress mechanism, is carried out to constrain the optimization trajectory of sequential model adaption. Beyond that, to enforce intertask class discrimination with little memory burden, we present a between-class regularization method that expands the decision space of few-shot classes for constructing unbiased feature representation. Final, we deeply investigate the incomplete annotation issue considering the realistic scenario of incremental few-shot object detection (iFSOD) and propose a semisupervised object labeling mechanism to accurately recover the missing annotations for previously encountered classes, which further enhances the robustness of the target detector to counteract catastrophic forgetting. Extensive experiments conducted on both Pascal visual object classes dataset (VOC) and microsoft common objects in context dataset (MS-COCO) datasets demonstrate the effectiveness of our method. © 2020 IEEE.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Collaborative Strategy for Software engineering Courses at a South American University 13

Collaborative Strategy for Software Engineering Courses at a...

引用

13th International Conference on Computer Supported Education, CSEDU 2021

作者： Feijóo-García, Miguel Alfonso Ramírez-Arévalo, Helio Henry Feijóo-García, Pedro Guillermo Program of Systems Engineering Universidad El Bosque Bogotá Colombia Department of CISE University of Florida GainesvilleFL United States

ISBN: (纸本)9789897585029

Software engineering (SE) is the discipline that integrates theory, methods, and tools to promote the development of new informatic solutions for multiple contexts. The discipline is generally introduced in Computer Science (CS) programs between the sophomore and junior years, adding the human being as an actor who participates in teamwork strategies to optimize time and effort. We report on an inter-curricular collaborative instructional strategy between two subsequent SE core courses—SE1 and SE2, at Universidad El Bosque, Colombia. We evaluated our strategy considering students’ performance and perceptions, basing our analysis on their grades, Likert scale (1-5) responses, and the sentiment of their open-ended feedback—we calculated it with Natural Language Processing (NLP) techniques. Our findings suggest that an inter-curricular strategy like the one we present can foster students’ performance, engagement, and motivation. Moreover, the strategy supports the promotion of SE skills, such as communication and teamwork. Copyright © 2021 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved.

关键词： Software engineering

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：