检索结果-内蒙古大学图书馆

Dual-View Pyramid Pooling in Deep Neural Networks for Improved Medical Image Classification and Confidence Calibration

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhang, Xiaoqing Nie, Qiushi Xiao, Zunjie Zhao, Jilu Wu, Xiao Guo, Pengxin Li, Runzhi Liu, Jin Wei, Yanjie Pan, Yi Center for High Performance Computing Shenzhen Key Laboratory of Intelligent Bioinformatics Shenzhen Institute of Advanced Technology Chinese Academy of Sciences Guangdong Shenzhen518055 China Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen518055 China Department of Statistics and Actuarial Science The University of Hong Kong 999077 Hong Kong Cooperative Innovation Center of Internet Healthcare Zhengzhou University Zhengzhou450001 China Hunan Provincial Key Lab on Bioinformatics School of Computer Science and Engineering Central South University Changsha410083 China Xinjiang Engineering Research Center of Big Data and Intelligent Software School of software Xinjiang University Wulumuqi830046 China Faculty of Computer Science and Control Engineering Shenzhen University of Advanced Technology Shenzhen China

Spatial pooling (SP) and cross-channel pooling (CCP) operators have been applied to aggregate spatial features and pixel-wise features from feature maps in deep neural networks (DNNs), respectively. Their main goal is to reduce computation and memory overhead without visibly weakening the performance of DNNs. However, SP often faces the problem of losing the subtle feature representations, while CCP has a high possibility of ignoring salient feature representations, which may lead to both miscalibration of confidence issues and suboptimal medical classification results. To address these problems, we propose a novel dual-view framework, the first to systematically investigate the relative roles of SP and CCP by analyzing the difference between spatial features and pixel-wise features. Based on this framework, we propose a new pooling method, termed dual-view pyramid pooling (DVPP), to aggregate multi-scale dual-view features. DVPP aims to boost both medical image classification and confidence calibration performance by fully leveraging the merits of SP and CCP operators from a dual-axis perspective. Additionally, we discuss how to fulfill DVPP with five parameter-free implementations. Extensive experiments on six 2D/3D medical image classification tasks show that our DVPP surpasses state-of-the-art pooling methods in terms of medical image classification results and confidence calibration across different DNNs. © 2024, CC BY.

关键词： Deep neural networks

Incentive-Scheduling Algorithms to Provide Green Computational Data Center

学校读者我要写书评

暂无评论

SN computer science 2021年第4期2卷 252页

作者： Haruna, Ahmed Abba Jung, Low Tan Arputharaj, Vijay Muhammad, L.J. College of Computer Science and Engineering University of Hafr Al Batin Al Jamiah Hafar Al Batin 39524d Saudi Arabia High Performance Computing Cloud Centre Universiti Teknologi PETRONAS Seri Iskandar Perak 32610 Malaysia Faculty of Science and Information Technology Skyline University Nigeria No. 2 Zaria Road Kano Nigeria Mathematics and Computer Science Department Federal University Kashere Gombe State Kashere Nigeria

The increased computational loads in grid servers are dissipating more heat to eventually amplifies the cooling demand in the data center (DC). This can lead to more submitted jobs missing their job completion deadlines. Unfortunately, the conventional DC job scheduling approaches do not provide compensation to the resource users on their submitted jobs that missed the deadlines. The absence of compensation may dissuade users from submitting jobs to the DCs. While the free air-cooling strategy is used elsewhere, it is not generally applicable in tropical countries such as Malaysia. A constant artificial cooling (air conditioning) is needed to sustain the DC long-hour operation. To solve this issue in the tropical region, green incentive-scheduling algorithms were devised to significantly save the cooling electricity consumption cost in DC and to be able to compensate users (as an incentive) for their submitted jobs that missed the job completion deadline(s). However, there is a need to further analyse the performance of the proposed green incentive-scheduling algorithms using a different benchmark traces file and to conduct a cost-saving comparison analysis between the green incentive-scheduling and the non-green incentive-scheduling algorithms. Therefore, this paper's authors observed and analysed the performance comparison of the proposed green incentive-based job scheduling algorithms using a benchmark traces file and conducted a cost-saving comparison analysis between the green incentive and the non-green incentive-based scheduling algorithms. Ultimately, the green incentive-scheduling algorithms are to show their impact on reducing the DC operation costs (e.g. electricity bills for cooling) and to provide fair compensation to grid users for a win–win situation. © 2021, The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd.

关键词： Data center Green computing Incentive-based job scheduling algorithm Server Temperature

Optical Anisotropy in van der Waals materials: Impact on Direct Excitation of Plasmons and Photons by Quantum Tunneling

学校读者我要写书评

暂无评论

Light(science & Applications) 2021年第12期10卷 2387-2398页

作者： Zhe Wang Vijith Kalathingal Thanh Xuan Hoang Hong-Son Chu Christian A.Nijhuis Department of Chemistry National University of Singapore3 Science Drive 3Singapore 117543Singapore Centre for Advanced 2D Materials National University of Singapore6 Science Drive 2Singapore 117564Singapore Department of Electrical and Computer Engineering National University of Singapore4 Engineering Drive 3Singapore 117583Singapore Department of Electronics and Photonics Institute of High Performance ComputingA*STAR(Agency for ScienceTechnology and Research)1 Fusionopolis WaySingapore 138632Singapore Hybrid Materials for Opto-Electronics Group Department of Molecules and MaterialsMESA+Institute for Nanotechnology and Center for Brain-Inspired Nano SystemsFaculty of Science and TechnologyUniversity of Twente7500 AE EnschedeThe Netherlands

Inelastic quantum mechanical tunneling of electrons across plasmonic tunnel junctions can lead to surface plasmon polariton(SPP)and photon *** far,the optical properties of such junctions have been controlled by changing the shape,or the type of the material,of the electrodes,primarily with the aim to improve SPP or photon emission *** we show that by tuning the tunneling barrier itself,the efficiency of the inelastic tunneling rates can be improved by a factor of *** exploit the anisotropic nature of hexagonal boron nitride(hBN)as the tunneling barrier material in Au//hBN//graphene tunnel junctions where the Au electrode also serves as a plasmonic strip *** this junction constitutes an optically transparent hBN–graphene heterostructure on a glass substrate,it forms an open plasmonic system where the SPPs are directly coupled to the dedicated strip waveguide and photons outcouple to the far *** experimentally and analytically show that the photon emission rate per tunneling electron is significantly improved(~×3)in Au//hBN//graphene tunnel junction due to the enhancement in the local density of optical states(LDOS)arising from the hBN *** the dedicated strip waveguide,SPP outcoupling efficiency is quantified and is found to be∼80%stronger than the radiative outcoupling in Au//hBN//graphene due to the high LDOS of the SPP decay channel associated with the inelastic *** new insights elucidated here deepen our understanding of plasmonic tunnel junctions beyond the isotropic models with enhanced LDOS.

关键词： tunneling strip coupling

performance analysis of ACO-based improved virtual machine allocation in cloud for IoT-enabled healthcare

学校读者我要写书评

暂无评论

Performance analysis of ACO-based improved virtual machine a...

作者： Kavitha, Kadarla Sharma, S.C. High Performance Computing Lab Department of Computer Science and Engineering Indian Institute of Technology Roorkee Roorkee India Electronics and Computer Discipline Indian Institute of Technology Roorkee Roorkee India

The Internet of Things (IoT)–enabled healthcare environment irregularly requires the resources from the Cloud to handle massive amounts of data, which impacts the response times of the Cloud. A typical healthcare application scenario could be to monitor the heart-patients who require the highest response times, immediate attention of all the stakeholders, and very quick decisions. This paper replaces the default First-Come-First-Serve (FCFS) Virtual Machine (VM) allocation scheme of the Cloud computing in CloudSim with the implementation of Ant Colony Optimization (ACO) by the efficient tuning of parameters. An IoT-enabled healthcare environment is designed, which irregularly requires the resources from the Cloud to handle massive amounts of data to assess the impact of the response times of the Cloud. Several experiments have been carried out to assess the optimal VM allocation using ACO by varying different parameters like the variation of the number of ants, the strength of pheromone, the number of VMs, the number of hosts, the strength of computing power of processing elements, the number of users, and the size of the workloads. Experimental results indicate that the ACO optimally utilizes the resources in a Cloud when compared to the default FCFS strategy. The results indicate that on an average, the ACO gives 25% to 30% improved response times than FCFS in the given healthcare and Cloud scenarios. © 2019 John Wiley & Sons, Ltd.

关键词： Ant colony optimization

Rethinking Client Drift in Federated Learning: A Logit Perspective

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Yan, Yunlu Feng, Chun-Mei Ye, Mang Zuo, Wangmeng Li, Ping Goh, Rick Siow Mong Zhu, Lei Chen, C.L. Philip Guangzhou Nansha511400 China The Institute of High Performance Computing A*STAR Singapore138632 Singapore The Hubei Luojia Laboratory National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Wuhan430072 China The School of Computer Science and Technology Harbin Institute of Technology Harbin130407 China The Hong Kong University of Science and Technology Hong Kong The Department of Computing the School of Design The Hong Kong Polytechnic University Hong Kong The School of Computer Science and Engineering South China University of Technology Guangzhou510006 China The Pazhou Lab Guangzhou510335 China

Federated Learning (FL) enables multiple clients to collaboratively learn in a distributed way, allowing for privacy protection. However, the real-world non-IID data will lead to client drift which degrades the performance of FL. Interestingly, we find that the difference in logits between the local and global models increases as the model is continuously updated, thus seriously deteriorating FL performance. This is mainly due to catastrophic forgetting caused by data heterogeneity between clients. To alleviate this problem, we propose a new algorithm, named FedCSD, a Class prototype Similarity Distillation in a federated framework to align the local and global models. FedCSD does not simply transfer global knowledge to local clients, as an undertrained global model cannot provide reliable knowledge, i.e., class similarity information, and its wrong soft labels will mislead the optimization of local models. Concretely, FedCSD introduces a class prototype similarity distillation to align the local logits with the refined global logits that are weighted by the similarity between local logits and the global prototype. To enhance the quality of global logits, FedCSD adopts an adaptive mask to filter out the terrible soft labels of the global models, thereby preventing them to mislead local optimization. Extensive experiments demonstrate the superiority of our method over the state-of-the-art federated learning approaches in various heterogeneous settings. The source code will be released. Copyright © 2023, The Authors. All rights reserved.

关键词： Distillation

New Perspectives on Recommender Systems for Industries

学校读者我要写书评

暂无评论

New Perspectives on Recommender Systems for Industries

Artificial Intelligence for Industries (AI4I)

作者： Mouzhi Ge Giovanni Pilato Fabio Persia Daniela D’Auria Faculty of European Campus Rottal-Inn Deggendorf Institute of Technology Deggendorf Germany National Research Council of Italy Institute for High Performance Computing and Networking Palermo italy Department of Information Engineering Computer Science and Mathematics University of L’Aquila L’Aquila italy Faculty of Computer Science Free University of Bozen-Bolzano Bozen-Bolzano italy

Nowadays, recommender systems are increasingly being exploited in many industrial applications, including virtual museums and movie streaming platforms. In the last few years, some new perspectives provided by research paradigms such as deep learning or quantum computing, have arisen. As a result, this paper identifies four new perspectives on recommender systems: e-health, tourism, deep-learning-based, and recommender systems exploiting quantum computing. After discussing them, the paper provides the current state of the art and highlights the possible future directions for industries.

关键词： Industries Deep learning Training Quantum computing Scalability Computational modeling Virtual museums

Thermal-aware Energy Efficient Task Scheduling Framework

学校读者我要写书评

暂无评论

Thermal-aware Energy Efficient Task Scheduling Framework

IEEE International Symposium on Parallel and Distributed Processing with Applications and IEEE International Conference on Ubiquitous computing and Communications (ISPA/IUCC)

作者： Jian Nong Jia Chen Yinqing Wang Wei Qin Xi He Faculty of Innovation Engineering-School of Computer Science and Engineering Macau University of Science and Technology Macau China Guangxi Key Laboratory of Machine Vision and Intelligent Control Wuzhou University Wuzhou China Guangxi Colleges and Universities Key Laboratory of Industry Software Technology Wuzhou University Wuzhou China High Performance Computing Lab Wuzhou University Wuzhou China

Large energy consumption in data centers has become a challenging problem with the emergence of cloud computing and large scale data centers. In this paper, we present an architectural framework for thermal-aware resource management while considering energy efficiency. The framework consists of a layered architecture and integrates a set of easy-to-use client tools and a thermal-aware task management middleware to schedule tasks based on thermal conditions within a cluster and among different data centers. As part of this paper we focus on the development of a thermal-aware task scheduling component for a single data center. This component is fundamental to our future activities, while considering to balance the temperature distribution in a single data center, thus implicitly minimizing energy cost in data centers.

关键词： Data centers Temperature distribution Schedules Processor scheduling Thermal management Ubiquitous computing Energy efficiency

Polar-Net: A Clinical-Friendly Model for Alzheimer’s Disease Detection in OCTA Images

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Liu, Shouyue Hao, Jinkui Xu, Yanwu Fu, Huazhu Guo, Xinyu Liu, Jiang Zheng, Yalin Liu, Yonghuai Zhang, Jiong Zhao, Yitian Cixi Institute of Biomedical Engineering Ningbo Institute of Materials Technology and Engineering Chinese Academy of Sciences China Cixi Biomedical Research Institute Wenzhou Medical University China School of Future Technology South China University of Technology Guangzhou China Pazhou Lab Guangzhou China Institute of High-Performance Computing Agency for Science Technology and Research Singapore Department of Computer Science Southern University of Science and Technology China Department of Eye and Vision Science University of Liverpool United Kingdom Department of Computer Science Edge Hill University United Kingdom

Optical Coherence Tomography Angiography (OCTA) is a promising tool for detecting Alzheimer’s disease (AD) by imaging the retinal microvasculature. Ophthalmologists commonly use region-based analysis, such as the ETDRS grid, to study OCTA image biomarkers and understand the correlation with AD. In this work, we propose a novel deep-learning framework called Polar-Net. Our approach involves mapping OCTA images from Cartesian coordinates to polar coordinates, which allows for the use of approximate sector convolution and enables the implementation of the ETDRS grid-based regional analysis method commonly used in clinical practice. Furthermore, Polar-Net incorporates clinical prior information of each sector region into the training process, which further enhances its performance. Additionally, our framework adapts to acquire the importance of the corresponding retinal region, which helps researchers and clinicians understand the model’s decision-making process in detecting AD and assess its conformity to clinical observations. Through evaluations on private and public datasets, we have demonstrated that Polar-Net outperforms existing state-of-the-art methods and provides more valuable pathological evidence for the association between retinal vascular changes and AD. In addition, we also show that the two innovative modules introduced in our framework have a significant impact on improving overall performance. Copyright © 2023, The Authors. All rights reserved.

关键词： Optical tomography

Regularly Truncated M-estimators for Learning with Noisy labels

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Xia, Xiaobo Lu, Pengqian Gong, Chen Han, Bo Yu, Jun Yu, Jun Liu, Tongliang The Sydney AI Center School of Computer Science Faculty of Engineering The University of Sydney DarlingtonNSW2008 Australia The Australian AI Institute Faculty of Engineering and IT The University of Technology Sydney BroadwayNSW2007 Australia The PCA Lab The Key Laboratory of Intelligent Perception and Systems for High-Dimensional Information of Ministry of Education School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing210094 China The Department of Computing Hong Kong Polytechnic University Hong Kong The Department of Computer Science Hong Kong Baptist University Hong Kong The School of Computer Science and Technology Hangzhou Dianzi University Hangzhou310018 China The Department of Automation University of Science and Technology of China Hefei230026 China

The sample selection approach is very popular in learning with noisy labels. As deep networks "learn pattern first", prior methods built on sample selection share a similar training procedure: the small-loss examples can be regarded as clean examples and used for helping generalization, while the large-loss examples are treated as mislabeled ones and excluded from network parameter updates. However, such a procedure is arguably debatable from two folds: (a) it does not consider the bad influence of noisy labels in selected small-loss examples;(b) it does not make good use of the discarded large-loss examples, which may be clean or have meaningful information for generalization. In this paper, we propose regularly truncated M-estimators (RTME) to address the above two issues simultaneously. Specifically, RTME can alternately switch modes between truncated M-estimators and original M-estimators. The former can adaptively select small-losses examples without knowing the noise rate and reduce the side-effects of noisy labels in them. The latter makes the possibly clean examples but with large losses involved to help generalization. Theoretically, we demonstrate that our strategies are label-noise-tolerant. Empirically, comprehensive experimental results show that our method can outperform multiple baselines and is robust to broad noise types and levels. The implementation is available at https://***/xiaoboxia/RTM_LNL. Copyright © 2023, The Authors. All rights reserved.

关键词： Machine learning