检索结果-内蒙古大学图书馆

Multi-core based parallel computing technique for content-based image retrieval

Journal of Shanghai University(English Edition) 2010年第1期14卷 55-59页

作者：陈文浩方昱春姚继锋张武 School of Computer Engineering and Science Shanghai University Shanghai Supercomputer Center High Performance Computing Center Shanghai University

In this paper, we propose a parallel computing technique for content-based image retrieval （CBIR） system. This technique is mainly used for single node with multi-core processor, which is different from those based on cluster or network computing architecture. Due to its specific applications （such as medical image processing） and the harsh terms of hardware resource requirement, the CBIR system has been prevented from being widely used. With the increasing volume of the image database, the widespread use of multi-core processors, and the requirement of the retrieval accuracy and speed, we need to achieve a retrieval strategy which is based on multi-core processor to make the retrieval faster and more convenient than before. Experimental results demonstrate that this parallel architecture can significantly improve the performance of retrieval system. In addition, we also propose an efficient parallel technique with the combinations of the cluster and the multi-core techniques, which is supposed to gear to the new trend of the cloud computing.

关键词： content-based image retrieval （CBIR） parallel computing shared-memory feature extraction similarity comparison

来源：评论

学校读者我要写书评

暂无评论

MedBLIP: A Multimodal Method of Medical Question-Answering Based on Fine-Tuning Large Language Model

引用

computerized Medical Imaging and Graphics 2025年

作者： Lejun Gong Jiaming Yang Shengyuan Han Yimu Ji Jiangsu Key Lab of Big Data Security & Intelligent Processing School of computer science Nanjing University of Posts and Telecommunications Nanjing210023 China Institute of High Performance Computing and Bigdata Nanjing University of Posts and Telecommunications Nanjing 210023 China

Medical visual question answering is crucial for effectively interpreting medical images containing clinically relevant information. This study proposes a method called MedBLIP (Medical Treatment Bootstrapping Language-Image Pretraining) to tackle visual language generation tasks related to chest X-rays in the medical field. The method combine an image encoder with a large-scale language model, and effectively generates medical question-answering text through a strategy of freezing the image encoder based on the BLIP-2 model. Firstly, chest X-ray images are preprocessed, and an image sample generation algorithm is used to enhance the text data of doctor-patient question-answering, thereby increasing data diversity. Then, a multi-layer convolutional image feature extractor is introduced to better capture the feature representation of medical images. During the fine-tuning process of the large language generation model, a new unfreezing strategy is proposed, which is to unfreeze different proportions of the weights of the fully connected layer to adapt to the data in the medical field. The image feature extractor is responsible for extracting key features from images, providing the model with rich visual information, while the text feature extractor accurately captures the essential requirements of the user's question. Through their synergistic interaction, the model can more effectively integrate medical images and user inquiries, thereby generating more accurate and relevant output content. The experimental results show that unfreezing 31.25% of the weights of the fully connected layer can significantly improve the performance of the model, with ROUGE-L reaching 66.12%, and providing a more accurate and efficient answer generation solution for the medical field. The method of this study has potential applications in the field of medical language generation tasks. Although the proposed model cannot yet fully replace human radiologists, it plays an indispensable role

关键词：

来源：评论

学校读者我要写书评

暂无评论

An algorithm design and implementation of Connect6

An algorithm design and implementation of Connect6

引用

2012 International Conference on Electrical Insulating Materials and Electrical engineering, EIMEE 2012

作者： Shi, Weiyi Shen, Yunfu School of Computer Engineering and Science China High Performance Computing Center Shanghai University Shanghai 200072 China

ISBN: (纸本)9783037854518

The new Connect6 game was launched and received the widespread attention in recent years. For the problems existing in pattern judgment way and in 6-8 windows in some literatures, the algorithm of Connect6 is studied by considering the attack and defensive characteristics of the Connect6. By the new 6-8 window approach proposed here, the evaluation function which determines the evaluation value is designed and implemented by calculating the changing quantity of the assessed values of the chessboard before and after the falling a stone. Based on this, an optimal Connect6 program is designed and implemented based on the PVS algorithm and the transposition table optimization, history heuristic optimization, null-move pruning optimization as well as iterative deepening optimization. The pruning rate of nodes in game tree comes to more than 98%. The actual playing of the chess indicates that the program has the strong chess strength. © (2012) Trans Tech Publications, Switzerland.

关键词： Optimization

来源：评论

学校读者我要写书评

暂无评论

Using spatial correlations in distributed video coding

Using spatial correlations in distributed video coding

引用

International Symposium on Artificial Intelligence and Signal Processing

作者： Fathy, Mahmood Haqqani, Mohammad High Performance Computing Lab. Computer Engineering Department Iran University of Science and Technology Tehran Iran

ISBN: (纸本)9781424498345

Distributed video coding (DVC) is a new paradigm of coding that makes very interest to itself in the past decade. It's usually based on temporal correlations between successive frames which are called key frames. But this approach failed for high object motion in a scene and need very complex decoder to cover it. In this paper we propose a novel technique to use spatial correlation instead of temporal correlation in a frame by dividing a single frame to two sub frames which one of them is used as a key frame for generate side information at the decoder. Simulation results show a significant improvement on rate distortion parameter in DVC schema. © 2011 IEEE.

关键词： Decoding

来源：评论

学校读者我要写书评

暂无评论

Optimized Reduce Communication performance with the Tree Topology 20

Optimized Reduce Communication Performance with the Tree Top...

引用

4th high performance computing and Cluster Technologies Conference, HPCCT 2020 and the 3rd International Conference on Big Data and Artificial Intelligence, BDAI 2020

作者： Zhao, Tianhai Wang, Yunlan Wang, Xu Northwestern Polytechnical University School of Computer Science and Engineering Center for High Performance Computing Xi'an Shaanxi China

ISBN: (纸本)9781450375603

Communication plays an important role in MPI applications, and reduce operations are heavily used part of MPI. In this paper, we propose a k-nomial tree topology and a hierarchy tree topology to optimize the Reduce operation in MPI. The k-nomial tree can effectively decrease the communication steps and is suitable for lots of processes. Compared with the binomial tree algorithm in small and medium size messages, the Reduce operation performed by the k-nomial tree can improve communication performance by 46%. Hierarchy trees can dynamically group processes at run time to take advantage of high bandwidth to communicate as much as possible within nodes. The test results show that compared with the binomial tree algorithm, the performance of the hierarchy tree algorithm is stable. For Reduce operation, we can get a 30% performance improvement. © 2020 ACM.

关键词： Trees (mathematics)

来源：评论

学校读者我要写书评

暂无评论

Two-Level Scheduling Technology for Heterogeneous Clusters Using Analytical Hierarchy Processes 6

Two-Level Scheduling Technology for Heterogeneous Clusters U...

引用

6th IEEE International Conference on computer and Communication Systems, ICCCS 2021

作者： Zhao, Tianhai Gu, Jianhua Zhang, Xiaoyuan Northwestern Polytechnical University School of Computer Science and Engineering Center for High Performance Computing Xi'an Shanxi China

ISBN: (纸本)9780738126043

The cluster system presents multi-level and complex heterogeneous features with GPUs computing. Job scheduling technology determines computing resource utilization and throughput of the cluster system. On the basis of heterogeneous cluster system, the two-level scheduling technology is proposed in this paper. In the first level, a job priority scheduling algorithm based on user value is designed at the cluster level using analytical hierarchy processes. The second level of scheduling is to act on the computing nodes that contain GPUs, and design a scheduling algorithm based on GPU sharing policy that manages GPU jobs in a more fine-grained manner. Through experiments, it shows that the job priority scheduling algorithm based on user value can ensure that jobs with high user value are executed first, and fairness is guaranteed. The scheduling algorithm based on GPU sharing technology can significantly improve the utilization of GPU and video memory, reducing the average waiting time of jobs by 55.54% and the average turnaround time of jobs by 28.83%. © 2021 IEEE.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Combining RTT and ECN for RoCEv2 Protocol 20

Combining RTT and ECN for RoCEv2 Protocol

引用

4th high performance computing and Cluster Technologies Conference, HPCCT 2020 and the 3rd International Conference on Big Data and Artificial Intelligence, BDAI 2020

作者： Wang, Yunlan Lan, Ming Zhao, Tianhai Gao, Zhaokui Xie, Min School of Computer Science and Engineering Center for High Performance Computing Northwestern Polytechnical University Xi'an Shaanxi China

ISBN: (纸本)9781450375603

Remote direct memory access (RDMA) has the advantages of direct user-level access to HW, asynchronous communication, etc. RoCEv2 protocol enables RDMA technology to be used in large-scale data centers over Ethernet. It provides high throughput and low latency for data center, but it faces the problem of network congestion caused by RoCEv2 data flow. Efficient congestion control strategy plays an important role in the performance of RoCEv2 based data center network. At present, the network congestion control algorithms for RoCEv2 protocol are mainly based on ECN marked DCQCN and RTT based TIMELY. ECN and RTT have different characteristics and advantages. ECN is more effective to prevent packet losses;whereas RTT controls end-to-end queuing delay better. In this paper, a new algorithm RTT_DCQCN is presented for congestion control which is based on DCQCN. In RTT_DCQCN, the CNP generation mechanism and rate adjustment strategy are optimized;and the scheme of introducing RTT signal into DCQCN is proposed, in which the congestion degree is judged by using ECN and RTT together. The RoCEv2 protocol is implemented on ns-3 network simulation platform. The simulation experiment on ns-3 proved that RTT_DCQCN is valid for the aim of improving stability, fairness and throughput of the RoCEv2 data flows. © 2020 ACM.

关键词： Simulation platform

来源：评论

学校读者我要写书评

暂无评论

performance evaluation of container-based high performance computing ecosystem using OpenPOWER 32nd

Performance evaluation of container-based high performance c...

引用

32nd International Conference on high performance computing, ISC high performance 2017

作者： Kuity, Animesh Peddoju, Sateesh Kumar High Performance Computing Lab Department of Computer Science and Engineering Indian Institute of Technology Roorkee247667 India

ISBN: (纸本)9783319676296

Container-based high performance computing (HPC) has started gaining popularity due to its almost negligible performance penalty compared to the BareMetal hardware. Although HPC hardware architectures and programming models are continuously evolving, the platform models are suffering from the HPC community’s awareness. Power-awareness, hardware and application-aware co-design along with security related concerns have attracted the most in recent time to empower the platform models. Otherside, the OpenPOWER ecosystems have stepped into our life to fulfill the thirst of exploiting the last drop of a performance benefit from our invested system. It has Power8 compliant processor with a larger cache, big fat instructions and data path accompanied by the popular coherent accelerator. In this paper, we have proposed a container-based HPC ecosystem established using OpenPOWER machine. The performance of the designed and developed ecosystem is evaluated stressing on different subcomponents of the system such as processor, memory, IO, and interconnect. Finally, the results are compared with the performance of the equivalent environments made of virtual machines and BareMetal hardware. © Springer International Publishing AG 2017.

关键词： Ecosystems

来源：评论

学校读者我要写书评

暂无评论

performance prediction based on hierarchy parallel features captured in multi-processing system 09

Performance prediction based on hierarchy parallel features ...

引用

18th ACM International Symposium on high performance Distributed computing, HPDC 09, Co-located with the 2009 International Symposium on high performance Distributed computing Conference, HPDC'09

作者： Jiaxin, Li Feng, Shi Ning, Deng Qi, Zuo High Performance Embedded Computing Lab. School of Computer Science and Technology Beijing Institute of Technology China

ISBN: (纸本)9781605585871

As the computing ability of high performance computers are improved by increasing the number of computing elements, how to utilize the available computing resources becomes an important issue. Different strategies to solve an problem based on a multi-processing system can bring about distinct performance. In this paper, we propose a method to predict the performance of parallel applications. The method describes the parallel features of the multi-processing systems in a hierarchy way, and evaluates solutions based on the description. In this way, programmers can find the better solution of an application before real programming.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

User-preferences based service selection model in large-scale services with various QoS properties

引用

Journal of Shanghai University(English Edition) 2011年第5期15卷 420-425页

作者：刘慧许东许德震雷州张龙 School of Computer Engineering and Science Shanghai University High Performance Computing Center Shanghai University Lenovo Mobile Communication Technology Limited

With the rapid growth of service scale, there are many services with the same functional properties but different non-flmctional properties on the Internet. There have been some global optimizing service selection algorithms for service selection. However, most of those approaches cannot fully reflect users＇ preferences or are not fully suitable for large-scale services selection. In this paper, an ant colony optimization （ACO） algorithm for the model of global optimizing service selection with various quality of srevice （QoS） properties is employed, and a user-preference based large-scale service selection algorithm is proposed. This algorithm aims at optimizing user-preferred QoS properties and selecting services that meet all user-defined QoS thresholds. Experiment results prove that this algorithm is very efficient in this regard.

关键词： service selection quilty of service （QoS） ant colony optimization （ACO）

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：