检索结果-内蒙古大学图书馆

22nd IEEE International Conference on Communication technology, ICCT 2022

作者： Chen, Shangshang Zhao, Yaqian Li, Ruyang Li, Xuelei Zhao, Jian Liu, Kai National Supercomputing Center in Zhengzhou Zhengzhou University Zhengzhou China Inspur Electronic Information Industry Co. Ltd Jinan China Electronic Information Industry Co. Ltd Beijing China Electronic Information Industry Co. Ltd State Key Laboratory of High-end Server & Storage Technology Beijing China

ISBN: (纸本)9781665470674

Electronic voting (e-voting) is more economical and efficient than traditional voting, however, it also faces the challenges of privacy and security. Existing e-voting schemes take advantage of homomorphic encryption to solve the privacy and security issues for {0,1} type e-voting applications. Unfortunately, existing schemes cannot meet the requirements in (real number) score-type e-voting applications, e.g., in competitive competitions or satisfaction surveys. This paper focuses on solving the privacy and security issues for score-type e-voting applications based on cryptographic technology. In order to solve such issues, we propose an electronic scoring scheme based on the CKKS(Cheon-Kim-Kim-Song) homomorphic encryption algorithm. It supports floating-point computing for exploiting homomorphic encryption in scoring and evaluating procedures. In addition, the security analysis demonstrates that the proposal achieves data privacy and security protection. Furthermore, we design an electronic scoring experiment based on the CKKS scheme in the library of SEAL. The experimental results indicate that it only takes 0.42 seconds to complete the scoring evaluation for 5000 people. Finally, the comparison results show that the proposal is more efficient, flexible, and practical. © 2022 IEEE.

关键词： Privacy-preserving techniques

来源：评论

学校读者我要写书评

暂无评论

A Clustering-Anonymity Approach for Trajectory Data Publishing Considering both Distance and Direction

引用

Journal of Computing and Information technology 2022年第1期29卷

作者： Jiang, Huo-wen Hu, Ke-kun College of Mathematics and Computer Science Jiangxi Science and Technology Normal University Nanchang China State Key Laboratory of High-end Server and Storage Technology Inspur Group Co. Ltd Jinan China

Trajectory data contains rich spatio-temporal information of moving objects. Directly publishing it for mining and analysis will result in severe privacy disclosure problems. Most existing clustering-anonymity methods cluster trajectories according to either distance or direction-based similarities, leading to a high information loss. To bridge this gap, in this paper, we present a clustering-anonymity approach considering both these two types of similarities. As trajectories may not be synchronized, we first design a trajectory synchronization algorithm to synchronize them. Then, two similarity metrics between trajectories are quantitatively defined, followed by a comprehensive one. Furthermore, a clustering-anonymity algorithm for trajectory data publishing with privacy-preserving is proposed. It groups trajectories into clusters according to the comprehensive similarity metric. These clusters are finally anonymized. Experimental results show that our algorithm is effective in preserving privacy with low information loss. ACM CCS (2012) Classification: Security and privacy → Database and storage security → Data anonymization and sanitization Security and privacy → Human and societal aspects of security and privacy → Privacy protections © 2021. Journal of Computing and Information *** Rights Reserved.

关键词： Trajectories

来源：评论

学校读者我要写书评

暂无评论

Post Training Mixed-Precision Quantization Based on key Layers Selection 16th

Post Training Mixed-Precision Quantization Based on Key Laye...

引用

Workshops held at the 16th European Conference on Computer Vision, ECCV 2020

作者： Liang, Lingyan State Key Laboratory of High-End Server and Storage Technology Inspur Group Company Limited Beijing China

ISBN: (纸本)9783030682378

Model quantization has been extensively used to compress and accelerate deep neural network inference. Because post-training quantization methods are simple to use, they have gained considerable attention. However, when the model is quantized below 8-bits, significant accuracy degradation will be involved. This paper seeks to address this problem by building mixed-precision inference networks based on key activation layers selection. In post training quantization process, key activation layers are quantized by 8-bit precision, and non-key activation layers are quantized by 4-bit precision. The experimental results indicate an impressive promotion with our method. Relative to ResNet-50(W8A8) and VGG-16(W8A8), our proposed method can accelerate inference with lower power consumption and a little accuracy loss. © 2020, Springer Nature Switzerland AG.

关键词： Chemical activation

来源：评论

学校读者我要写书评

暂无评论

Distribution-restrained Softmax Loss for the Model Robustness

Distribution-restrained Softmax Loss for the Model Robustnes...

引用

2023 IJCAI Joint Workshop on Artificial Intelligence Safety and Safe Reinforcement Learning, AISafety-SafeRL 2023

作者： Wang, Hao Li, Chen Jiang, Jinzhe Zhang, Xin Zhao, Yaqian Gong, Weifeng School of Computer and Artificial Intelligence Zhengzhou University 100 Science Rd. Zhengzhou China Inspur Electronic Information Industry Co. Ltd 1036 Langchao Rd. Jinan China State Key Laboratory of High-End Server & Storage Technology 1036 Langchao Rd. Jinan China Shandong Hailiang Information Technology Institutes 1768 Xinli St. Jinan China Zhengzhou Yunhai Information Technology Co. Ltd 278 Xinyi Rd. Zhengzhou China

Recently, the issue of robustness in deep learning models has garnered considerable attention, prompting the development of diverse methods aimed at enhancing model robustness. These approaches encompass adversarial training, architectural modifications, the design of novel loss functions, as well as certified defenses, among others. Despite these efforts, a comprehensive understanding of the underlying principles governing robustness against attacks remains elusive, and related research in this area is still not sufficient. Here, we have identified a significant factor that affects the robustness of models: the distribution characteristics of softmax values for non-real label samples. We found that the results after an attack are highly correlated with the distribution characteristics. Leveraging this observation, we introduce a novel loss function that effectively mitigates the diversity in softmax distribution. Extensive experimental evaluations demonstrate that our proposed method significantly enhances model robustness without significant time consumption. These findings are poised to contribute valuable insights to the realm of AI safety. © 2023 CEUR-WS. All rights reserved.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Towards interpreting computer vision based on transformation invariant optimization

arXiv

引用

arXiv 2021年

作者： Li, Chen Jiang, Jinzhe Zhang, Xin Zhao, Yaqian Jiang, Dongdong Zhang, Tonghuan Li, RenGang State Key Laboratory of High-end Server & Storage Technology Beijing100085 China Shandong Hailiang Information Technology Institutes Jinan250014 China State Key Laboratory of High-End Server & Storage Technology Jinan250014 China

Interpreting how does deep neural networks (DNNs) make predictions is a vital field in artificial intelligence, which hinders wide applications of DNNs. Visualization of learned representations helps we humans understand the vision of DNNs. In this work, visualized images that can activate the neural network to the target classes are generated by back-propagation method. Here, rotation and scaling operations are applied to introduce the transformation invariance in the image generating process, which we find a significant improvement on visualization effect. Finally, we show some cases that such method can help us to gain insight into neural *** Codes 68T45 Copyright © 2021, The Authors. All rights reserved.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

UAV Flight Attitude Acquisition Algorithm Based on Multi-Camera 6

UAV Flight Attitude Acquisition Algorithm Based on Multi-Cam...

引用

6th International Conference on Signal and Image Processing, ICSIP 2021

作者： Yuehang, Sun Gewen, Kang Kehan, Zhang Kaiya, Liu Chunmeng, Zhao Shuli, Lou University of Electronic Science and Technology of China School of Aeronautics and Astronautics Chengdu China University of Electronic Science and Technology of China School of Foreign Lanhuages Chengdu China Yantai University State Key Laboratory of High-end Server and Storage Technology Yantai China

ISBN: (纸本)9780738133737

In this paper, an algorithm of UAV flight attitude acquisition based on multi-camera is proposed. Compared with the previous work, this method can realize the recognition and attitude capture of external flying objects and small targets. An algorithm is constructed to capture the flight attitude of UAV by building a camera interface database, and the whole interface pixel database is obtained by regularly calibrating the pitch angle and yaw angle data of a small number of pixels. The whole interface pixel database is obtained by Gaussian dynamic threshold filtering and interpolation. The attitude position of the UAV can be determined by the pitch and yaw line of sight rendezvous of the multi-camera, and the spatial position error caused by the error between the interpolated pitch yaw data and the real pitch yaw data can be expressed by the spatial position offset. In the flight test, when the maximum distance is 50m, the maximum absolute error of pitch angle is less than 1.3°, the average absolute error is less than 0.32°, the maximum absolute error of yaw angle is less than 1.5°, and the average absolute error is less than 0.4°. The spatial position offset of 96% of the data points is less than 0.25m, which has achieved excellent results. Flight test experiments show that this method is feasible, the model is universal and has a wide application prospect. © 2021 IEEE.

关键词： Interpolation

来源：评论

学校读者我要写书评

暂无评论

ROWE-tree: A Read-Optimized and Write-Efficient B+-tree for Persistent Memory 22

ROWE-tree: A Read-Optimized and Write-Efficient B+-tree for ...

引用

Proceedings of the 51st International Conference on Parallel Processing

作者： Xiaomin Zou Fang Wang Dan Feng Tianjin Guan Nan Su Wuhan National Laboratory for Optoelectronics Huazhong University of Science and Technology China Wuhan National Laboratory for Optoelectronics Huazhong University of Science and Technology China and Shenzhen Huazhong University of Science and Technology Research Institute China State Key Laboratory of High-end Server & Storage Technology China and shandong massive information technology research institute China

ISBN: (纸本)9781450397339

Persistent memory (PM) exhibits a huge potential to provide B+-tree indexes with high performance, efficient persistence, and instant recovery. A large number of PM-optimized B+-tree indexes have been proposed, but most of them fail to provide high performance for both read and write operations because: (1) their designs of search optimization and insert improvement are often traded off against each other, and (2) they overlook the read/write interference problem of PM which incurs unpredictable performance degradation. In this paper, we propose ROWE-tree, a read-optimized and write-efficient B+-tree for PM. The designs of our ROWE-tree consist of three key points. First, we propose two techniques to make a good trade-off between write and read performance: self-verifying insertion, which reduces consistency overhead by using the key itself as a persist mark instead of additional metadata, and semi-sorted leaf nodes, which use append-only insertion to avoid the shifting overhead of sorting nodes but keep intra-cache-line items sorted to accelerate the lookup. Second, based on the observation that data accesses are highly skewed in real-world workloads, we build an in-DRAM cache of hot items to outsource accesses to hot items to DRAM. By doing so, we can alleviate the read/write interference of PM and significantly improve overall performance. Third, to cope with the dynamic changes of hot items, we exploit a lightweight mechanism to track such changes at run-time. Using Intel Optane DCPMM, our evaluations show that ROWE-tree obtains up to 3.86 × higher performance than the state-of-the-art PM B+-tree indexes under YCSB workloads.

关键词： hotspot awareness B+-tree read/write optimization persistent memory

来源：评论

学校读者我要写书评

暂无评论

A Silicon-Based MEMS Transducer with Enhanced Frame Rate for Dynamic Infrared Scene Projection

A Silicon-Based MEMS Transducer with Enhanced Frame Rate for...

引用

2021 IEEE Conference on Telecommunications, Optics and Computer Science, TOCS 2021

作者： Zhou, Lang Li, Tuo Li, Zhuo Liu, Kai Man, Hongtao Li, Qifeng State Key Laboratory of High-end Server Storage Technology Jinan250013 China Electronic Information Industry Co Ltd Beijing100085 China Beijing Institute of Technology School of Optics and Photonics Beijing100081 China Beijing Key Lab for Precision Optoelectronic Measurement Instrument and Technology Beijing100081 China

ISBN: (纸本)9781665424981

This paper reports a novel MEMS transducer used for dynamic infrared scene projection. It can absorb the heat from an incident visible light scene and convert it into a dynamic infrared scene quickly. To improve heat dissipation, a silicon-based microcavity array was introduced as the substrate by a dry-etching fabrication process. Illuminated by a 50 Hz laser, the transient infrared response was obtained with the rising time and falling time of both 4 ms. The highest frame rate of 125 Hz was calculated. The simulation and measured results of transient thermal properties were well consistent. high resolution infrared images and an infrared video with the frequency of 120 Hz were successfully captured in long wave infrared band. These results proved the silicon-based MEMS transducer was a promising candidate in the dynamic infrared scene projection. © 2021 IEEE.

关键词： Silicon

来源：评论

学校读者我要写书评

暂无评论

Electrical Transport Properties of Few-Layer SnS2 Field-effect Transistors

Electrical Transport Properties of Few-Layer SnS2 Field-effe...

引用

2022 International Conference on Electrical, Electronics and Information Engineering, EEIE 2022

作者： Zhang, Lu Li, Tuo Yao, Peng Zou, Xiaofeng Ji, Chao Zhan, Yongzheng He, Jiaqi Zhang, Qing State Key Laboratory of High-end Server & Storage Technology Jinan250013 China Inspur Electronic Information Industry Co. Ltd. Jinan250013 China Shandong Yunhai Guochuang Cloud Computing Equipment Industry Innovation Co. Ltd. Jinan250013 China Beijing Institute of Computer Technology and Application Beijing100854 China College of Mathematics and Physics Beijing University of Chemical Technology Beijing100029 China

After the discovery of graphene in 2004, two dimensional (2D) materials have fascinated a lot of view due to the excellent properties. Nowadays, the research on 2D materials has spread to other graphene-like layer structured materials, especially transition metal dichalcogenides (TMDCs). Tin disulfide (SnS2) is a kind of TMDCs with a sizable bandgap. Here we introduce few-layer SnS2 field-effect transistors (FETs) fabricated using micromechanical exfoliation method. The FETs show n-type behavior, the on/off ratio exceeding 0.54×104 and the carrier mobility is 0.61 cm2V-1s-1. The electronic and optical characteristics of SnS2 flakes with a finite bandgap illustrate their potential applications in optoelectronics device. © Published under licence by IOP Publishing Ltd.

关键词： Field effect transistors

来源：评论

学校读者我要写书评

暂无评论

Recommender for Channel Pruning with Weakly Supervised Meta Learning 7

Recommender for Channel Pruning with Weakly Supervised Meta ...

引用

7th International Conference on Information Science and Control Engineering, ICISCE 2020

作者： Yin, Wenfeng Dong, Gang Zhao, Yaqian Li, Rengang Inspur Electronic Information Industry Co. Ltd. State Key Laboratory of High-end Server Storage Technology Jinan250101 China

ISBN: (纸本)9781728164069

The switchable neural network makes it feasible to deploy a deep neural network to a specific hardware platform by adjusting architectures but without retraining parameters. However, searching out the suitable architecture to adjust is time consuming and computation-tense. To accelerate searching architectures, this paper designs a recommender which takes one round to search and evaluate without repeating. In training stages, the recommender is alternately trained with a meta network which offers the recommender gradients as weakly supervision. In searching stages, the trained recommender extracts features from a specified layer's outputs, and generates suggested channels to prune for every layer. The recommender is deployed and evaluated on three common classification neural networks, including ResNet, MobileNet-v1 and MobileNet-v2. Experimental results validate that the proposed recommender is able to obtain compressed architectures whose accuracies surpass the switchable neural network's accuracy at a speed faster than evolutionary algorithm based searching methods. © 2020 IEEE.

关键词： Deep neural networks

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：