检索结果-内蒙古大学图书馆

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Zhang, Weichao Zhang, Ruqing Guo, Jiafeng de Rijke, Maarten Fan, Yixing Cheng, Xueqi CAS Key Lab of Network Data Science and Technology ICT CAS Beijing China University of Chinese Academy of Sciences Beijing China Zhongguancun Laboratory Beijing China University of Amsterdam Amsterdam Netherlands

ISBN: (纸本)9798891761643

As the scale of training corpora for large language models (LLMs) grows, model developers become increasingly reluctant to disclose details on their data. This lack of transparency poses challenges to scientific evaluation and ethical deployment. Recently, pretraining data detection approaches, which infer whether a given text was part of an LLM's training data through black-box access, have been explored. The Min-K% Prob method, which has achieved state-of-the-art results, assumes that a non-training example tends to contain a few outlier words with low token probabilities. However, the effectiveness may be limited as it tends to misclassify non-training texts that contain many common words with high probabilities predicted by LLMs. To address this issue, we introduce a divergence-based calibration method, inspired by the divergence-from-randomness concept, to calibrate token probabilities for pretraining data detection. We compute the cross-entropy (i.e., the divergence) between the token probability distribution and the token frequency distribution to derive a detection score. We have developed a Chinese-language benchmark, PatentMIA, to assess the performance of detection approaches for LLMs on Chinese text. Experimental results on English-language benchmarks and PatentMIA demonstrate that our proposed method significantly outperforms existing methods. Our code and PatentMIA benchmark are available at https://***/zhang-wei-chao/DC-PDD. © 2024 Association for Computational Linguistics.

关键词： Calibration

来源：评论

学校读者我要写书评

暂无评论

Learning Dual-Domain Calibration and Distance-Driven Correlation Filter:A Probabilistic Perspective for UAV Tracking

引用

Computers, Materials & Continua 2023年第12期77卷 3741-3764页

作者： Taiyu Yan Yuxin Cao Guoxia Xu Xiaoran Zhao Hu Zhu Lizhen Deng Jiangsu Province Key Lab on Image Processing and Image Communication Nanjing University of Posts and TelecommunicationsNanjingChina School of Computer Science Qufu Normal UniversityQufuChina National Engineering Research Center of Communication and Network Technology Nanjing University of Posts and TelecommunicationsNanjingChina

Unmanned Aerial Vehicle(UAV)tracking has been possible because of the growth of intelligent information technology in smart cities,making it simple to gather data at any time by dynamically monitoring events,people,the environment,and other aspects in the *** traditional filter creates a model to address the boundary effect and time filter degradation issues in UAV tracking *** these methods ignore the loss of data integrity terms since they are overly dependent on numerous explicit previous regularization *** light of the aforementioned issues,this work suggests a dual-domain Jensen-Shannon divergence correlation filter(DJSCF)model address the probability-based distance measuring issue in the event of filter *** two-domain weighting matrix and JS divergence constraint are combined to lessen the impact of sample imbalance and *** new tracking models that are based on the perspectives of the actual probability filter distribution and observation probability filter distribution are proposed to translate the statistical distance in the online tracking model into response *** model is roughly transformed into a linear equality constraint issue in the iterative solution,which is then solved by the alternate direction multiplier method(ADMM).The usefulness and superiority of the suggested strategy have been shown by a vast number of experimental findings.

关键词： Dual-domain weighting distance measure correlation filter ADMM

来源：评论

学校读者我要写书评

暂无评论

CVCA: A Complex-Valued Classifiable Autoencoder for MmWave Massive MIMO Physical Layer Authentication

CVCA: A Complex-Valued Classifiable Autoencoder for MmWave M...

引用

2023 IEEE INFOCOM Conference on Computer Communications Workshops, INFOCOM WKSHPS 2023

作者： Zeng, Xinyuan Wang, Chao Wang, Cheng-Cai Li, Zan Xidian University State Key Laboratory of Integrated Service Network Xi'an710071 China Key Lab of Broadband Wireless Communication and Sensor Network Technology Nanjing210003 China School of Automation Science and Electrical Engineering Beihang University Beijing100191 China

ISBN: (纸本)9781665494274

For protecting millimeter wave (mmWave) communications from clone attacks, this paper employs the deep learning to propose a physical layer authentication (PLA) approach for detecting attackers and classifying multiple legitimate nodes simultaneously. Different from conventional upper-layer authentication mechanisms, the proposed PLA approach exploits the spatial and temporal characteristics of mmWave channels to extract the unique fingerprints for building a lightweight channel-based authentication method. However, the existing threshold-based PLA methods could not discriminate multiple nodes, and supervised learning based approaches have limited application due to the unavailability of attackers' channel state information (CSI) in practice. Besides, traditional real-valued deep neural networks cannot exploit the phase information of complex channels efficiently, which is unsuitable for designing the PLA scheme. Considering these, we propose a complex-valued classifiable autoencoder induced PLA scheme that includes a novel complex-valued long short-term memory (LSTM) module. Simulation results validate the superiority of our proposed PLA approach by comparing it with existing approaches and demonstrate that the detection probability of clone attacks positively correlates with antenna number. The classification performance is satisfactory even under the challenging experimental condition. © 2023 IEEE.

关键词： Authentication

来源：评论

学校读者我要写书评

暂无评论

Properties of focused Laguerre–Gaussian beam propagating in anisotropic ocean turbulence

引用

Chinese Physics B 2024年第2期33卷 386-393页

作者：王新光马洋斌袁邱杰陈伟王乐赵生妹 College of Science Nanjing University of Posts and TelecommunicationsNanjing 210003China Institute of Signal Processing and Transmission Nanjing University of Posts and TelecommunicationsNanjing 210003China Key Lab of Broadband Wireless Communication and Sensor Network Technology(Ministry of Education) Nanjing University of Posts and TelecommunicationsNanjing 210003China National Laboratory of Solid State Microstructures Nanjing UniversityNanjing 210093China

We analyze the properties of a focused Laguerre–Gaussian(LG)beam propagating through anisotropic ocean turbulence based on the Huygens–Fresnel *** the Rytov approximation theory,we derive the analytical formula of the channel capacity of the focused LG beam in the anisotropic ocean turbulence,and analyze the relationship between the capacity and the light source parameters as well as the turbulent ocean *** is found that the focusing mirror can greatly enhance the channel capacity of the system at the geometric focal plane in oceanic *** results also demonstrate that the communication link can obtain high channel capacity by adopting longer beam wavelength,greater initial beam waist radius,and larger number of transmission ***,the capacity of the system increases with the decrease of the mean squared temperature dissipation rate,temperature-salinity contribution ratio and turbulence outer scale factor,and with the increase of the kinetic energy dissipation rate per unit mass of fluid,turbulence inner scale factor and anisotropy *** to a Hankel–Bessel beam with diffraction-free characteristics and unfocused LG beam,the focused LG beam shows superior anti-turbulence interference properties,which provide a theoretical reference for research and development of underwater optical communication links using focused LG beams.

关键词： vortex beam orbital angular momentum focusing mirror anisotropic turbulence

来源：评论

学校读者我要写书评

暂无评论

A Contrastive Pre-training Approach to Discriminative Autoencoder for Dense Retrieval 22

A Contrastive Pre-training Approach to Discriminative Autoen...

引用

31st ACM International Conference on Information and Knowledge Management, CIKM 2022

作者： Ma, Xinyu Zhang, Ruqing Guo, Jiafeng Fan, Yixing Cheng, Xueqi Cas Key Lab of Network Data Science and Technology Ict Cas University of Chinese Academy of Sciences Beijing China

ISBN: (纸本)9781450392365

Dense retrieval (DR) has shown promising results in information retrieval. In essence, DR requires high-quality text representations to support effective search in the representation space. Recent studies have shown that pre-trained autoencoder-based language models with a weak decoder can provide high-quality text representations, boosting the effectiveness and few-shot ability of DR models. However, even a weak autoregressive decoder has the bypass effect on the encoder. More importantly, the discriminative ability of learned representations may be limited since each token is treated equally important in decoding the input texts. To address the above problems, in this paper, we propose a contrastive pre-training approach to learn a discriminative autoencoder with a lightweight multi-layer perception (MLP) decoder. The basic idea is to generate word distributions of input text in a non-autoregressive fashion and pull the word distributions of two masked versions of one text close while pushing away from others. We theoretically show that our contrastive strategy can suppress the common words and highlight the representative words in decoding, leading to discriminative representations. Empirical results show that our method can significantly outperform the state-of-the-art autoencoder-based language models and other pre-trained models for dense retrieval. © 2022 ACM.

关键词： Decoding

来源：评论

学校读者我要写书评

暂无评论

Track Irregularity Prediction Based on DWT-DLSTM Model

Track Irregularity Prediction Based on DWT-DLSTM Model

引用

2022 International Conference on networking and network Applications, NaNA 2022

作者： Meng, Haining Zhang, Jiawei Zheng, Yi Ji, Wenjiang Tong, Xinyu Hei, Xinhong School of Computer Science and Engineering Xi'an University of Technology Xi'an China Shaanxi Key Lab Network Computer and Security Technology Xi'an China

ISBN: (数字)9781665461313

ISBN: (纸本)9781665461313

The long-term operation of high-speed railway will lead to track irregularity that will cause random vibration of the track system and affect driving safety. The accurate prediction of track irregularity is of great significance to the quality of high-speed railway. In this paper, we proposed a DWT-DLSTM model to predict the track irregularity for high-speed railway. Firstly, the track irregularity time series data is denoised through the discrete wavelet transform (DWT). Then the deep long short-term memory (DLSTM) neural network is adopted to predict the denoised data. Finally, the experiment results show that the proposed DWT-DLSTM model outperforms other traditional models and obtain more accurate prediction results for track irregularity. © 2022 IEEE.

关键词： Discrete wavelet transforms

来源：评论

学校读者我要写书评

暂无评论

AMTrans:Auto-Correlation Multi-Head Attention Transformer for Infrared Spectral Deconvolution

引用

清华大学学报自然科学版（英文版） 2025年第3期30卷 1329-1341页

作者： Lei Gao Liyuan Cui Shuwen Chen Lizhen Deng Xiaokang Wang Xiaohong Yan Hu Zhu College of Electronic and Optical Engineering & College of Flexible Electronics(Future Technology) Nanjing University of Posts and TelecommunicationsNanjing 210003China Jiangsu Province Key Lab on Image Processing and Image Communication Nanjing University of Posts and TelecommunicationsNanjing 210003China National Engineering Research Center of Communication and Network Technology Nanjing University of Posts and TelecommunicationsNanjing 210003China School of Computer Science and Technology Hainan UniversityHaikou 570228China

Infrared spectroscopy analysis has found widespread applications in various fields due to advancements in technology and industry *** improve the quality and reliability of infrared spectroscopy signals,deconvolution is a crucial preprocessing *** by the transformer model,we propose an Auto-correlation Multi-head attention Transformer(AMTrans)for infrared spectrum sequence *** auto-correlation attention model improves the scaled dot-product attention in the *** utilizes attention mechanism for feature extraction and implements attention computation using the auto-correlation *** auto-correlation attention model is used to exploit the inherent sequence nature of spectral data and to effectively recovery spectra by capturing auto-correlation patterns in the *** proposed model is trained using supervised learning and demonstrates promising results in infrared spectroscopic *** comparing the experiments with other deconvolution techniques,the experimental results show that the method has excellent deconvolution performance and can effectively recover the texture details of the infrared spectrum.

关键词： spectroscopy spectral deconvolution transformer auto-correlation mechanism

来源：评论

学校读者我要写书评

暂无评论

On the MDS Condition of Generalized Expanded-Blaum-Roth Codes 18

On the MDS Condition of Generalized Expanded-Blaum-Roth Code...

引用

18th International Symposium on Problems of Redundancy in Information and Control Systems, REDUNDANCY 2023

作者： Hou, Hanxu Song, Linqi Dongguan University of Technology Dongguan Key Lab of Artificial Information Network School of Electrical Engineering & Intelligentization China City University of Hong Kong Department of Computer Science Hong Kong

ISBN: (纸本)9798350381528

Generalized Expanded-Blaum-Roth (GEBR) codes [You et al. 2020] are designed for large-scale distributed storage systems that have larger recoverability for single-symbol failures, multi-column failures and multi-row failures, compared with locally recoverable codes (LRC). GEBR codes encode an α × k information array into a pτ ×(k + r) array such that lines of slope i with 0 ≤ i ≤ r -1 have even parity and each column contains pτ-α local parity symbols, where p is an odd prime and k +r ≤ pτ. Necessary and sufficient conditions for GEBR codes to be (n, k) recoverable (i.e., any k out of n = k+r columns can retrieve all information symbols) are given in [Hou et al. 2023] for α = (p -1)τ. However, the (n, k) recoverable condition of GEBR codes is unknown when α © 2023 IEEE.

关键词： Multiprocessing systems

来源：评论

学校读者我要写书评

暂无评论

Triple-Fault-Tolerant Binary MDS Array Codes with Lower Encoding Complexity 18

Triple-Fault-Tolerant Binary MDS Array Codes with Lower Enco...

引用

18th International Symposium on Problems of Redundancy in Information and Control Systems, REDUNDANCY 2023

ISBN: (纸本)9798350381528

In this paper, we present a new construction of maximal distance separable (MDS) array codes with three parity columns. We present an efficient encoding algorithm for our new codes and show that our new codes have lower encoding complexity than all the existing MDS codes under the same parameters. The essential reasons of obtaining lower encoding complexity for our codes are of two-folds: (i) our codes are well-designed in a cyclic polynomial ring with only cyclic shifts and XORs being involved in the operations;(ii) our codes are designed based on the efficient schedule algorithm that can minimize the number of operations in the coding procedure. Moreover, our codes can support much more parameters compared with the existing MDS array codes, such as the EVENODD codes and STAR codes. We also present an efficient decoding method for the proposed codes. © 2023 IEEE.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

An enhanced genetic algorithm for computation task offloading in MEC scenario

引用

International Journal of Wireless and Mobile Computing 2023年第2期25卷 118-127页

作者： Zhao, Jiacheng Li, Wenzao Liu, Hantao Yu, Peizhen Li, Hanyun Wen, Zhan School of Communication Engineering Chengdu University of Information Technology Sichuan Chengdu China Chengdu University of Information Technology Sichuan Chengdu China Network and Data Security Key Lab of Sichuan Pro. University of Electronic Science and Technology of China Sichuan Chengdu China Educational Informationisation and Big Data Centre Education Department of Sichuan Province Sichuan Chengdu China

The explosive growth of Internet of Things (IoT) and 5G communication technologies has driven the increasing computing demands for wireless devices. Mobile edge computing in the 5G scenario is a promising solution for energy-efficient and low latency applications. However, due to limited bandwidth, the selection of appropriate computing tasks greatly affects the user experience and system performance. Under the wireless bandwidth constraint, the reasonable choice of offloading objects is an NP-hard problem. The genetic algorithm has a great ability to solve this problem, but the performance of the algorithm varies with different scenarios. This paper proposes a task offloading strategy based on an enhanced genetic algorithm for small-scale computing tasks with an ultra-dense terminal distribution. Numerical experiments show that the convergence speed and optimisation effect of the enhanced genetic algorithm are significantly improved compared to the conventional genetic algorithm. © 2023 Inderscience Enterprises Ltd.

关键词： Genetic algorithms

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：