检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Cai, Yinqiong Bi, Keping Fan, Yixing Guo, Jiafeng Chen, Wei Cheng, Xueqi CAS Key Lab of Network Data Science and Technology ICT CAS University of Chinese Academy of Sciences Beijing China

First-stage retrieval is a critical task that aims to retrieve relevant document candidates from a large-scale collection. While existing retrieval models have achieved impressive performance, they are mostly studied on static data sets, ignoring that in the real-world, the data on the Web is continuously growing with potential distribution drift. Consequently, retrievers trained on static old data may not suit new-coming data well and inevitably produce sub-optimal results. In this work, we study lifelong learning for first-stage retrieval, especially focusing on the setting where the emerging documents are unlabeled since relevance annotation is expensive and may not keep up with data emergence. Under this setting, we aim to develop model updating with two goals: (1) to effectively adapt to the evolving distribution with the unlabeled new-coming data, and (2) to avoid re-inferring all embeddings of old documents to efficiently update the index each time the model is updated. We first formalize the task and then propose a novel Lifelong Learning method for the first-stage Retrieval, namely L2R. L2R adopts the typical memory mechanism for lifelong learning, and incorporates two crucial components: (1) selecting diverse support negatives for model training and memory updating for effective model adaptation, and (2) a ranking alignment objective to ensure the backward-compatibility of representations to save the cost of index rebuilding without hurting the model performance. For evaluation, we construct two new benchmarks from LoTTE and Multi-CPR © 2023, CC BY-NC-ND.

关键词： Learning systems

来源：评论

学校读者我要写书评

暂无评论

Speaker Recognition Model with Robust Front-end Processing Algorithm

Speaker Recognition Model with Robust Front-end Processing A...

引用

2022 International Conference on Neural networks, Information, and Communication Engineering, NNICE 2022

作者： Lian, Yingzi Pang, Jing College of Science Inner Mongolia University of Technology Hohhot China Inner Mongolia Key Laboratory of Statistical Analysis Theory for Life Data and Neural Network Modeling Inner Mongolia University of Technology Hohhot China

ISBN: (纸本)9781510655171

The end-to-end speaker recognition model has made a great breakthrough in the research field and practical application scenarios. However, in practical application, we often suffer the diversity of noise, and the interference of noise will affect the performance of speaker recognition and classification model, and the model usually degrades in unseen scenes with noise. In this paper, a speaker recognition model combined with front-end enhancement is proposed. The front-end enhancement model (DTLN or CRNN) is combined with the back-end speaker recognition model (Res2Net-GhostVLAD) to improve the robustness of the model against noisy scenes, and the generalization capability of the model is increased by the data augmentation method (SpecAugment). Our proposed method was trained on VoxCeleb and AISHELL datasets and tested on VoxCeleb datasets. The test results show that the proposed method significantly improves the performance of the speaker recognition model in noisy scenarios, and the relative improvement of different front-end models is 9% and 13%, respectively. © 2022 SPIE.

关键词： Speech recognition

来源：评论

学校读者我要写书评

暂无评论

A feature extraction with pose-auxiliary method for Person Reidentification 2

A feature extraction with pose-auxiliary method for Person R...

引用

2nd International Conference on Advanced Algorithms and Signal Image Processing, AASIP 2022

作者： Ma, ZePing Network and Data Security Key Laboratory of Sichuan Province University of Electronic Science and Technology of China Sichuan Chengdu610054 China School of Information and Software Engineering University of Electronic Science and Technology of China Sichuan Chengdu610054 China

ISBN: (纸本)9781510660571

Person re-identification is widely considered as a sub problem of image retrieval. It is a technology that uses computer vision technology to determine whether there is a specific pedestrian in an image or video, given a monitored pedestrian image, the pedestrian image under the cross device is retrieved. Many researches have been made to study how to match one people from two picture, but how to locating proper body parts has become a problem which could lead to misalign and wrongly divided sample. To solve this problem, we introduce a pose-auxiliary network to extract the position feature of pedestrian. We use a post estimation network to estimate the person's body location, then we decide whether it's a body part or a static obstacle by using a separated trained classifier. Then we compare these features from two picture and calculate their loss. Meanwhile, we use a graph-matching strategy to compare two pedestrians pose structure, which is also used into jugging the match problem. By using a multi-verification strategy, our method has some improvement than some previous study using Market-1501 and DukeMTMC-Reid datasets, but still remains room for improvement. © 2022 SPIE.

关键词： Image retrieval

来源：评论

学校读者我要写书评

暂无评论

Are Large Language Models More Honest in Their Probabilistic or Verbalized Confidence?

arXiv

引用

arXiv 2024年

作者： Ni, Shiyu Bi, Keping Yu, Lulu Guo, Jiafeng CAS Key Lab of Network Data Science and Technology ICT CAS China University of Chinese Academy of Sciences China

Large language models (LLMs) have been found to produce hallucinations when the question exceeds their internal knowledge boundaries. A reliable model should have a clear perception of its knowledge boundaries, providing correct answers within its scope and refusing to answer when it lacks knowledge. Existing research on LLMs’ perception of their knowledge boundaries typically uses either the probability of the generated tokens or the verbalized confidence as the model’s confidence in its response. However, these studies overlook the differences and connections between the two. In this paper, we conduct a comprehensive analysis and comparison of LLMs’ probabilistic perception and verbalized perception of their factual knowledge boundaries. First, we investigate the pros and cons of these two perceptions. Then, we study how they change under questions of varying frequencies. Finally, we measure the correlation between LLMs’ probabilistic confidence and verbalized confidence. Experimental results show that 1) LLMs’ probabilistic perception is generally more accurate than verbalized perception but requires an in-domain validation set to adjust the confidence threshold. 2) Both perceptions perform better on less frequent questions. 3) It is challenging for LLMs to accurately express their internal confidence in natural language. © 2024, CC BY.

关键词： Depth perception

来源：评论

学校读者我要写书评

暂无评论

A Novel 3D Medical Image Segmentation Model Using Improved SAM

A Novel 3D Medical Image Segmentation Model Using Improved S...

引用

2024 IEEE International Conference on Systems, Man, and Cybernetics, SMC 2024

作者： Kuang, Yuansen Ma, Xitong Zhao, Jing Wang, Guangchen Zeng, Yijie Liu, Song Key Laboratory of Computing Power Network and Information Security Jinan China Shandong Engineering Research Center of Big Data Applied Technology Faculty of Computer Science and Technology Jinan China Shandong Fundamental Research Center for Computer Science Shandong Provincial Key Laboratory of Computer Networks Jinan China

ISBN: (纸本)9781665410205

3D medical image segmentation is an essential task in the medical image field, which aims to segment organs or tumours into different labels. A number of issues exist with the current 3D medical image segmentation task: existing models cannot simultaneously obtain the space correlation and depth correlation of 3D slices;previous models suffer from local detail loss of positional embedding in 3D images;previous approaches often have blurring of boundaries in segmenting 3D images. To solve these shortcomings, we propose a 3D medical image segmentation model named TPM-SAM. In our model, we design a twinchannel image encoder to simultaneously capture the space correlation and depth correlation of 3D slices through a multi-head attention mechanism and improved adapters. Furthermore, we design a prompt encoding generator, which divides the volumetric image into small blocks and better captures the local detail information. In addition, we introduce a multi-layer aggregation decoder by employing U-Net with multi-level skip connection to solve the blurring of boundaries in processing 3D images. Finally, we experimented and evaluated our model on KiTS21 and LiTS17 datasets to compare with other baseline models. © 2024 IEEE.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

A Verifiable Privacy-Preserving Outsourced Prediction Scheme Based on Blockchain in Smart Healthcare

A Verifiable Privacy-Preserving Outsourced Prediction Scheme...

引用

2023 IEEE International Conference on E-Health networking, Application and Services, Healthcom 2023

作者： Li, Ta Tian, Youliang Xiong, Jinbo Wang, Linjie College of Computer Science and Technology Guizhou University State Key Laboratory of Public Big Data Guiyang550025 China College of Computer and Cyber Security Fujian Provincial Key Laboratory of Network Security and Cryptology Fujian Normal University Fuzhou350117 China School of Data Science Tongren University Tongren554300 China

ISBN: (纸本)9798350302301

The swift progression of the Internet of Things and the extensive integration of machine learning have spurred the growth of intelligent healthcare. Many intelligent healthcare devices, limited by their own computing and storage resources, require outsourcing data analysis tasks to cloud platforms for efficient and accurate results. Unfortunately, malicious cloud services lead to privacy breaches in outsourced data and untrustworthiness in learning models. To address these challenges, this paper proposes a verifiable privacy-preserving outsourced prediction scheme based on blockchain in smart healthcare (VPOL). Specifically, by incorporating blockchain technology into VPOL, we build a robust and scalable framework to prevent falsification of outsourced data and learning models in a decentralized and transparent manner. Then, we design a training committee approach to ensure the reliability of outsourced prediction and employ homomorphic encryption and commitment scheme to protect the privacy and integrity of the data. Finally, theoretical analysis proves the effectiveness and security of VPOL. Sufficient experiments demonstrate that VPOL achieves the approximate accuracy of the plaintext. © 2023 IEEE.

关键词： Blockchain

来源：评论

学校读者我要写书评

暂无评论

Evaluating Implicit Bias in Large Language Models by Attacking From a Psychometric Perspective

arXiv

引用

arXiv 2024年

作者： Wen, Yuchen Bi, Keping Chen, Wei Guo, Jiafeng Cheng, Xueqi CAS Key Laboratory of AI Safety ICT CAS Beijing China CAS Key Lab of Network Data Science and Technology ICT CAS Beijing China University of Chinese Academy of Sciences Beijing China

As large language models (LLMs) become an important way of information access, there have been increasing concerns that LLMs may intensify the spread of unethical content, including implicit bias that hurts certain populations without explicit harmful words. In this paper, we conduct a rigorous evaluation of LLMs’ implicit bias towards certain demographics by attacking them from a psychometric perspective to elicit agreements to biased viewpoints. Inspired by psychometric principles in cognitive and social psychology, we propose three attack approaches, i.e., Disguise, Deception, and Teaching. Incorporating the corresponding attack instructions, we built two benchmarks: (1) a bilingual dataset with biased statements covering four bias types (2.7K instances) for extensive comparative analysis, and (2) BUMBLE, a larger benchmark spanning nine common bias types (12.7K instances) for comprehensive evaluation. Extensive evaluation of popular commercial and open-source LLMs shows that our methods can elicit LLMs’ inner bias more effectively than competitive baselines. Our attack methodology and benchmarks offer an effective means of assessing the ethical risks of LLMs, driving progress toward greater accountability in their development. * © 2024, CC BY-NC-SA.

关键词： Risk assessment

来源：评论

学校读者我要写书评

暂无评论

Detection and mitigation of RFI in SBRS observation data

引用

Research in Astronomy and Astrophysics 2021年第8期21卷 111-123页

作者： Zhen-Ping Qiang Jun Cheng Zhen-Hong Shang Kai-Fan Ji Fei Dai Hui Liu College of Big Data and Intelligent Engineering Southwest Forestry UniversityKunming 650224China CAS Key Laboratory of Solar Activity National Astronomical ObservatoriesBeijing 100101China Faculty of Information Engineering and Automation Kunming University of Science and TechnologyKunming 650504China Yunnan Observatories Chinese Academy of SciencesKunming 650216China

In view of the inconsistency of channel gains and a large amount of interference noise in Solar Broadband Radio Spectrometer(SBRS) observation data,they will seriously affect the analysis of SBRS *** this paper,a method of Radio Frequency Interference(RFI) detection and mitigation for SBRS observation data is ***,the SBRS observation data are preprocessed,a part of the observation data was selected to calculate the mean and variance to achieve the normalization of the entire observation data,which can avoid the influence of strong noise on the normalization ***,we proposed an adaptive threshold RFI detection method based on fusion wavelet transform reconstruction and an RFI elimination method based on neighborhood weighted *** is worth mentioning that to detect RFI interference signals of different magnitudes,we adopted an iterative approach to the RFI detection and mitigation *** qualitative analysis of real observation data and quantitative analysis of simulated data,it is shown that the method proposed in this paper can effectively eliminate RFI in SBRS observation data,and improve the quality of observation data for further scientific analysis.

关键词： spectrographs:SBRS techniques:image processing and spectroscopic Sun:radio radiation

来源：评论

学校读者我要写书评

暂无评论

The 600 keV electron injections in the Earth's outer radiation belt:A statistical study

引用

Earth and Planetary Physics 2022年第2期6卷 149-160页

作者： ChaoLing Tang Xu Wang BinBin Ni ZhengPeng Su JiChun Zhang Shandong Key Laboratory of Optical Astronomy and Solar-Terrestrial Environment Institute of Space SciencesShandong UniversityWeihai Shandong 264209China School of Space Science and Physics Shandong UniversityWeihai Shandong 264209China Department of Space Physics School of Electronic InformationWuhan UniversityWuhan 430072China CAS Center for Excellence in Comparative Planetology Hefei 230026China CAS Key Laboratory of Geospace Environment Department of Geophysics and Planetary SciencesUniversity of Science and Technology of ChinaHefei 230026China Staunch Data Analytics Lab DurhamNew HampshireUSA

Relativistic electron injections are one of the mechanisms of relativistic(≥0.5 MeV) electron enhancements in the Earth’s outer radiation belt. In this study, we present a statistical observation of 600 keV electron injections in the outer radiation belt by using data from the Van Allen Probes. On the basis of the characteristics of different injections, 600 keV electron injections in the outer radiation belt were divided into pulsed electron injections and nonpulsed electron injections. The 600 keV electron injections were observed at 4.5 < L <6.4 under the geomagnetic conditions of 450 nT < AE < 1,450 nT. An L of ~4.5 is an inward limit for 600 keV electron injections. Before the electron injections, a flux negative L shell gradient for ≤0.6 MeV electrons or low electron fluxes in the injected region were observed. For600 keV electron injections at different L shells, the source populations from the Earth’s plasma sheet were different. For 600 keV electron injections at higher L shells, the source populations were higher energy electrons(~200 keV at X ~–9 R_(E)), whereas the source populations for 600 keV electron injections at lower L shells were lower energy electrons(~80 keV at X ~–9 R_(E)). These results are important to further our understanding of electron injections and rapid enhancements of 600 keV electrons in the Earth’s outer radiation belt.

关键词： electron injections relativistic electrons the Earth’s outer radiation belt plasma sheet Van Allen Probes

来源：评论

学校读者我要写书评

暂无评论

DFS-NER: Description Enhanced Few-shot NER via Prompt Learning and Meta-Learning

DFS-NER: Description Enhanced Few-shot NER via Prompt Learni...

引用

2022 IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent technology (WI-IAT)

作者： Huinan Huang Yuming Feng Xiaolong Jin Saiping Guan Jiafeng Guo School of Computer Science and Technology University of Chinese Academy of Sciences CAS Key Laboratory of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences

Named Entity Recognition (NER) is a very common task in many social good related domains. Recently, deep learning based NER has gradually matured, but still faces the scarcity problem of labeled data in specific domains. Therefore, researchers focus on few-shot NER to reduce the model’s data dependence and enhance the transferability of the model. However, existing works usually cannot adapt to new entity types and are prone to the so-called negative transfer problem. Therefore, in this paper we propose a type- Description-enhanced Few Shot NER model, called DFS-NER, which effectively integrates the prompt learning paradigm and the meta-learning framework. DFS-NER performs well under frozen pre-training model parameters through continuous templates. It realizes efficient source domain training and target domain parameter fine-tuning through the metalearning framework. We enhance the robustness of entity- type prototype representations by introducing word-word- level and word-type-level contrastive learning objectives and capsule networks as the induction module. Simultaneously, based on discrete prompt learning, a masked-language model learning objective guided by type description is proposed, which can well absorb the semantic information of entity types. Experiments on commonly used datasets, including, SNIPS, Few-NERD, and MIT Movie show that DFS-NER basically surpasses baseline models and achieves the state-of-the-art performance.

关键词： Deep learning Training Adaptation models Analytical models Semantics Prototypes Robustness

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：