检索结果-内蒙古大学图书馆

Paradigm Shift in Natural Language processing

Machine Intelligence Research 2022年第3期19卷 169-183页

作者： Tian-Xiang Sun Xiang-Yang Liu Xi-Peng Qiu Xuan-Jing Huang School of Computer Science Fudan UniversityShanghai 200438China Shanghai Key Laboratory of Intelligent Information Processing Fudan UniversityShanghai 200438China

In the era of deep learning, modeling for most natural language processing (NLP) tasks has converged into several mainstream paradigms. For example, we usually adopt the sequence labeling paradigm to solve a bundle of tasks such as POS-tagging, named entity recognition (NER), and chunking, and adopt the classification paradigm to solve tasks like sentiment analysis. With the rapid progress of pre-trained language models, recent years have witnessed a rising trend of paradigm shift, which is solving one NLP task in a new paradigm by reformulating the task. The paradigm shift has achieved great success on many tasks and is becoming a promising way to improve model performance. Moreover, some of these paradigms have shown great potential to unify a large number of NLP tasks, making it possible to build a single model to handle diverse tasks. In this paper, we review such phenomenon of paradigm shifts in recent years, highlighting several paradigms that have the potential to solve different NLP tasks.

关键词： Natural language processing pre-trained language models deep learning sequence-to-sequence paradigm shift

来源：评论

学校读者我要写书评

暂无评论

A transform domain communication method based on combinatorial basis function

A transform domain communication method based on combinatori...

引用

2023 International Conference on Communication Network and Machine Learning, CNML 2023

作者： Li, Beiming Xu, Jiahao Key Laboratory of Intelligent Information Processing Technology Ministry of Industry and Information Technology Harbin Engineering University Harbin150001 China

ISBN: (纸本)9798400716683

With the development of communication countermeasure technology, the covertness, anti-interception and anti-jamming of information transmission are more and more important. In this paper, on the basis of researching and analyzing the existing covert anti-jamming communication technology, we propose a new type of covert anti-jamming communication system, a kind of transform-domain covert communication based on combining basis functions. The communication system maps part of the information into a combination of frequency points, and the same information maps different combinations of frequency points, the combination of frequency points as the amplitude spectrum of the basis function, and then another part of the information through the transform domain modulation technology modulated into both strong concealment and strong anti-jamming ability, but also has a higher information transmission efficiency of the time domain transmission waveform, to provide a new idea for China's satellite covert communication. The simulation results verify the effectiveness of the method. © 2023 ACM.

关键词： Efficiency

来源：评论

学校读者我要写书评

暂无评论

NONMONOTONE LOCAL MINIMAX METHODS FOR FINDING MULTIPLE SADDLE POINTS

引用

Journal of Computational Mathematics 2024年第3期42卷 851-884页

作者： Wei Liu Ziqing Xie Wenfan Yi South China Research Center for Applied Mathematics and Interdisciplinary Studies South China Normal UniversityGuangzhou 510631China Key Laboratory of Computing and Stochastic Mathematics(Ministry of Education) Hunan Normal UniversityChangshaHunan 410081China Hunan Provincial Key Laboratory of Intelligent Information Processing and Applied Mathematics School of MathematicsHunan UniversityChangshaHunan 410082China

In this paper,by designing a normalized nonmonotone search strategy with the BarzilaiBorwein-type step-size,a novel local minimax method(LMM),which is a globally convergent iterative method,is proposed and analyzed to find multiple(unstable)saddle points of nonconvex functionals in Hilbert *** to traditional LMMs with monotone search strategies,this approach,which does not require strict decrease of the objective functional value at each iterative step,is observed to converge faster with less ***,based on a normalized iterative scheme coupled with a local peak selection that pulls the iterative point back onto the solution submanifold,by generalizing the Zhang-Hager(ZH)search strategy in the optimization theory to the LMM framework,a kind of normalized ZH-type nonmonotone step-size search strategy is introduced,and then a novel nonmonotone LMM is *** feasibility and global convergence results are rigorously carried out under the relaxation of the monotonicity for the functional at the iterative ***,in order to speed up the convergence of the nonmonotone LMM,a globally convergent Barzilai-Borwein-type LMM(GBBLMM)is presented by explicitly constructing the Barzilai-Borwein-type step-size as a trial step-size of the normalized ZH-type nonmonotone step-size search strategy in each ***,the GBBLMM algorithm is implemented to find multiple unstable solutions of two classes of semilinear elliptic boundary value problems with variational structures:one is the semilinear elliptic equations with the homogeneous Dirichlet boundary condition and another is the linear elliptic equations with semilinear Neumann boundary *** numerical results indicate that our approach is very effective and speeds up the LMMs significantly.

关键词： Multiple saddle points Local minimax method Barzilai-Borwein gradient method Normalized nonmonotone search strategy Global convergence

来源：评论

学校读者我要写书评

暂无评论

TOWARDS CODABLE WATERMARKING FOR INJECTING MULTI-BITS information TO LLMS 12

TOWARDS CODABLE WATERMARKING FOR INJECTING MULTI-BITS INFORM...

引用

12th International Conference on Learning Representations, ICLR 2024

作者： Wang, Lean Yang, Wenkai Chen, Deli Zhou, Hao Lin, Yankai Meng, Fandong Zhou, Jie Sun, Xu National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University China Gaoling School of Artificial Intelligence Renmin University of China China Pattern Recognition Center WeChat AI Tencent Inc. China DeepSeek AI China

As large language models (LLMs) generate texts with increasing fluency and realism, there is a growing need to identify the source of texts to prevent the abuse of LLMs. Text watermarking techniques have proven reliable in distinguishing whether a text is generated by LLMs by injecting hidden patterns. However, we argue that existing LLM watermarking methods are encoding-inefficient and cannot flexibly meet the diverse information encoding needs (such as encoding model version, generation time, user id, etc.). In this work, we conduct the first systematic study on the topic of Codable Text Watermarking for LLMs (CTWL) that allows text watermarks to carry multi-bit customizable information. First of all, we study the taxonomy of LLM watermarking technologies and give a mathematical formulation for CTWL. Additionally, we provide a comprehensive evaluation system for CTWL: (1) watermarking success rate, (2) robustness against various corruptions, (3) coding rate of payload information, (4) encoding and decoding efficiency, (5) impacts on the quality of the generated text. To meet the requirements of these non-Pareto-improving metrics, we follow the most prominent vocabulary partition-based watermarking direction, and devise an advanced CTWL method named Balance-Marking. The core idea of our method is to use a proxy language model to split the vocabulary into probability-balanced parts, thereby effectively maintaining the quality of the watermarked text. Extensive experimental results show that our method outperforms the baseline under comprehensive evaluation. Our code is available at https://***/lancopku/codable-watermarking-for-llm. © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

Speech emotion recognition based on CNN-MGU-attention

Speech emotion recognition based on CNN-MGU-attention

引用

2023 International Conference on Image, Signal processing, and pattern recognition, ISPP 2023

作者： Wang, Yanni Qinghai Normal University The State Key Laboratory of Tibetan Intelligent Information Processing and Application Xining810008 China

ISBN: (纸本)9781510666351

Speech is the most natural way of human communication, carrying a wealth of emotional information, and emotion is an important part of human cognitive process, so speech emotion recognition (SER) has become an important research direction in the field of pattern recognition. SER technology enables computers to understand human emotions, which is an indispensable step to realize artificial intelligence. However, there are some problems in the field of SER, such as the lack of effective emotion feature set and effective emotion recognition model. Starting from improving the recognition performance of SER model, this paper builds a new model CNN-MGU-Attention on the basis of deep neural network to improve the performance of speech emotion recognition system. The model is composed of convolutional neural network (CNN), minimum gated unit (MGU) and attention mechanism. Firstly, MGU can simplify the gate structure to the smallest possible state, and has far fewer parameters than long Short-Term memory (LSTM) and gated unit (GRU), which greatly reduces the training complexity and improves the training speed. Secondly, the attention mechanism was used to learn the correlation degree between the input and output sequences of the model, so as to pay more attention to the effective information. Finally, the Softmax layer was used to classify the sentiment. Experiments show that the CNN-MGU-Attention network model proposed in this study achieves an average recognition accuracy of 88.90%and 86.21% on the CASIA and EMODB databases respectively, which achieves better recognition performance compared with previous research results. © 2023 SPIE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

A Transformer-Assisted Cascade Learning Network for Choroidal Vessel Segmentation

引用

Journal of Computer Science & Technology 2024年第2期39卷 286-304页

作者：温阳吴依林毕磊石武祯刘潇骁许毓鹏许迅曹文明冯大淦 Guangdong Provincial Key Laboratory of Intelligent Information Processing School of Electronics and Information EngineeringShenzhen UniversityShenzhen 518060China Institute of Translational Medicine Shanghai Jiao Tong UniversityShanghai 200240China Department of Ophthalmology Shanghai General HospitalShanghai Jiao Tong University School of MedicineShanghai 200080China School of Information Technologies The University of SydneySydney 2006Australia

As a highly vascular eye part,the choroid is crucial in various eye disease ***,limited research has focused on the inner structure of the choroid due to the challenges in obtaining sufficient accurate label data,particularly for the choroidal ***,the existing direct choroidal vessel segmentation methods for the intelligent diagnosis of vascular assisted ophthalmic diseases are still unsatisfactory due to noise data,while the synergistic segmentation methods compromise vessel segmentation performance for the choroid layer segmentation *** cascaded structures grapple with error propagation during *** address these challenges,we propose a cascade learning segmentation method for the inner vessel structures of the choroid in this ***,we propose TransformerAssisted Cascade Learning Network(TACLNet)for choroidal vessel segmentation,which comprises a two-stage training strategy:pre-training for choroid layer segmentation and joint training for choroid layer and choroidal vessel *** also enhance the skip connection structures by introducing a multi-scale subtraction connection module designated as MSC,capturing differential and detailed information ***,we implement an auxiliary Transformer branch named ATB to integrate global features into the segmentation *** results exhibit that our method achieves the state-of-the-art performance for choroidal vessel ***,we further validate the significant superiority of the proposed method for retinal fluid segmentation in optical coherence tomography(OCT)scans on a publicly available *** these fully prove that our TACLNet contributes to the advancement of choroidal vessel segmentation and is of great significance for ophthalmic research and clinical application.

关键词： choroidal vessel segmentation optical coherence tomography(OCT) Transformer-assisted cascade learning retinal fluid segmentation multi-scale feature extraction

来源：评论

学校读者我要写书评

暂无评论

Dynamic Student Classification with Forgetting Mechanisms on Memory Networks for Knowledge Tracing 5

Dynamic Student Classification with Forgetting Mechanisms on...

引用

5th International Conference on intelligent Computing and Human-Computer Interaction, ICHCI 2024

作者： Huang, Guimin Qin, Xiangqian School of Computer Science and Information Security Guilin University of Electronic Technology Guangxi Key Laboratory of Image and Graphic Intelligent Processing Guilin China

ISBN: (纸本)9798350368284

Knowledge Tracing (KT), a technique for modeling students' knowledge levels and predicting their future question-answering performance based on their historical answer data, is one of the key research areas to strengthen the ability of personalized education. In recent years, memory networks have received more and more attention and application in the field of KT, however, the current KT model based on memory networks ignores the effect of students' learning sequence on the level of forgetting, and fails to model the forgetting behavior of students in the process of learning by using the characteristics of time intervals in the interaction data. Therefore, in this paper, we propose a dynamic student classification with forgetting mechanisms on memory networks model (DSCFMN), which enhance the existing model by dynamically classifying students similar to their learning abilities at specified time intervals, incorporating forgetting factors and introducing a weight decay strategy. Experiments show that our model performs well on online education datasets, and the proposed model achieves better prediction results than existing knowledge tracking methods. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

引用

5th International Conference on Big Data and Artificial Intelligence and Software Engineering, ICBASE 2024

作者： Zhou, Ya Chen, Chen Huang, Guimin Guilin University of Electronic Technology School of Computer Science and Information Security Guangxi Key Laboratory of Image and Graphic Intelligent Processing Guilin China

ISBN: (纸本)9798331506612

In the research methods of sentence similarity, sentence similarity is often calculated from the semantic aspect, while the influence of syntactic structure is ignored. We propose an enhanced knowledge language representation model (ExtKBRCNN) based on CNN and Bi-GRU, which effectively uses the fine-grained word relations in the knowledge base to evaluate semantic similarity and models the relationship between knowledge structure and text structure. In order to make full use of the syntactic information of the sentence, we also propose a dependency tree kernel-based method (Dep-SIF), which combines syntactic information and semantic features to evaluate syntactic similarity. Finally, we propose a comprehensive model that integrates semantic and syntactic information to comprehensively evaluate sentence similarity. Experimental results show that the accuracy of the model on the MRPC dataset is 77.63% and the F1 value is 83.90%. © 2024 IEEE.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Underwater Image Restoration Based on Improved SwinIR Model 13

Underwater Image Restoration Based on Improved SwinIR Model

引用

13th International Conference on intelligent Control and information processing, ICICIP 2025

作者： Lu, Yanke Li, Fang Luo, Xiaonan Guangxi Key Laboratory of Image and Graphic Intelligent Processing Guilin University of Electronic Technology Guilin541004 China School of Computer Science and Information Security Guilin University of Electronic Technology Guilin541004 China Guilin541004 China Guangxi Key Laboratory of Image and Graphic Intelligent Processing Guilin541004 China

ISBN: (纸本)9798331516147

Image restoration is a classic foundational visual task, aimed at recovering damaged images, such as those affected by compression, blurring, or noise, to high-definition clarity. Although current image enhancement techniques largely rely on convolutional neural networks, the efficient capabilities of Transformers demonstrated in complex visual tasks have not yet been extensively applied in the field of image restoration. This paper presents an optimized image restoration strategy based on the SwinIR model, which is applied to the task of underwater image restoration. In the shallow feature extraction phase, deep separable convolution technology is employed;while in the deep feature extraction phase, a Spatial Gated Feedforward Network (SGFN) is utilized. The experimental data suggests that the novel approach surpasses conventional techniques in effectiveness and has effectively minimized both the model's parameter quantity and computational demands. © 2025 IEEE.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Grain condition inspection: one improved grey wolf algorithm for nodes deployment optimization combining with path planning of multi-inspection robots

引用

Soft Computing 2024年第20期28卷 11971-11986页

作者： Zhu, Chunhua Zhang, Jilong Yang, Jing Key Laboratory of Grain Information Processing and Control Ministry of Education Henan University of Technology Henan Zhengzhou450001 China Henan Key Laboratory of Grain Photoelectric Detection and Control Henan University of Technology Henan Zhengzhou450001 China Henan Engineering Laboratory of Grain Condition Intelligent Detection and Application Henan University of Technology Henan Zhengzhou450001 China

Grain inspection robots can detect abnormal information such as grain insects and impurities on the grain surface by deploying detection nodes, it has a significant for the safety of grain storage. Generally, the grey wolf optimization (GWO) algorithm has significant advantages such as fewer control parameters and easy implementation, is one of the commonly employed approaches aiming to solve the problem in many fields. However, the traditional GWO has the shortcomings such as poor population initialization diversity, slow convergence speed and easily fall into local optimal values, it is difficult to solve deployment of inspection nodes problem in the grain surface. In this study, an improved Latin Hypercube grey wolf optimization (LH-GWO) is presented to overcome the disadvantages of the traditional GWO algorithm in solving inspection node deployment problem. The LH-GWO algorithm contains two novel mechanisms which are the Latin Hypercube Sampling population initialization and an improved nonlinear convergence factor. Moreover, a grain surface boundary inspection nodes deployment strategy is proposed to optimize the location of nodes at the boundary. For verifying the effectiveness of the proposed LH-GWO, comparing with six commonly used algorithms based on benchmark functions experiment and coverage rate experiment. The experimental results demonstrate the merits of LH-GWO in terms of convergence speed, optimization accuracy and area coverage. Finally, the optimal inspection nodes are divided into three inspection regional areas by k-means algorithm clustering, and in each inspection regional area, the optimal inspection path of each robot is determined by solving the travel salesman problem (TSP). © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.

关键词： Motion planning

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：