检索结果-内蒙古大学图书馆

Proceedings of the 38th International Conference on Neural information processing Systems

作者： Zhuofan Wen Shangtong Gui Yang Feng Key Laboratory of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences and University of Chinese Academy of Sciences Beijing China Key Laboratory of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences and State Key Lab of Processors Institute of Computing Technology Chinese Academy of Sciences and University of Chinese Academy of Sciences Beijing China Key Laboratory of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences and Key Laboratory of Al Safety Chinese Academy of Sciences and University of Chinese Academy of Sciences Beijing China

ISBN: (纸本)9798331314385

Inference acceleration of large language models (LLMs) has been put forward in many application scenarios and speculative decoding has shown its advantage in addressing inference acceleration. Speculative decoding usually introduces a draft model to assist the base LLM where the draft model produces drafts and the base LLM verifies the draft for acceptance or rejection. In this framework, the final inference speed is decided by the decoding speed of the draft model and the acceptance rate of the draft provided by the draft model. Currently the widely used draft models usually generate draft tokens for the next several positions in a non-autoregressive way without considering the correlations between draft tokens. Therefore, it has a high decoding speed but an unsatisfactory acceptance rate. In this paper, we focus on how to improve the performance of the draft model and aim to accelerate inference via a high acceptance rate. To this end, we propose a CTC-based draft model which strengthens the correlations between draft tokens during the draft phase, thereby generating higher-quality draft candidate sequences. Experiment results show that compared to strong baselines, the proposed method can achieve a higher acceptance rate and hence a faster inference speed.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Research on Machine Reading Comprehension Model Based on Bidirectional Attention

Research on Machine Reading Comprehension Model Based on Bid...

引用

2023 IEEE International Conference on Sensors, Electronics and Computer Engineering, ICSECE 2023

作者： Jiang, Chenglin Ma, Ning Wan, Fucheng Key Laboratory of China's Ethnic Languages and Information Technology of Ministry of Education Northwest Minzu University Gansu Lanzhou730000 China Northwest Minzu University Key Laboratory of China's Ethnic Languages and Intelligent Processing of Gansu Province Gansu Lanzhou730000 China

ISBN: (纸本)9798350313734

Machine Reading Comprehension (MRC) is a significant subfield of Natural Language processing (NLP) and has become a hot research topic in recent years. It is an algorithm that allows a machine to understand natural language text and answer based on questions. With the rapid development of deep learning, MRC, especially attention mechanisms, has emerged as a primary technique in *** paper focuses on the development and principles of attention mechanisms, along with introducing three MRC models based on bidirectional attention. Bidirectional attention helps the models extract crucial information, enhancing accuracy and enabling broader applications in various NLP tasks. © 2023 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Two-Archive-Based Competitive Swarm Optimizer for Constrained Multiobjective Optimization 5

Two-Archive-Based Competitive Swarm Optimizer for Constraine...

引用

5th International Conference on Data-Driven Optimization of Complex Systems, DOCS 2023

作者： Li, Lianghao He, Cheng Lin, Jianqing Pan, Linqiang School of Artificial Intelligence and Automation Huazhong University of Science and Technology Key Laboratory of Image Information Processing and Intelligent Control of Education Ministry of China Wuhan China School of Electrical and Electronic Engineering Huazhong University of Science and Technology Wuhan China

ISBN: (纸本)9798350393521

Constrained multiobjective optimization problems (CMOPs) are prevalent in various real-world applications, presenting a formidable challenge to existing evolutionary algorithms when faced with intricate constraints. When solving CMOPs, a crucial concern is achieving a balance between convergence, diversity, and feasibility. To address these challenges, this paper proposes a two-archive-based constrained multiobjective competitive swarm optimizer. The algorithm preserves two collaborative and complementary archives to improve population convergence and feasibility. By implementing a competitive particle swarm mechanism, offspring solutions are generated by drawing upon solutions from both archives, thus capitalizing on their synergistic effect and exceptional information. An adaptive parameter is also used to adjust the bias in choosing the winner in the paired competition during the evolution. The experimental results demonstrate the effectiveness of the proposed algorithm in handling constrained multiobjective optimization problems. © 2023 IEEE.

关键词： Constraint handling

来源：评论

学校读者我要写书评

暂无评论

Research on patent recommendation method based on graph neural network 26

Research on patent recommendation method based on graph neur...

引用

26th International Conference on Computer Supported Cooperative Work in Design, CSCWD 2023

作者： Zhang, Enchao Liu, Weidong Liu, Shi Cao, Yan Institute of Scientific and Technical Information of China China College of Computer Science Inner Mongolia University China Natl. Loc. Jt. Eng. Research Center of Intelligent Information Processing Technology for Mongolian China Inner Mongolia Key Laboratory of Social Computing and Data Processing China Inner Mongolia Engineering Laboratory for Big Data Analysis Technology China

ISBN: (纸本)9798350331684

Intellectual property transactions have shown a strong growth momentum in recent years, but the patent transaction market has been plagued by the matching degree of consumers and sellers, resulting in frequent problems such as low patent transformation efficiency and poor transaction quality. This paper proposes a method of recommending patents to consumers by experts to improve the environment of patent transactions. Through the analysis of the past transaction information of the patent, the effective path information of the target is extracted. The graph neural network is used to describe the characteristics and semantics among experts, patents and consumers, and then capture the potential weight among them through the common attention mechanism, and then dynamically integrate them to predict the occurrence of recommendation behavior. The paper makes reasonable use of social information and expert information in the transaction, which significantly improves the rationality and accuracy of expert recommendation. © 2023 IEEE.

关键词： Patents and inventions

来源：评论

学校读者我要写书评

暂无评论

Multi-feature Chinese Semantic Role Labeling combined with global information

Multi-feature Chinese Semantic Role Labeling combined with g...

引用

2023 International Conference on Electronic information Engineering and Data processing, EIEDP 2023

作者： Ma, Ning Wang, Jiahao Zhu, Ao Key Laboratory of China's Ethnic Languages and Information Technology of Ministry of Education Northwest Minzu University Gansu Lanzhou730000 China Key Laboratory of China's Ethnic Languages and Intelligent Processing of Gansu Province Northwest Minzu University Gansu Lanzhou730000 China

ISBN: (纸本)9781510666160

Semantic role labeling serves as a crucial approach to obtaining semantic information and enabling shallow semantic analysis. Currently, the BiLSTM-CRF model is the primary method used for semantic role labeling. However, this model has many network parameters and is unable to effectively capture semantic information in long sentences. To address these issues, this paper proposes a CNN-BiLSTM-MaxPool-CRF fusion model for Chinese semantic role labeling. The model utilizes MaxPool to sample and extract the output of the BiLSTM network to optimize the network structure. Convolution kernels of differing sizes are utilized to capture local features of sentences, and these features are then combined through average pooling to form new feature vectors. These new features incorporate semantic information of the sentence context and are inputted into the model alongside multi-level linguistic feature groups such as part of speech and sentence phrase structure. Through multiple sets of experimental demonstrations, the method proposed in this paper has demonstrated significant improvements in the performance of the semantic role labeling model. © 2023 SPIE.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Versatile Teacher: A Class-aware Teacher-student Framework for Cross-domain Adaptation

arXiv

引用

arXiv 2024年

作者： Yang, Runou Tian, Tian Tian, Jinwen National Key Laboratory of Multispectral Information Intelligent Processing Technology School of Artificial Intelligence and Automation Huazhong University of Science and Technology China

Addressing the challenge of domain shift between datasets is vital in maintaining model performance. In the context of cross-domain object detection, the teacher-student framework, a widely-used semi-supervised model, has shown significant accuracy improvements. However, existing methods often overlook class differences, treating all classes equally, resulting in suboptimal results. Furthermore, the integration of instance-level alignment with a one-stage detector, essential due to the absence of a Region Proposal Network (RPN), remains unexplored in this framework. In response to these shortcomings, we introduce a novel teacher-student model named Versatile Teacher (VT). VT differs from previous works by considering class-specific detection difficulty and employing a two-step pseudo-label selection mechanism, referred to as Class-aware Pseudo-label Adaptive Selection (CAPS), to generate more reliable pseudo labels. These labels are leveraged as saliency matrices to guide the discriminator for targeted instance-level alignment. Our method demonstrates promising results on three benchmark datasets, and extends the alignment methods for widely-used one-stage detectors, presenting significant potential for practical applications. Code is available at https://***/RicardooYoung/VersatileTeacher. Copyright © 2024, The Authors. All rights reserved.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Heavy Rainfall Prediction Model Using Sample Entropy Derived from GNSS-PWV and PSO-SVM 14th

Heavy Rainfall Prediction Model Using Sample Entropy Derived...

引用

14th China Satellite Navigation Conference, CSNC 2024

作者： Wu, Fanming Li, Dengao Zhao, Jinhua Feng, Ran Shi, Danyang Zhang, Xinfang Zhao, Jumin College of Data Science Taiyuan University of Technology Taiyuan030024 China College of Information and Computer Taiyuan University of Technology Taiyuan030024 China Key Laboratory of Big Data Fusion Analysis and Application of Shanxi Province Taiyuan030024 China Intelligent Perception Engineering Technology Center of Shanxi Taiyuan030024 China Shanxi Province Engineering Technology Research Center of Spatial Information Network Taiyuan030024 China

ISBN: (纸本)9789819969272

There is a growing interest to use Global Navigation Satellite System (GNSS) inversed PWV for heavy rainfall prediction. When heavy rainfall occurs, it requires the atmosphere to contain sufficient water vapour and undergo strong upward motion. However, existing models using GNSS-PWV for heavy rainfall prediction have not accounted for the effect of the complex motion of water vapour. In this paper, an hourly heavy rainfall prediction model using sample entropy derived from GNSS-PWV and PSO-SVM is proposed. The sample entropy of GNSS-PWV is used to measure the complexity of the water vapour movement process before heavy rainfall occurs. Meanwhile, combining GNSS-PWV time domain features and co-located meteorological data, the occurrence of hourly heavy rainfall events is predicted by support vector machine optimized with particle swarm. To verify the validity and feasibility of the proposed algorithm, Hong Kong's HKSC station in Sham Shui Po is used for the model to train and test. The result shows that probability of detection, false alarm rate and critical success index of the proposed algorithm in this paper have significantly improved over other heavy rainfall prediction models. © 2024, Aerospace information Research Institute.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Open-Vocabulary Semantic Segmentation with Image Embedding Balancing

arXiv

引用

arXiv 2024年

作者： Shan, Xiangheng Wu, Dongyue Zhu, Guilin Shao, Yuanjie Sang, Nong Gao, Changxin National Key Laboratory of Multispectral Information Intelligent Processing Technology School of Artificial Intelligence and Automation Huazhong University of Science and Technology China

Open-vocabulary semantic segmentation is a challenging task, which requires the model to output semantic masks of an image beyond a close-set vocabulary. Although many efforts have been made to utilize powerful CLIP models to accomplish this task, they are still easily overfitting to training classes due to the natural gaps in semantic information between training and new classes. To overcome this challenge, we propose a novel framework for open-vocabulary semantic segmentation called EBSeg, incorporating an Adaptively Balanced Decoder (AdaB Decoder) and a Semantic Structure Consistency loss (SSC Loss). The AdaB Decoder is designed to generate different image embeddings for both training and new classes. Subsequently, these two types of embeddings are adaptively balanced to fully exploit their ability to recognize training classes and generalization ability for new classes. To learn a consistent semantic structure from CLIP, the SSC Loss aligns the inter-classes affinity in the image feature space with that in the text feature space of CLIP, thereby improving the generalization ability of our model. Furthermore, we employ a frozen SAM image encoder to complement the spatial information that CLIP features lack due to the low training image resolution and image-level supervision inherent in CLIP. Extensive experiments conducted across various benchmarks demonstrate that the proposed EBSeg outperforms the state-of-the-art methods. Our code and trained models will be here: https://***/slonetime/EBSeg. © 2024, CC BY.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

Prediction of Ionospheric TEC Based on BLS-LSTM-GRU Hybrid Model 14th

Prediction of Ionospheric TEC Based on BLS-LSTM-GRU Hybrid M...

引用

14th China Satellite Navigation Conference, CSNC 2024

作者： Li, Dengao Zhang, Xinfang Zhao, Jumin Wu, Fanming Feng, Ran Zhao, Jinhua Shi, Danyang College of Data Science Taiyuan University of Technology Taiyuan030024 China College of Information and Computer Taiyuan University of Technology Taiyuan030024 China Key Laboratory of Big Data Fusion Analysis and Application of Shanxi Province Taiyuan030024 China Intelligent Perception Engineering Technology Center of Shanxi Taiyuan030024 China Shanxi Province Engineering Technology Research Center of Spatial Information Network Taiyuan030024 China

ISBN: (纸本)9789819969272

In satellite communications, global navigation satellite systems(GNSS) and other important space activities, the value of total electron content (TEC) in the ionosphere directly affects the size of ionospheric delay. Achieving accurate prediction of ionospheric total electron content through modeling can effectively improve the reliability and accuracy of GNSS positioning. Considering the high prediction accuracy of long short-term memory neural network (LSTM) and the fast training speed of gated recurrent unit network (GRU), but it is easy to fall into the local optimum, while the width learning system(BLS) can effectively avoid this drawback. Therefore, this paper proposes a model for predicting the total electron content of the ionosphere, which consists of a broad learning system, an LSTM network and a GRU networks. Six parameters of TEC, geomagnetic index Dst, Kp, Ap, solar activity index F10.7, and hour (HD) are selected as input features from the observed data to predict the variation of the total ionospheric electron content. The results generate prediction maps of the total ionospheric electron content and are compared with empirical models and conventional neural network models at different times and solar activity conditions. The experimental results show that the root mean square error (RMSE) of the prediction results of this BLS-LSTM-GRU model is decreased by 23.07%, 17.48% and 9.46% compared with the IRI-2016, NeQuick and CNN-LSTM prediction models, respectively, with higher prediction accuracy. © 2024, Aerospace information Research Institute.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

REPAIR: Rank Correlation and Noisy Pair Half-replacing with Memory for Noisy Correspondence

arXiv

引用

arXiv 2024年

作者： Zheng, Ruochen Hong, Jiahao Gao, Changxin Sang, Nong National Key Laboratory of Multispectral Information Intelligent Processing Technology School of Artificial Intelligence and Automation Huazhong University of Science and Technology China

The presence of noise in acquired data invariably leads to performance degradation in cross-modal matching. Unfortunately, obtaining precise annotations in the multimodal field is expensive, which has prompted some methods to tackle the mismatched data pair issue in cross-modal matching contexts, termed as noisy correspondence. However, most of these existing noisy correspondence methods exhibit the following limitations: a) the problem of self-reinforcing error accumulation, and b) improper handling of noisy data pair. To tackle the two problems, we propose a generalized framework termed as Rank corrElation and noisy Pair hAlf-replacing wIth memoRy (REPAIR), which benefits from maintaining a memory bank for features of matched pairs. Specifically, we calculate the distances between the features in the memory bank and those of the target pair for each respective modality, and use the rank correlation of these two sets of distances to estimate the soft correspondence label of the target pair. Estimating soft correspondence based on memory bank features rather than using a similarity network can avoid the accumulation of errors due to incorrect network identifications. For pairs that are completely mismatched, REPAIR searches the memory bank for the most matching feature to replace one feature of one modality, instead of using the original pair directly or merely discarding the mismatched pair. We conduct experiments on three cross-modal datasets, i.e., Flickr30K, MS-COCO, and CC152K, proving the effectiveness and robustness of our REPAIR on synthetic and real-world noise. © 2024, CC BY.

关键词： Data handling

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：