检索结果-内蒙古大学图书馆

Unveiling factuality and injecting knowledge for LLMs via reinforcement learning and data proportion

Science China(Information Sciences) 2024年第10期67卷 389-390页

作者： Wenjun KE Ziyu SHANG Zhizhao LUO Peng WANG Yikai GUO Qi LIU Yuxuan CHEN School of Computer Science and Engineering Southeast University Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications(Southeast University) Beijing Institute of Technology Zhuhai Beijing Institute of Computer Technology and Application

Large language models(LLMs) have demonstrated remarkable effectiveness across various natural language processing(NLP) tasks, as evidenced by recent studies [1, 2]. However, these models often produce responses that conflict with reality due to the unreliable distribution of facts within their training data, which is particularly critical for applications requiring high credibility and accuracy [3].

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Novel Collision Detection Protocol for Wireless Full-duplex Networks 20

A Novel Collision Detection Protocol for Wireless Full-duple...

引用

20th IEEE International Conference on Networking, Sensing and Control, ICNSC 2023

作者： Zhao, Qinglin Feng, Li Zhou, MengChu Yu, Haisheng Sun, Yi Macau University of Science and Technology School of Computer Science and Engineering China New Jersey Institute of Technology Department of Electrical and Computer Engineering Newark United States Macau University of Science and Technology International Institute of Next Generation Internet China Chinese Academy of Sciences Institute of Computing Technology Beijing China

ISBN: (纸本)9798350369502

Conventional wireless networks are half-duplex and most of them use contention-based protocols. These protocols usually adopt a principle of contention with collision avoidance and infer a collision occurrence very late from the absence of an acknowledgement after data transmission, thus causing low network performance. Wireless full-duplex enables simultaneous transmission (TX) and reception (RX) on the same channel. By exploiting this functionality, this work proposes a novel design that enables contention with collision detection (CCD) to improve the network performance. With this design, in contention, a node exploits its TX antenna to transmit a signal for channel contention, while exploiting its RX antenna to sense if other nodes are transmitting too. By checking the status of TX and RX antennas, the node can detect the contention collision before data transmission and hence obtains an opportunity to avoid data collision effectively. This work then develops a theoretical model to analyze the collision probability of this design. Extensive simulation results verify the effectiveness of our proposed design in improving wireless network performance. © 2023 IEEE.

关键词： Network performance

来源：评论

学校读者我要写书评

暂无评论

Multi-Scale Cascaded with Cross-Attention Network-Based Deformation Vector Field Estimation for Motion-Compensated 4D-CBCT Reconstruction

IEEE Transactions on Computational Imaging

引用

IEEE Transactions on Computational Imaging 2025年 11卷 717-731页

作者： Yuan, Peng Lyu, Fei Gao, Zhiqiang Yang, Chunfeng Hu, Dianlin Zhu, Jian Wu, Zhan Lyu, Tianling Zhao, Wei Dong, Jianmin Chen, Yang Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications Ministry of Education Southeast University Nanjing210096 China Laboratory of Image Science and Technology Southeast University Nanjing210096 China Department of Computer Science Hong Kong Baptist University Hong Kong Hong Kong Academy for Advanced Interdisciplinary Science and Technology Zhejiang University of Technology Hangzhou China School of Physics Beihang University Beijing China Beihang Hangzhou Innovation Institute Zhejiang Hangzhou China School of Shandong Cancer Hospital and Institute Shandong First Medical University Shandong Academy of Medical Sciences Jinan Shandong China School of Information Engineering Xizang Minzu University Xianyang712082 China Jiangsu Provincial Joint International Research Laboratory of Medical Information Processing The Laboratory of Image Science and Technology The School of Computer Science and Engineering Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications Southeast University Ministry of Education Nanjing210096 China

Four-Dimensional Cone Beam Computed Tomography (4D-CBCT) imaging technology offers enhanced image quality and spatial resolution for intraoperative guidance, facilitating real-time tracking of tumor position changes during radiotherapy. However, this is still a task of great challenges due to insufficient projections at each respiratory phase after phase-sorting, and the image phases reconstructed directly from phase-sorted data are discrete and discontinuous. To generate high-quality 4D-CBCT deformation vector fields (DVFs), this paper leverages the preoperative static prior image to guide intraoperative dynamic sparse-view reconstruction images for reducing anatomical structure differences, ultimately achieving continuous and dynamic 4D-CBCT imaging. In this paper, we propose a Multi-scale Cascaded residual deformable vector field estimation framework based on Cross-attention in Motioncompensated 4D-CBCT reconstruction (MCCM), which combines Multi-Scale Cascaded residual registration network (MSCNet), Cross-Attention Enhanced feature Fusion (CAEF) module and Structure-Enhanced Motion-Compensated (SEMC) module: 1) the MCCM employs a multi-scale cascaded residual network strategy, merging multi-receptive fields and multi-resolution feature maps for large-scale internal changes. 2) the CAEF is embedded into MSC-Net to facilitate effective communication and learning between features and promote the flow of information. 3) the SEMC is developed to reduce artifacts after intraoperative CBCT sparse-view reconstruction, restore global lung motion changes and local details, and enhance structural information through image fusion. The proposed method has been evaluated using simulated and clinical datasets and has shown promising results by comparative experiment. Our approach exhibits significant improvements across various evaluation metrics. © 2024 IEEE.

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

Towards Synchronous Memorizability and Generalizability with Site-Modulated Diffusion Replay for Cross-Site Continual Segmentation

引用

IEEE Transactions on Medical Imaging 2025年 PP卷 PP页

作者： Xu, Dunyuan Wang, Xi Li, Jinpeng Zhang, Jingyang Heng, Pheng-Ann Department of Computer Science and Engineering Chinese University of Hong Kong Hong Kong Institute of Medical Intelligence and XR Chinese University of Hong Kong Hong Kong Hong Kong University of Science and Technology Department of Computer Science and Engineering China Southeast University School of Computer Science and Engineering Key Laboratory of New Generation Artiﬁcial Intelligence Technology and Interdisciplinary Applications China

The ability to learn sequentially from different data sites is crucial for a deep network in solving practical medical image diagnosis problems due to privacy restrictions and storage limitations. However, adapting to the incoming site leads to catastrophic forgetting on past sites and decreases generalizability on unseen sites. Existing Continual Learning (CL) and Domain Generalization (DG) methods have been proposed to solve these two challenges respectively, but none of them can address both simultaneously. Recognizing this limitation, this paper proposes a novel training paradigm, learning towards Synchronous Memorizability and Generalizability (SMG-Learning). To achieve this, we create the orientational gradient alignment to ensure memorizability on previous sites, and arbitrary gradient alignment to enhance generalizability on unseen sites. This approach is called Parallel Gradient Alignment (PGA). Furthermore, we approximate the PGA as dual meta-objectives using the first-order Taylor expansion to reduce the computational cost of aligning gradients. Considering that performing gradient alignments, especially for previous sites, is not feasible due to privacy constraints, we design a Site-Modulated Diffusion (SMD) model to generate images with site-specific learnable prompts, replaying images have similar data distributions as previous sites. We evaluate our method on two medical image segmentation tasks, where data from different sites arrive sequentially. Experimental results show that our method efficiently enhances both memorizability and generalizability better than other state-of-the-art methods, delivering satisfactory performance across all sites. © 1982-2012 IEEE.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

Towards Continual Knowledge Graph Embedding via Incremental Distillation

arXiv

引用

arXiv 2024年

作者： Liu, Jiajun Ke, Wenjun Wang, Peng Shang, Ziyu Gao, Jinhua Li, Guozheng Ji, Ke Liu, Yanhe School of Computer Science and Engineering Southeast University China Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications [Southeast University Ministry of Education China Institute of Computing Technology Chinese Academy of Sciences China

Traditional knowledge graph embedding (KGE) methods typically require preserving the entire knowledge graph (KG) with significant training costs when new knowledge emerges. To address this issue, the continual knowledge graph embedding (CKGE) task has been proposed to train the KGE model by learning emerging knowledge efficiently while simultaneously preserving decent old knowledge. However, the explicit graph structure in KGs, which is critical for the above goal, has been heavily ignored by existing CKGE methods. On the one hand, existing methods usually learn new triples in a random order, destroying the inner structure of new KGs. On the other hand, old triples are preserved with equal priority, failing to alleviate catastrophic forgetting effectively. In this paper, we propose a competitive method for CKGE based on incremental distillation (IncDE), which considers the full use of the explicit graph structure in KGs. First, to optimize the learning order, we introduce a hierarchical strategy, ranking new triples for layer-by-layer learning. By employing the inter- and intra-hierarchical orders together, new triples are grouped into layers based on the graph structure features. Secondly, to preserve the old knowledge effectively, we devise a novel incremental distillation mechanism, which facilitates the seamless transfer of entity representations from the previous layer to the next one, promoting old knowledge preservation. Finally, we adopt a two-stage training paradigm to avoid the over-corruption of old knowledge influenced by under-trained new knowledge. Experimental results demonstrate the superiority of IncDE over state-of-the-art baselines. Notably, the incremental distillation mechanism contributes to improvements of 0.2%-6.5% in the mean reciprocal rank (MRR) score. More exploratory experiments validate the effectiveness of IncDE in proficiently learning new knowledge while preserving old knowledge across all time steps. Copyright © 2024, The Authors. A

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Beyond Overfitting: Doubly Adaptive Dropout for Generalizable AU Detection

arXiv

引用

arXiv 2025年

作者： Li, Yong Ren, Yi Niu, Xuesong Ding, Yi Wei, Xiu-Shen Guan, Cuntai School of Computer Science and Engineering The Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications Southeast University Nanjing210096 China School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing210094 China Beijing Institute for General Artificial Intelligence Beijing China School of Computer Science and Engineering Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications Southeast University Nanjing210096 China School of Computer Science and Engineering Nanyang Technological University 50 Nanyang Avenue 639798 Singapore

Facial Action Units (AUs) are essential for conveying psychological states and emotional expressions. While automatic AU detection systems leveraging deep learning have progressed, they often overfit to specific datasets and individual features, limiting their cross-domain applicability. To overcome these limitations, we propose a doubly adaptive dropout approach for cross-domain AU detection, which enhances the robustness of convolutional feature maps and spatial tokens against domain shifts. This approach includes a Channel Drop Unit (CD-Unit) and a Token Drop Unit (TD-Unit), which work together to reduce domain-specific noise at both the channel and token levels. The CD-Unit preserves domain-agnostic local patterns in feature maps, while the TD-Unit helps the model identify AU relationships generalizable across domains. An auxiliary domain classifier, integrated at each layer, guides the selective omission of domain-sensitive features. To prevent excessive feature dropout, a progressive training strategy is used, allowing for selective exclusion of sensitive features at any model layer. Our method consistently outperforms existing techniques in cross-domain AU detection, as demonstrated by extensive experimental evaluations. Visualizations of attention maps also highlight clear and meaningful patterns related to both individual and combined AUs, further validating the approach’s effectiveness. © 2025, CC BY-NC-SA.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Beyond Overfitting: Doubly Adaptive Dropout for Generalizable AU Detection

引用

IEEE Transactions on Affective Computing 2025年

作者： Li, Yong Ren, Yi Niu, Xuesong Ding, Yi Wei, Xiu-Shen Guan, Cuntai School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing210094 China Beijing Institute for General Artificial Intelligence Beijing China School of Computer Science and Engineering Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications Southeast University China School of Computer Science and Engineering Nanyang Technological University 50 Nanyang Avenue Singapore639798 Singapore

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

A Novel Collision Detection Protocol for Wireless Full-duplex Networks

A Novel Collision Detection Protocol for Wireless Full-duple...

引用

IEEE International Conference on Networking, Sensing and Control

作者： Qinglin Zhao Li Feng MengChu Zhou Haisheng Yu Yi Sun School of Computer Science and Engineering Macau University of Science and Technology Macau China Department of Electrical and Computer Engineering New Jersey Institute of Technology Newark USA International Institute of Next Generation Internet Macau University of Science and Technology Macau China Institute of Computing Technology Chinese Academy of Sciences Beijing China

关键词：

来源：评论

学校读者我要写书评

暂无评论

Robust 3D Face Alignment with Multi-Path Neural Architecture Search

Robust 3D Face Alignment with Multi-Path Neural Architecture...

引用

IEEE International Conference on Multimedia and Expo (ICME)

作者： Zhichao Jiang Hongsong Wang Xi Teng Baopu Li Institute of Deep Learning (IDL) Baidu Beijing China Department of Computer Science and Engineering Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications Southeast University Nanjing China Computer Vision Technology Institution Baidu Beijing China Baidu Research Baidu Sunnyvale USA

ISBN: (数字)9798350390155

ISBN: (纸本)9798350390162

3D face alignment is a very challenging and fundamental problem in computer vision. Existing deep learning-based methods manually design different networks to regress either parameters of a 3D face model or 3D positions of face vertices. However, designing such networks relies on expert knowledge, and these methods often struggle to produce consistent results across various face poses. To address this limitation, we employ Neural Architecture Search (NAS) to automatically discover the optimal architecture for 3D face alignment. We propose a novel Multi-path One-shot Neural Architecture Search (MONAS) framework that leverages multi-scale features and contextual information to enhance face alignment across various poses. The MONAS comprises two key algorithms: Multi-path Networks Unbiased Sampling Based Training and Simulated Annealing based Multi-path One-shot Search. Experimental results on three popular benchmarks demonstrate the superior performance of the MONAS for both sparse alignment and dense alignment.

关键词： Training Learning systems Solid modeling computer vision Three-dimensional displays Neural networks Simulated annealing Network architecture Search problems Faces

来源：评论

学校读者我要写书评

暂无评论

Boosting LLMS with Ontology-Aware Prompt for Ner Data Augmentation

Boosting LLMS with Ontology-Aware Prompt for Ner Data Augmen...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Zhizhao Luo Youchen Wang Wenjun Ke Rui Qi Yikai Guo Peng Wang Beijing Institute of Computer Technology and Application Beijing China School of Computer Science and Engineering Southeast University Nanjing China Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University) Ministry of Education Nanjing China China Life Property & Casualty Insurance Company Limited Changsha China

Named Entity Recognition (NER) data augmentation (DA) aims to improve the performance and generalization capabilities of NER models by generating scalable training data. The key challenge lies in ensuring the generated samples maintain contextual diversity while preserving label consistency. However, existing dominant methods fail to simultaneously satisfy both criteria. Inspired by the extensive generative capabilities of large language models (LLMs), we propose ANGEL, a frAmework integrating the oNtoloGy structure and instructivE prompting within LLMs. Specifically, the hierarchical ontology structure guides prompt ranking, while instructive prompting enhances LLMs’ mastery of domain knowledge, empowering synthetic sample generation and annotation. Experiments show ANGEL surpasses state-of-the-art (SOTA) baselines, conferring absolute F1 increases of 2.86% and 0.93% on two benchmark datasets, respectively.

关键词：

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：