检索结果-内蒙古大学图书馆

HyFit: Hybrid Fine-Tuning With Diverse Sampling for Abstractive Summarization

IEEE Transactions on Big Data 2025年第3期11卷 1054-1065页

作者： Zhao, Shu Cheng, Yuanfang Zhang, Yanping Chen, Jie Duan, Zhen Sun, Yang Wang, Xinyuan Anhui University Ministry of Education Key Laboratory of Intelligent Computing and Signal Processing Hefei230601 China Anhui University School of Computer Science and Technology Hefei230601 China Anhui University Information Materials and Intelligent Sensing Laboratory of Anhui Province Hefei230601 China

Abstractive summarization has made significant progress in recent years, which aims to generate a concise and coherent summary that contains the most important facts from the source document. Current fine-tuning approaches based on pre-training models typically rely on autoregressive and maximum likelihood estimation, which may result in inconsistent historical distributions generated during the training and inference stages, i.e., exposure bias problem. To alleviate this problem, we propose a hybrid fine-tuning model(HyFit), which combines contrastive learning and reinforcement learning in a diverse sampling space. Firstly, we introduce reparameterization and probability-based sampling methods to generate a set of summary candidates called candidates bank, which improves the diversity and quality of the decoding sampling space and incorporates the potential for uncertainty. Secondly, hybrid fine-tuning with sampled candidates bank, upweighting confident summaries and downweighting unconfident ones. Experiments demonstrate that HyFit significantly outperforms the state-of-the-art models on SAMSum and DialogSum. HyFit also shows good performance on low-resource summarization, on DialogSum dataset, using only approximate 8% of the examples exceed the performance of the base model trained on all examples. © 2015 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Dark area recognition algorithm for SAR images based on superpixel statistical features

Dark area recognition algorithm for SAR images based on supe...

引用

2024 Advanced Fiber Laser Conference, AFL 2024

作者： Liu, Peng Liu, Yanyang Li, Xuemao Dong, Xuezhi Xiao, Qian Li, Zhenfang National Key Laboratory of Radar Signal Processing Xidian University Xi'an710071 China Shanghai Institute of Satellite Engineering Shanghai200090 China Aerospace Stellar Technology Co. Ltd. Beijing100095 China

ISBN: (数字)9781510688889

ISBN: (纸本)9781510688872

The dark areas in SAR images are primarily low backscatter regions dominated by noise signals, characterized by significant texture, low echo energy, and large phase interference errors in the echo signals. Accurately identifying these dark regions is crucial for high-precision SAR image target detection and classification, SAR cross-radiometric calibration of satellite constellations, SAR image filtering, and the generation of digital surface models based on SAR technology. This paper proposes a dark region recognition algorithm for SAR images based on superpixel statistical features(SSF). The algorithm constructs a bimodal distribution curve using the statistical features of superpixels and searches for the region with the minimum amplitude proportion within the bimodal distribution, which serves as a constraint to ensure that the OTSU thresholding method converges to the global optimal solution. This approach addresses the issue of inaccurate threshold detection in traditional algorithms, significantly improving the accuracy of dark region detection in SAR images. The detection results can further be used to identify low backscatter areas, such as shadows, water bodies, and asphalt roads. © 2025 SPIE.

关键词： Superpixels

来源：评论

学校读者我要写书评

暂无评论

CAU-Net: U-Shaped medical image segmentation network based on compound attention

CAU-Net: U-Shaped medical image segmentation network based o...

引用

2024 International Conference on Computer Vision and Image processing, CVIP 2024

作者： Liu, Jiaqi Mo, Jiaqing Xinjiang University School of Computer Science and Technology Key Laboratory of Signal Detection and Processing No.777 Huarui Street Shuimogou District Xinjiang Urumqi830017 China

ISBN: (纸本)9781510688384

In the context of the ongoing advancement of deep learning technologies, traditional image segmentation approaches are gradually become difficult for adapted to the requirements of medical imaging. The importance of medical image segmentation models is increasing, and they play a key role in accurately delineating anatomical structures and lesion regions. This is crucial for accurate diagnosis of diseases, formulation of treatment plans, and monitoring of disease course. In this paper, we propose a compound attention image segmentation model, CAU-Net, which aims to improve the segmentation accuracy of skin lesion images by introducing the receptive field module (RFB), the double squeezing-excitation (DSE) module, and our innovatively proposed compound attention (CA) module lesion image segmentation accuracy. The design of CAU-Net focuses on enhancing the feature extraction capability and contextual information capturing ability, and improving the edge segmentation capability and play a strong role in the processing of medical images with complex edges and chaotic backgrounds. The model reached an mDice to 0.9408 and a recall to 0.9418 on the ISIC 2018 dataset. Future work can further optimize these modules, extend to more medical image tasks. © 2025 SPIE.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

Infrared Small Target Detection with Feature Refinement and Context Enhancement 31st

Infrared Small Target Detection with Feature Refinement and...

引用

31st International Conference on Multimedia Modeling, MMM 2025

作者： Li, Xiuhong Zhu, Xinyue Li, Boyuan Li, Songlin Wang, Luyao Jia, Zhenhong School of Computer Science and Technology Xinjiang University Urumgi China Key Laboratory of Signal Detection and Processing Xinjiang University Urumqi China College of Mathematics and System Science Xinjiang University Urumqi China

ISBN: (纸本)9789819620609

Infrared small target detection has received widespread application and attention in both civilian and military fields. However, due to the very small size and lack of unique features of these targets, existing methods often suffer from inaccurate edge localization, and the targets are easily overwhelmed by complex backgrounds. To effectively address these issues, we designed a Feature Refinement and Context Enhancement Network (FCNet). The network consists of a multi-branch feature extraction module (MFEM), which integrates ordinary convolution and center difference convolution for diversified feature extraction, and further refines features through attention mechanism to enhance feature expression ability. In addition, in order to better capture the contextual information of the target, we have introduced the Context Enhancement Module (CEM). The CEM improves the robustness and detection accuracy of the network in complex backgrounds by preserving and enhancing important information in the image. Especially when processing infrared images with complex backgrounds and low signal-to-noise ratios, CEM can effectively highlight the target area and reduce false positives and false negatives. The experimental results on the SIRST dataset show that our FCNet performs well on multiple evaluation metrics. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Infrared imaging

来源：评论

学校读者我要写书评

暂无评论

Arbitrary-Shape Text Spotting Based on Global, Pixel and Sequence Semantics 4th

Arbitrary-Shape Text Spotting Based on Global, Pixel and S...

引用

4th International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2024

作者： Zhang, Chunhu Ibrayim, Mayire Hamdulla, Askar Deng, Qilin Xinjiang Key Laboratory of Signal Detection and Processing Urumqi China Xinjiang University School of Information Science and Engineering Urumqi China Xinjiang University College of Future Technology Urumqi China

ISBN: (纸本)9789819787012

The field of end-to-end text spotting has garnered significant interest in recent years, propelled by the revealed intrinsic synergies between scene text detection and recognition. While advancements have been made, the challenge of arbitrarily shaped scene text spotting persists. This paper introduces an innovative feature augmentation module that addresses the issues of limited receptive fields and weak representation typical of lightweight backbone networks, while also enhancing multi-scale information more effectively and reducing information loss during feature aggregation. Furthermore, to extract a richer set of backbone features, we propose a dual information attention mechanism that enables the backbone network to neuronally focus on salient information. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Dual-MambaNet: A Lightweight Dual-Branch Brain Image Segmentation Network Based on Local Attention and Mamba 27th

Dual-MambaNet: A Lightweight Dual-Branch Brain Image Segment...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Zhang, Feifei Shi, Fei Ren, Dayong Jia, Zhenhong Wang, Jianyi School of Computer Science and Technology Xinjiang University Urumqi Xinjiang830046 China Key Laboratory of Signal Detection and Processing Xinjiang University Urumqi Xinjiang830046 China National Key Laboratory for Novel Software Technology Nanjing University Nanjing210023 China

ISBN: (纸本)9783031781032

Brain tissue segmentation is critical for diagnosing and treating brain diseases. While Mamba-based models excel in the medical field, they face performance bottlenecks with high-resolution MRI images, often losing local feature information in complex texture structures. To address these challenges and enable deployment in resource-limited settings, we propose Dual-MambaNet, a lightweight segmentation model based on Mamba. In Dual-MambaNet, we introduce the Outlook attention module to capture local complex textures and structures in brain MRI images. Subsequently, we combined it with the Mamba block to construct a feature extractor (FE) encoder layer to couple local and global features. Additionally, we integrate dual decoder branches and a multi-level pixel contrastive loss function(MPCL) to better integrate local and global features. This method optimizes global feature representation by refining local complex textures and structural details, effectively capturing multi-level features in MRI images. Experimental results on public brain MRI datasets OASIS1 and MRBrainS13 demonstrate that Dual-MambaNet achieves high segmentation accuracy with minimal parameters and computational complexity, making it suitable for deployment in resource-limited medical environments. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

A Novel Construction of Two-Dimensional Z-Complementary Array Code Sets Based on Extended Boolean Functions

引用

IEEE Transactions on Vehicular technology 2025年第5期74卷 7821-7834页

作者： Tian, Liying Song, Xiaoshi Li, Yubo Zhang, Haijun Northeastern University School of Computer Science and Engineering Shenyang110819 China Northeastern University Key Laboratory of Intelligent Computing in Medical Image Ministry of Education Shenyang110189 China Yanshan University School of Information Science and Engineering Qinhuangdao066004 China Hebei Key Laboratory of Information Transmission and Signal Processing Qinhuangdao066004 China University of Science and Technology Beijing School of Computer and Communication Engineering Beijing100083 China

Two-dimensional (2-D) array sets with good 2-D correlation properties have received considerable attention in wireless communication systems. This paper focuses on 2-D Z-complementary array code sets (ZCACSs), which have ideal 2-D correlation properties within a rectangular zone around the origin. By developing a theory linking 2-D extended Boolean functions with 2-D arrays, we establish an efficient construction of 2-D ZCACSs. In contrast to previous work, the proposed construction is capable of producing 2-D ZCACSs with new flexible parameters. Constructions of three-dimensional (3-D) and higher-dimensional ZCACSs are further investigated. We remark that 2-D and higher-dimensional complete complementary codes (CCCs) can also be generated from the proposed constructions. © 1967-2012 IEEE.

关键词： Codes (symbols)

来源：评论

学校读者我要写书评

暂无评论

An efficient design of LPI Transmit beampattern for MIMO radar via Lp-norm smoothing

引用

IEEE Transactions on Aerospace and Electronic Systems 2025年

作者： Lan, Zhengyu Zuo, Lei Tang, Bo Hu, Juan Li, Xiang National Key Laboratory of Radar Signal Processing Xidian University China College of Electronic Engineering National University of Defense Technology China Tsinghua University China

The design of transmit beampattern may generate large instantaneous amplitude fluctuation in the synthesized waveforms at probed region, which distorts the low probability of intercept (LPI) performance of the multi-input multi-output (MIMO) radar systems. To avoid the radar signal being intercepted and identified by electronic interception measures (EIM), it is crucial to design LPI transmit beampattern with smoothed amplitude. In this paper, an effective LPI transmit beampattern design scheme is proposed for MIMO radar systems. First, we formulate the peak-to-average power ratio (PAR) model of synthetic waveform. Second, we establish an LPI transmit beampattern scheme to minimize the instantaneous amplitude fluctuation of the synthesized waveform and control the transmit beampattern, which is a non-convex, non-smooth min-max problem. To smooth the min-max problem, we approximate the max function by using the Lp norm. Then, we propose a majorization method of Lp norm, which transforms the Lp norm into a quadratic polynomial. Finally, to solve the min-max problem efficiently, a quadratic function majorization (QFMA) method is proposed. Numerical results demonstrate that the proposed method designs a LPI transmit beampattern efficiently. © 1965-2011 IEEE All rights reserved.

关键词： MIMO radar

来源：评论

学校读者我要写书评

暂无评论

RGR-KBQA: Generating Logical Forms for Question Answering Using Knowledge-Graph-Enhanced Large Language Model 31

RGR-KBQA: Generating Logical Forms for Question Answering Us...

引用

31st International Conference on Computational Linguistics, COLING 2025

作者： Feng, Tengfei He, Liang School of Computer Science and Technology Xinjiang University Urumqi830017 China Xinjiang Key Laboratory of Signal Detection and Processing Urumqi830017 China Department of Electronic Engineering Beijing National Research Center for Information Science and Technology Tsinghua University Beijing100084 China

ISBN: (纸本)9798891761964

In the field of natural language processing, Knowledge Base Question Answering (KBQA) is a challenging task that involves accurately retrieving answers from structured knowledge. Existing methods often face issues when generating query statements using LLMs, as the knowledge introduced may be imprecise and the models themselves may exhibit hallucination problems, leading to low accuracy, particularly when dealing with complex questions. To address these challenges, we introduce a novel semantic parsing approach called RGR-KBQA, which adopts a Retrieve-Generate-Retrieve framework. The first retrieval step introduces factual knowledge from a knowledge graph to enhance the semantic understanding capabilities of LLMs, thereby improving generation accuracy of logical form. The second step uses a fine-tuned model to generate the logical form, and the final step involves unsupervised relation and entity retrieval to further enhance generation accuracy. These two retrieval steps help alleviate the hallucination problems inherent in LLMs. Experimental results show that RGR-KBQA demonstrate promising performance on CWQ and WebQSP datasets. © 2025 Association for Computational Linguistics.

关键词： Question answering

来源：评论

学校读者我要写书评

暂无评论

Prompt Fusion and Aspect-Oriented Filtration for Aspect-Based Multimodal Sentiment Analysis

Prompt Fusion and Aspect-Oriented Filtration for Aspect-Base...

引用

2025 IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP 2025

作者： Sun, Jiachang Li, Xiuhong Zhu, Fuxian School of Computer Science and Technology XinJiang University Urumqi830046 China Xinjiang Key Laboratory of Signal Detection and Processing Urumqi830046 China School of Economics and Management XinJiang University Urumqi830046 China

ISBN: (纸本)9798350368741

Currently, aspect-based multimodal sentiment analysis (ABMSA) remains a highly hot research field, aiming to leverage various modalities such as images and text to determine the sentiment orientation of viewpoint entities. Although deep learning methods have made significant progress in this field, some challenges still exist: Image information unrelated to the viewpoint entity can introduce noise, insufficient interaction during modal fusion. To solve these problems, we propose a novel ABMSA model with prompt fusion and noise filtration. Specifically, we introduced an Image Noise Filtering (INF) module that can filter the extracted image features to reduce the noise introduced by the irrelevant between viewpoint entity and the image. Additionally, we implemented Multimodal Prompt Fusion (MPF) module to enhance the interaction between modalities, addressing the issue of minimal effective features related to viewpoint entity during the fusion stage of text and image features. Experimental results show that the model outperforms the baseline model on two public datasets, Twitter-2015 and Twitter-2017, demonstrating that our proposed approach effectively enhances the accuracy of aspect-based multimodal sentiment analysis tasks. © 2025 IEEE.

关键词： Aspect-Based Sentiment Analysis Multimodal Sentiment Analysis Noise Filtration Prompt Fusion

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：