检索结果-内蒙古大学图书馆

MedBLIP: A Multimodal Method of Medical Question-Answering Based on Fine-Tuning Large Language Model

computerized Medical Imaging and Graphics 2025年

作者： Lejun Gong Jiaming Yang Shengyuan Han Yimu Ji Jiangsu Key Lab of Big Data Security & Intelligent Processing School of computer science Nanjing University of Posts and Telecommunications Nanjing210023 China Institute of High Performance Computing and Bigdata Nanjing University of Posts and Telecommunications Nanjing 210023 China

Medical visual question answering is crucial for effectively interpreting medical images containing clinically relevant information. This study proposes a method called MedBLIP (Medical Treatment Bootstrapping Language-Image Pretraining) to tackle visual language generation tasks related to chest X-rays in the medical field. The method combine an image encoder with a large-scale language model, and effectively generates medical question-answering text through a strategy of freezing the image encoder based on the BLIP-2 model. Firstly, chest X-ray images are preprocessed, and an image sample generation algorithm is used to enhance the text data of doctor-patient question-answering, thereby increasing data diversity. Then, a multi-layer convolutional image feature extractor is introduced to better capture the feature representation of medical images. During the fine-tuning process of the large language generation model, a new unfreezing strategy is proposed, which is to unfreeze different proportions of the weights of the fully connected layer to adapt to the data in the medical field. The image feature extractor is responsible for extracting key features from images, providing the model with rich visual information, while the text feature extractor accurately captures the essential requirements of the user's question. Through their synergistic interaction, the model can more effectively integrate medical images and user inquiries, thereby generating more accurate and relevant output content. The experimental results show that unfreezing 31.25% of the weights of the fully connected layer can significantly improve the performance of the model, with ROUGE-L reaching 66.12%, and providing a more accurate and efficient answer generation solution for the medical field. The method of this study has potential applications in the field of medical language generation tasks. Although the proposed model cannot yet fully replace human radiologists, it plays an indispensable role

关键词：

来源：评论

学校读者我要写书评

暂无评论

Weakly supervised learning for textbook question answering

TechRxiv

引用

TechRxiv 2021年

作者： Ma, Jie Chai, Qi Huang, Jingyue Liu, Jun You, Yang Zheng, Qinghua National Engineering Lab for Big Data Analytics School of Computer Science and Technology Xi’an Jiaotong University Shaanxi Xi’an710049 China Shaanxi Province Key Laboratory of Satellite and Terrestrial Network Tech. R&D School of Computer Science and Technology Xi’an Jiaotong University Shaanxi Xi’an710049 China The High Performance Computing for Artificial intelligence Lab Department of Computer Science National University of Singapore 117417 Singapore

Textbook Question Answering (TQA) is the task of answering diagram and non-diagram questions given large multimodal contexts consisting of abundant text and diagrams. Deep text understandings and effective learning of diagram semantics are important for this task due to its specificity. In this paper, we propose a Weakly Supervised learning method for TQA (WSTQ), which regards the incompletely accurate results of essential intermediate procedures for this task as supervision to develop Text Matching (TM) and Relation Detection (RD) tasks and then employs the tasks to motivate itself to learn strong text comprehension and excellent diagram semantics respectively. Specifically, we apply the result of text retrieval to build positive as well as negative text pairs. In order to learn deep text understandings, we first pre-train the text understanding module of WSTQ on TM and then fine-tune it on TQA. We build positive as well as negative relation pairs by checking whether there is any overlap between the items/regions detected from diagrams using object detection. The RD task forces our method to learn the relationships between regions, which are crucial to express the diagram semantics. We train WSTQ on RD and TQA simultaneously, i.e., multitask learning, to obtain effective diagram semantics and then improve the TQA performance. Extensive experiments are carried out on CK12-QA and AI2D to verify the effectiveness of WSTQ. Experimental results show that our method achieves significant accuracy improvements of 5.02% and 4.12% on test splits of the above datasets respectively than the current state-of-the-art baseline. We have released our code on https://***/dr-majie/WSTQ. © 2021, CC BY.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

DVO + LCLMF: A web service recommendation mechanism with QoS privacy preservation

引用

Concurrency and Computation: Practice and Experience 2022年第18期34卷 e7003-e7003页

作者： Li, Kui Ji, Yi-mu Liu, Shang-dong Wu, Fei Yao, Hai-chang He, Jing Liu, Qiang Liu, Yan-lan Shao, Si-si You, Shuai School of Computer Science Nanjing University of Posts and Telecommunications Nanjing China Nantong Bigdata Research Key Laboratory for Rehabilitation Medicine The Second People's Hospital of Nantong Nantong China Jiangsu High Technology Research Key Laboratory for Wireless Sensor Networks Nanjing China Institute of High Performance Computing and Bigdata Nanjing University of Posts and Telecommunications Nanjing China Jiangsu HPC and Intelligent Processing Engineer Research Center Nanjing China School of Software and Electrical Engineering Swinburne University of Technology MelbourneVIC Australia

QoS-aware based web service recommendation is one of the crucial solutions to help users find high-quality web services. To accurately predict the QoS values of candidate services, it is usually required to collect historical QoS data of users (QoS data for short). If these collected QoS data are improperly processed, QoS data privacy may be threatened. However, how to accurately predict the QoS values of candidate services while protecting QoS data privacy has not been well studied. In response to the situation, we propose a hybrid web service recommendation mechanism, which is divided into three parts. In the first part, the QoS data privacy preservation algorithm, which called DVO, is proposed based on keeping the cosine similarity of QoS data unchanged, that is, to realize the confusion of QoS data while ensuring the availability of QoS data remains unchanged. In the second part, a hybrid matrix factorization model based on location information and service features, which called LCLMF, is proposed to improve the accuracy of QoS values prediction. According to DVO and LCLMF, the DVO + LCLMF is designed in the third part, which can accurately predict QoS values while protecting QoS data privacy. The experimental results show that DVO + LCLMF can accurately predict the QoS values of candidate services on the basis of attaining QoS data privacy protection. © 2022 John Wiley & Sons, Ltd.

关键词： Data privacy

来源：评论

学校读者我要写书评

暂无评论

Span-based localizing network for natural language video localization

arXiv

引用

arXiv 2020年

作者： Zhang, Hao Sun, Aixin Jing, Wei Zhou, Joey Tianyi School of Computer Science and Engineering Nanyang Technological University Institute of High Performance Computing A*STAR Institute for Infocomm Research A*STAR

Given an untrimmed video and a text query, natural language video localization (NLVL) is to locate a matching span from the video that semantically corresponds to the query. Exist-ing Solutions formulate NLVL either as a rank-ing task and apply multimodal matching ar-chitecture, or as a regression task to directly regress the target video span. In this work, we address NLVL task with a span-based QA approach by treating the input video as text passage. We propose a video span localizing network (VSLNet), on top of the Standard span-based QA framework, to address NLVL. The proposed VSLNet tackles the differences between NLVL and span-based QA through a simple and yet effective query-guided high-lighting (QGH) strategy. The QGH guides VSLNet to search for matching video span within a highlighted region. Through extensive experiments on three benchmark datasets, we show that the proposed VSLNet outper-forms the state-of-the-art methods;and adopt-ing span-based QA framework is a promising direction to solve NLVL. Copyright © 2020, The Authors. All rights reserved.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Experimental Characterization and Yield Strength Model for Process-Structure-Property of Selective Laser Melted 316l

SSRN

引用

SSRN 2022年

作者： Chen, Yefeng Wang, Xiaowei Li, Dong Zhou, Dewen Jiang, Yong Yang, Xinyu Liu, Chenglu Leen, Sean B. Gong, Jianming School of Mechanical and Power Engineering Nanjing Tech University Nanjing211816 China Jiangsu Key Lab of Design and Manufacture of Extreme Pressure Equipment Nanjing211816 China Key Laboratory for Light-Weight Materials Nanjing Tech University Nanjing211816 China Institute of High Performance Computing A*STAR Research Entities Connexis 1 Fusionopolis Way #16-16 138632 Singapore Mechanical Engineering School of Engineering College of Science and Engineering NUI Galway Ireland

Heat treatment is the most common method to relieve residual stress and to adjust tensile properties in additively manufactured material. However, the well-known strengthening factors (e.g., dislocation density, cellular sub-structure, low angle grain boundaries, nano-oxide particle) in selective laser melted 316L (SLM 316L) are strongly influenced by heat-treatment temperature. In this work, horizontal and vertical oriented SLM 316L specimens are heat treated from 550 ℃ to 1150 ℃, followed by tensile tests and microstructural characterization. With increasing temperature, strength (e.g., yield strength, ultimate tensile strength) is found to decrease and elongation is found to increase. The heat treatments of 550 ℃, 650 ℃ and 750 ℃ lead to coarsening of cellular sub-structure. Furthermore, the cellular sub-structure and melt boundaries annihilate at 950 ℃. When the temperature reaches 1150 ℃, SLM 316L has finished recrystallization with volume fraction of nano-oxide particles increasing greatly. Based on microstructural characterization, the proposed model, considering strengthening mechanisms of cellular sub-structure, nano-oxide particle and grain boundaries, give satisfactory accuracy to predict yield strength for horizontal orientation. The yield strength for vertical orientation can also be predicted by this model when modified by consideration of anisotropy of columnar grains and lack-of-fusion defects. The anisotropy and lack-of-fusion lead to 5.4% reduction in yield strength for the vertical orientation. This work establishes quantitative relationships for heat treatment-microstructure-property, in which the anisotropy of tensile properties is considered. © 2022, The Authors. All rights reserved.

关键词： Yield stress

来源：评论

学校读者我要写书评

暂无评论

Intensity correlations in decoy-state BB84 quantum key distribution systems

arXiv

引用

arXiv 2024年

作者： Trefilov, Daniil Sixto, Xoel Zapatero, Víctor Huang, Anqi Curty, Marcos Makarov, Vadim Vigo Quantum Communication Center University of Vigo VigoE-36310 Spain School of Telecommunication Engineering Department of Signal Theory and Communications University of Vigo VigoE-36310 Spain atlanTTic Research Center University of Vigo VigoE-36310 Spain Russian Quantum Center Skolkovo Moscow121205 Russia National Research University Higher School of Economics Moscow101000 Russia Institute for Quantum Information State Key Laboratory of High Performance Computing College of Computer Science and Technology National University of Defense Technology Changsha410073 China NTI Center for Quantum Communications National University of Science and Technology MISiS Moscow119049 Russia

The decoy-state method is a prominent approach to enhance the performance of quantum key distribution (QKD) systems that operate with weak coherent laser sources. Due to the limited transmissivity of single photons in optical fiber, current experimental decoy-state QKD setups increase their secret key rate by raising the repetition rate of the transmitter. However, this usually leads to correlations between subsequent optical pulses. This phenomenon leaks information about the encoding settings, including the intensities of the generated signals, which invalidates a basic premise of decoy-state QKD. Here we characterize intensity correlations between the emitted optical pulses in two industrial prototypes of decoy-state BB84 QKD systems and show that they significantly reduce the asymptotic key rate. In contrast to what has been conjectured, we experimentally confirm that the impact of higher-order correlations on the intensity of the generated signals can be much higher than that of nearest-neighbour correlations. Copyright © 2024, The Authors. All rights reserved.

关键词： Quantum cryptography

来源：评论

学校读者我要写书评

暂无评论

Combined density functional theory and boundary element methods study on optical and electronic properties of interfacial Au/TiO2 defects

引用

Optical Materials: X 2023年 20卷

作者： Dang, Diem Thi-Xuan Vu, Nam Hoang Vu, Thu Thi-Hanh Thoai, Nam Kawazoe, Yoshiyuki Phan, Bach Thang Nguyen-Manh, Duc Ho Chi Minh City Viet Nam Ho Chi Minh City Viet Nam Faculty of Materials Science and Technology University of Science Ho Chi Minh City Viet Nam Faculty of Physics and Engineering Physics University of Science Ho Chi Minh City Viet Nam High Performance Computing Lab and Faculty of Computer Science & Engineering University of Technology Ho Chi Minh City Viet Nam New Industry Creation Hatchery Center Tohoku University Sendai980-8579 Japan Department of Physics Suranaree University of Technology Ratchasima Nakhon Thailand Department of Physics and Nanotechnology SRM Institute of Science and Technology Tamil Nadu Kattankulathur603203 India Culham Center for Fusion Energy United Kingdom Atomic Energy Authority OX14 3DB United Kingdom

This study investigates the optical and electrical properties of Au/TiO2 nanoparticles with different sizes and defects by combining micro- and macro-scale simulations through density functional theory (DFT) and boundary element method (BEM), respectively. The O-vacancy in the sample enhances the bonding and the electron transfer at the Au/TiO2 perimeter zone, as indicated by the binding energy and charge density difference in DFT calculations, respectively. BEM simulation indicates a stronger electromagnetic field in the Au/TiO2 interface of the sample with O-vacancy. DFT calculations demonstrate that the optical absorption of Au/TiO2 in the ultraviolet range is well described by the absorption coefficient spectra, whereas the BEM results well describe the range of the surface plasmon resonance absorption of Au/TiO2 in the wavelength range of 480–550 nm through the absorption cross-section. This study demonstrates the combined role of DFT and BEM in studying the electrical and optical properties of the nanocrystal system. © 2023 The Author(s)

关键词： Density functional theory

来源：评论

学校读者我要写书评

暂无评论

Tackling Instance-Dependent label Noise via a Universal Probabilistic Model

arXiv

引用

arXiv 2021年

作者： Wang, Qizhou Han, Bo Liu, Tongliang Niu, Gang Yang, Jian Gong, Chen Key Laboratory of Intelligent Perception and Systems for High-Dimensional Information of MoE School of Computer Science and Engineering Nanjing University of Science and Technology China Department of Computer Science Hong Kong Baptist University Hong Kong Trustworthy Machine Learning Lab School of Computer Science Faculty of Engineering The University of Sydney Australia Japan Department of Computing Hong Kong Polytechnic University Hong Kong

The drastic increase of data quantity often brings the severe decrease of data quality, such as incorrect label annotations, which poses a great challenge for robustly training Deep Neural Networks (DNNs). Existing learning methods with label noise either employ ad-hoc heuristics or restrict to specific noise assumptions. However, more general situations, such as instance-dependent label noise, have not been fully explored, as scarce studies focus on their label corruption process. By categorizing instances into confusing and unconfusing instances, this paper proposes a simple yet universal probabilistic model, which explicitly relates noisy labels to their instances. The resultant model can be realized by DNNs, where the training procedure is accomplished by employing an alternating optimization algorithm. Experiments on datasets with both synthetic and real-world label noise verify that the proposed method yields significant improvements on robustness over state-of-the-art counterparts. Copyright © 2021, The Authors. All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

performance Analysis of Change Detection Algorithms on Multispectral Imagery

Performance Analysis of Change Detection Algorithms on Multi...

引用

International Conference on Computational Intelligence and Communication Networks (CICN)

作者： Indira Bidari Satyadhyan Chickerur Rekha.M. Talikoti Smita.S. Kapali Sushmita Talawar Soumya Sangam School of Computer Science and Engineering KLE Technological University Hubballi India Center for High Performance Computing KLE Technological University Hubballi India

ISBN: (数字)9781728193939

ISBN: (纸本)9781728193946

Due to growing human population and technology, huge climate change occurs which impact the environment and lives. Satellite images are profound for monitoring the ground surface. Due to the enormous availability of multispectral images, various applications based on classification, change detection have emerged. Researchers have developed lots of Multispectral Change Detection Methods (MSCD) till date. These methods show essential sub-pixel level details, such as the abundance variation of each underlying material at a given location, or the shift in material distribution within the scene, with time or as a result of significant events such as a natural disaster. The main aim of the proposed work is to present a comparative study of various unsupervised methods for detecting binary changes in multispectral imagery. Through experimental study, we provide a comparative analysis of the algorithms. The various algorithms considered for our study are (1) Principal Component Analysis (PCA), (2) PCA with K-means clustering, (3) Multivariate Alteration Detection (MAD), (4) Iteratively Reweighted Multivariate Alteration Detection (IRMAD). Using real-world multi-temporal multispectral imaging dataset, we assess and compare the performance of all these algorithms and their time efficiency on Central Processing Unit (CPU) and Graphics Processing Unit (GPU). The empirical findings, accompanied by a description of each algorithms pros and cons, are intended to help researchers pick the procedures with the good features for MSCD applications.

关键词： Principal component analysis Graphics processing units Change detection algorithms Correlation Central Processing Unit Satellites Detection algorithms

来源：评论

学校读者我要写书评

暂无评论

Surgical video workflow analysis via visual-language learning

npj Health Systems

引用

npj Health Systems 2025年第1期2卷 1-12页

作者： Pengpeng Li Xiangbo Shu Jinhui Tang Chun-Mei Feng Yifei Feng Wangmeng Zuo School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing China Institute of High Performance Computing (IHPC) Agency for Science Technology and Research (A*STAR) Singapore Republic of Singapore The First School of Clinical Medicine Nanjing Medical University Nanjing China Department of General Surgery The First Affiliated Hospital of Nanjing Medical University Nanjing China Faculty of Computing Harbin Institute of Technology Harbin China

Surgical video workflow analysis has made intensive development in computer-assisted surgery by combining deep learning models, aiming to enhance surgical scene analysis and decision-making. However, previous research has primarily focused on coarse-grained analysis of surgical videos, e.g., phase recognition, instrument recognition, and triplet recognition that only considers relationships within surgical triplets. In order to provide a more comprehensive fine-grained analysis of surgical videos, this work focuses on accurately identifying triplets from surgical videos. Specifically, we propose a vision-language deep learning framework that incorporates intra- and inter- triplet modeling, termed I2TM, to explore the relationships among triplets and leverage the model understanding of the entire surgical process, thereby enhancing the accuracy and robustness of recognition. Besides, we also develop a new surgical triplet semantic enhancer (TSE) to establish semantic relationships, both intra- and inter-triplets, across visual and textual modalities. Extensive experimental results on surgical video benchmark datasets demonstrate that our approach can capture finer semantics, achieve effective surgical video understanding and analysis, with potential for widespread medical applications.

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：