检索结果-内蒙古大学图书馆

Context-aware Multi-level Question Embedding Fusion for visual question answering

Information Fusion 2024年 102卷

作者： Li, Shengdong Gong, Chen Zhu, Yuqing Luo, Chuanwen Hong, Yi Lv, Xueqiang School of Information Renmin University of China Beijing100872 China School of Computer Science and Engineering Nanjing University of Science and Technology Nanjing210094 China Zhejiang Lab Zhejiang Hangzhou311121 China Department of Computer Science California State University Los Angeles CA90032 United States School of Information Science and Technology Beijing Forestry University Beijing100083 China Engineering Research Center for Forestry-Oriented Intelligent Information Processing National Forestry and Grassland Administration 100083 China Beijing Key Laboratory of Internet Culture and Digital Dissemination Research Beijing Information Science and Technology University 100101 China

Question model has been widely concerned as the cornerstone of constructing Visual Question Answering (VQA) models. Existing question models attempt to exploit word context to extract multi-level concepts for modeling multi-level questions. However, they still have many defects. For example, most question models utilize simple fusion methods to fuse shallow modules and extract parameter-unshared low-level concepts, leading to poor modeling of multi-level questions;although some question models use deep bidirectional Transformer encoder in external knowledge transfer and BERT for multi-level questions, their complexity is still high. To solve these issues, we propose a novel low-complex multi-level contextual question model, termed Context-aware Multi-level Question Embedding Fusion (CMQEF). We formalize its concepts and theories, deduce its modeling process, optimization process and feature extraction process, analyze its low complexity and high expressiveness, and prove that it defines a new way to solve parameter non-sharing for extracting parameter-shared multi-level concepts and optimize the tradeoff between expressiveness and complexity in question models. Extensive experiments on VQAv2 and VQA-CPv2 validate that comparing with the state-of-the-art, our CMQEF outperforms it on SANs and UpDn, reduces the language priors of SANs and UpDn, and has preferable interpretability and applicability. Our code is available at https://***/lsdruc/CMQEF-for-VQA. © 2023 The Author(s)

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Imbalance Malware Classification by Decoupling Representation and Classifier 7th

Imbalance Malware Classification by Decoupling Representatio...

引用

7th International Conference on Artificial Intelligence and Security, ICAIS 2021

作者： Liu, Jiayin Zhuge, Chengchen Wang, Qun Guo, Xiangmin Li, Ziyuan Department of Computer Information and Cyber Security Jiangsu Police Institute Nanjing210031 China Jiangsu Electronic Data Forensics and Analysis Engineering Research Center Jiangsu Police Institute Nanjing210031 China Jiangsu Provincial Public Security Department Key Laboratory of Digital Forensics Jiangsu Police Institute Nanjing210031 China

ISBN: (纸本)9783030786205

In recent years, with the widespread utilization of automatic and semi-automatic malicious code generation tools, malware has been increased dramatically. Due to the growing malware variants, automatic detection and classification of malware samples is inevitable. A long-tailed distribution is one of the most common problems in malware dataset. However, existing general deep learning model is difficult to identify few-shot classes in the case of frequent classes. To address this challenge, we propose a new training scheme to classify unbalanced malware by decoupling representation learning from classifier learning. Specifically, in representation learning, the model is trained without balancing or oversampling the tail classes in order to learn the best representation. In classifier learning, we fine-tune the classifier under the class-balance sampling scheme. In addition, at this stage, we optimize the classifier with a fixed feature extractor during training to suppress the re-balancing effect on the backbone learned in the first stage. To assess the performance of the proposed approach, experiments are conducted on the BIG 2015 dataset. The experimental results demonstrate that our approach provides far higher classification performance, particularly in few-shot malware families than those proposed in existing literature. © 2021, Springer Nature Switzerland AG.

关键词： Malware

来源：评论

学校读者我要写书评

暂无评论

Strike a Balance in Continual Panoptic Segmentation

arXiv

引用

arXiv 2024年

作者： Chen, Jinpeng Cong, Runmin Luo, Yuxuan Ip, Horace Ho Shing Kwong, Sam Department of Computer Science City University of Hong Kong Hong Kong School of Control Science and Engineering Shandong University Shandong Jinan China Key Laboratory of Machine Intelligence and System Control Ministry of Education Shandong Jinan China Centre for Innovative Applications of Internet and Multimedia Technologies City University of Hong Kong Hong Kong Lingnan University Hong Kong

This study explores the emerging area of continual panoptic segmentation, highlighting three key balances. First, we introduce past-class backtrace distillation to balance the stability of existing knowledge with the adaptability to new information. This technique retraces the features associated with past classes based on the final label assignment results, performing knowledge distillation targeting these specific features from the previous model while allowing other features to flexibly adapt to new information. Additionally, we introduce a class-proportional memory strategy, which aligns the class distribution in the replay sample set with that of the historical training data. This strategy maintains a balanced class representation during replay, enhancing the utility of the limited-capacity replay sample set in recalling prior classes. Moreover, recognizing that replay samples are annotated only for the classes of their original step, we devise balanced anti-misguidance losses, which combat the impact of incomplete annotations without incurring classification bias. Building upon these innovations, we present a new method named Balanced Continual Panoptic Segmentation (BalConpas). Our evaluation on the challenging ADE20K dataset demonstrates its superior performance compared to existing state-of-the-art methods. Copyright © 2024, The Authors. All rights reserved.

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Challenges and Opportunities in the Responsible Use of AI and Human-computer Interaction

Challenges and Opportunities in the Responsible Use of AI an...

引用

2025 CHI Conference on Human Factors in Computing systems, CHI EA 2025

作者： Steckman, Laura Iyer, Nandini Kim, Boyoung Lyons, Joseph Cappuccio, Massimiliano Gaffley, Mark Richard Phillips, Elizabeth Sousa, Sonia Van Kleek, Max Air Force Research Laboratory ArlingtonVA United States Air Force Research Laboratory Air Force Office of Scientific Research European Office of Aerospace Research and Development London United Kingdom George Mason University Incheon Korea Republic of Air Force Research Laboratory Wright-Patterson AFBOH United States School of Engineering and Information Technology University of New South Wales CanberraNSW Australia University of Cape Town Western Province Cape Town South Africa George Mason University FairfaxVA United States School of Digital Technologies Tallinn University Tallinn Estonia Department of Computer Science University of Oxford Oxfordshire Oxford United Kingdom

ISBN: (纸本)9798400713958

The use of Artificial Intelligence (AI) is fraught with ethical challenges as we increasingly start to use and rely on AI in society. Despite the potential promises of AI to enhance human abilities and increase how Human-computer Interaction (HCI) utilizes HUMAN-centered inputs, issues arise wherein AI or its outputs violate human ethics and/or social norms. In response, the academic, industrial, and military research communities have advocated for increased research to understand how to best develop, test, and use AI technologies and their outputs in ways that are consistent with our ideals, values and morals. Fields of inquiry, such as machine ethics, have burgeoned as a result. Yet, moral and ethical challenges associated with AI and HCI persist. Recent literature on AI ethics and responsible use of AI remains limited to a small number of principles. HCI requires considering a larger number of principles for human-centered design, or even the incorporation of ethics and values more holistically, yet remains limited without a solid, more inclusive scientific foundation. The current panel will bring together diverse perspectives in discussing issues associated with research related to the responsible use of AI and HCI. The participants, along with the expert panelists, will play a key role in shaping the direction of the conversation through active participation. © 2025 Copyright held by the owner/author(s).

关键词： Industrial research

来源：评论

学校读者我要写书评

暂无评论

Jseun: Joint Semantic-Instance Segmentation of Point Cloud Via Neighbor Features for Indoor Scenes

SSRN

引用

SSRN 2023年

作者： Chen, Hui Jiang, Jinlong Chen, Wanlou Liu, Wanquan Liu, Peter Xiaoping College of Automation Engineering Shanghai University of Electric Power Shanghai200090 China Laboratory of System Control and Information Processing Ministry of Education Shanghai200240 China Nanovision Medical Technology Co. Ltd Shanghai200000 China School of Intelligent Systems Engineering Sun Yat-sen University Guangzhou510275 China Department of Systems and Computer Engineering Carleton University OttawaONK1S 5B6 Canada

Efficient and accurate segmentation of the point cloud for indoor scenes is an important task for 3D scene understanding. Due to the complexity of indoor environments. It is difficult to fully extract the details of point cloud, resulting in poor recognizability in existing models. In this paper, a novel framework called joint semantic-instance segmentation via enhancedutilization of neighbor features is proposed to achieve semantic and instance segmentation of indoor scenes. This framework mines point cloud information from a local scale, but it is different from other models in that it focuses on the full mining of point cloud neighbor information. In this regard, a new neighbor feature extraction module is developed to enrich the expression of neighborhood information and enlarge the construction of the neighborhood neighbor scope. Besides, semantic and instance segmentations take advantages of each other by constructing a mutual promotion model of semantic and instance, which can obtain more accurate semantic and instance predictions for indoor scenes. The proposed scheme is evaluated on the S3DIS dataset and compared with existing methods. Experimental results demonstrate that this method is more effective and efficient than previous approaches in joint 3D semantic and instance segmentation. © 2023, The Authors. All rights reserved.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Lesion-Guided Selective Multi-Modal Integration for Prostate Cancer Segmentation and Pi-Rads Grading in Mp-Mri

SSRN

引用

SSRN 2024年

作者： Wu, Menglin Li, Fan Tao, Yuhui Zhang, Yuhan Wang, Shanshan Burstein, Pablo D. Mu, Xuetao Zhu, Jie School of Computer Science and Information Engineering Nanjing Tech University Nanjing China Carbon Medical Device Ltd. Shenzhen China Digital Medical Research Center School of Basic Medical Sciences Fudan University Shanghai China Shanghai Key Laboratory of MICCAI Shanghai China Department of Computer Science and Engineering The Chinese University of Hong Kong Hong Kong Shenzhen Research Institute The Chinese University of Hong Kong Hong Kong Department of Radiology The Third Medical Center of Chinese PLA General Hospital Beijing China Department of Urology The Third Medical Center of Chinese PLA General Hospital Beijing China

Background and ObjectivesProstate cancer (PCa) presents a significant global health challenge affecting men. Accurate segmentation and grading of PCa lesions in multiparametric Magnetic Resonance Imaging (mp-MRI) are essential for effective diagnosis and treatment planning. In this work, we investigate the performance of automated model in PCa lesion segmentation and PI-RADS grading in *** that the lesion’s perceived characteristics are strongly related to both imaging modalities and lesion locations, we propose a Lesion-guided Selective Multi-modal Integration (LeSMI) module. This module incorporates two advanced attention mechanisms—Cross Modality Attention (CMA) and Localized Lesion Attention (LLA)—to dynamically integrate crucial information within the same and across modalities. Additionally, inspired by clinical workflows, our framework employs a two-stage Prostate Cancer Segmentation and Grading (PCaSG) strategy, leveraging knowledge from segmentation to improve PI-RADS grading *** on the Prostate158 and PI-CAI Challenge datasets, our method demonstrated superior performance, achieving a Dice Similarity Coefficient (DSC) of 0.5130 and a lesion-level quadratic-weighted kappa score (QWKl) of 0.6248 on Prostate158, and a DSC of 0.4381 and a QWKl of 0.4298 on PI-CAI. These results represent improvements of up to 2% in DSC and 17% in QWKl over current state-of-the-art models on Prostate158, and enhancements of 4% in DSC and 3% in QWKl on *** proposed model’s robustness in handling diverse lesion presentations and its reliable assessments underscore its clinical applicability, offering significant advancements in both segmentation accuracy and PI-RADS grading. © 2024, The Authors. All rights reserved.

关键词： Urology

来源：评论

学校读者我要写书评

暂无评论

Parallel chaotic sources based on microcombs

Parallel chaotic sources based on microcombs

引用

Conference on Lasers and Electro-Optics (CLEO)

作者： Bitao Shen Haowen Shu Weiqiang Xie Ruixuan Chen Zhi Liu Xuguang Zhang Yimeng Wang Yunhao Zhang Buwen Cheng Shaohua Yu John E. Bowers Lin Chang Xingjun Wang State Key Laboratory of Advanced Optical Communications System and Networks School of Electronics Peking University Beijing China Department of Electrical and Computer Engineering University of California Santa Barbara CA USA State Key Laboratory on Integrated Optoelectronics Institute of Semiconductors Chinese Academy of Sciences Beijing China Center of Materials Science and Optoelectronics Engineering University of Chinese Academy of Sciences Beijing China Peng Cheng Laboratory Shenzhen China Frontiers Science Center for Nano-Optoelectronics Peking University Beijing China

Massively parallel optical chaotic sources with large bandwidth are demonstrated based on chaotic microcombs in the AlGaAsOI platform. 30-channel parallel random bit generation with 30 Gbit/s per channel is realized b... 详细信息

ISBN: (纸本)9781957171258

关键词： Diode lasers Integrated photonics Kerr effect Nonlinear effects Optical microscopy Random number generation

来源：评论

学校读者我要写书评

暂无评论

PDCapsNet: A Convolution Neural Network Method for COVID-19 Detection from Chest X-Ray Images

PDCapsNet: A Convolution Neural Network Method for COVID-19 ...

引用

IEEE International Symposium on Information (IT) in Medicine and Education, ITME

作者： Zhihao Liang Huijuan Lu Cunqian You Wenjie Zhu Li Xie Yudong Yao Key Laboratory of Electromagnetic Wave Information Technology and Metrology College of Information Engineering China Jiliang University Hangzhou Zhejiang Province China College of Information Engineering China Jiliang University Drore Information and Technology Co. LTD Hangzhou Zhejiang Province China College of Information Engineering China Jiliang University Modern Educational Technology Center Hangzhou City Zhejiang Province China College of Biomedical Engineering and Instrumentation Science Zhejiang University Institute of Advanced Digital Technologies and Instrumentation Hangzhou Zhejiang Province China Department of Electrical and Computer Engineering Stevens Institute of Technology Hoboken New Jersey USA

The COVID-19 epidemic started in 2019 and it is still important to develop intelligent and efficient diagnostic methods for the detection of COVID-19. In the task of COVID-19 X-ray image recognition, conventional deep neural networks encounter challenges in effectively discerning distinctive features within pneumonia images, resulting in limited generalization capability when confronted with novel samples. To address this issue, we introduce a novel model in this paper, denoted as PDCapsNet. The model effectively exploits Pyconv's multi-scale fusion technique and leverages the capsule network's capacity to encapsulate diverse attributes of specific entities within COVID-19 X-ray images, resulting in enhancements in both accuracy and generalization. Experimental results show that the proposed model outperforms other previous related works.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Identification of Weather Phenomena Based on Lightweight Convolutional Neural Networks

引用

computers, Materials & Continua 2020年第9期64卷 2043-2055页

作者： Congcong Wang Pengyu Liu Kebin Jia Xiaowei Jia Yaoyao Li Faculty of Information Technology Beijing University of TechnologyBeijing100124China Beijing Laboratory of Advanced Information Networks Beijing100124China Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing University of TechnologyBeijing100124China Department of Computer Science and Engineering University of MinnesotaTwin CitiesUSA

Weather phenomenon recognition plays an important role in the field of ***,weather radars and weathers sensor have been widely used for weather ***,given the high cost in deploying and maintaining the devices,it is difficult to apply them to intensive weather phenomenon ***,advanced machine learning models such as Convolutional Neural Networks(CNNs)have shown a lot of promise in meteorology,but these models also require intensive computation and large memory,which make it difficult to use them in *** practice,lightweight models are often used to solve such ***,lightweight models often result in significant performance *** this end,after taking a deep dive into a large number of lightweight models and summarizing their shortcomings,we propose a novel lightweight CNNs model which is constructed based on new building *** experimental results show that the model proposed in this paper has comparable performance with the mainstream non-lightweight model while also saving 25 times of memory *** memory reduction is even better than that of existing lightweight models.

关键词： Deep learning convolution neural networks lightweight models weather identification

来源：评论

学校读者我要写书评

暂无评论

A Two-way Automatic Alignment Visible Light Communication system Based on Red Laser

A Two-way Automatic Alignment Visible Light Communication Sy...

引用

IEEE International Conference on Electrical engineering and Photonics (EExPolytech)

作者： Han Zhang Jiaming Li Zhen Liu Hongming Zhang Department of Electronic Engineering Beijing National Research Center for Information Science and Technology (BNRist) Tsinghua University Beijing China Zhengzhou Power Supply Company State Grid Henan Electric Power Company Zhengzhou China Key Laboratory of Digital TV System of Guangdong Province and Shenzhen City Research Institute of Tsinghua University in Shenzhen Shenzhen China

Visible light communication(VLC) has shown in explosive growth recent years, and LEDs based visible light wireless communication has the potential to change lives in the future with applications such as LiFi, smart lighting and indoor positioning. Due to the low bandwidth of LED devices, the VLC systems using LDs(laser diodes) as light sources can achieve faster transmission speed. In this paper, we propose a high-speed visible light communication system with automatic alignment function using red laser. Each tranceiver in the system can be moved in the angle of view of another, and be aligned in real time using cameras, computers and two-dimensional turntable. TCP data transmission rate more than 700 Mbps with a distance of 6 m free-space link, $\pm 10^{\circ}\text{angle}$ of view was achieved.

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：