检索结果-内蒙古大学图书馆

Unpaired Image Captioning With semantic-Constrained Self-Learning

ieee TRANSACTIONS ON MULTIMEDIA 2022年 24卷 904-916页

作者： Ben, Huixia Pan, Yingwei Li, Yehao Yao, Ting Hong, Richang Wang, Meng Mei, Tao Hefei Univ Technol Sch Comp & Informat Hefei 230009 Peoples R China JD AI Res CV Lab Beijing 100105 Peoples R China

Image captioning has been an emerging and fast-developing research topic. Nevertheless, most existing works heavily rely on large amounts of image-sentence pairs and therefore hinder the practical applications of captioning in the wild. In this paper, we present a novel Semantic-Constrained Self-learning (SCS) framework that explores an iterative self-learning strategy to learn an image captioner with only unpaired image and text data. Technically, SCS consists of two stages, i.e., pseudo pair generation and captioner re-training, iteratively producing "pseudo" image-sentence pairs via a pre-trained captioner and re-training the captioner with the pseudo pairs, respectively. Particularly, both stages are guided by the recognized objects in the image, that act as semantic constraint to strengthen the semantic alignment between the input image and the output sentence. We leverage a semantic-constrained beam search for pseudo pair generation to regularize the decoding process with the recognized objects via forcing the inclusion/exclusion of the recognized/irrelevant objects in output sentence. For captioner re-training, a self-supervised triplet loss is utilized to preserve the relative semantic similarity ordering among generated sentences with regard to the input image triplets. Moreover, an object inclusion reward and an adversarial reward are adopted to encourage the inclusion of the predicted objects in the output sentence and pursue the generation of more realistic sentences during self-critical training, respectively. Experiments conducted on both dependent and independent unpaired data validate the superiority of SCS. More remarkably, we obtain the best published CIDEr score to-date of 74.7\% on COCO Karpathy test split for unpaired image captioning.

关键词： Semantics Image recognition Training visualization Decoding Task analysis Dogs Encoder-decoder networks image captioning self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

A review of bearing fault diagnosis based on weakly supervised learning

A review of bearing fault diagnosis based on weakly supervis...

引用

2023 CAA symposium on Fault Detection, Supervision and Safety for Technical Processes, SAFEPROCESS 2023

作者： Zhao, Bichuang Cen, Jian Si, Weiwei Huang, Hankun Yang, Zhuohong Chen, Honghua Guangdong Polytechnic Normal University Guangzhou Intelligent Building Equipment Information Integration and Control Key Laboratory GuangZhou China

ISBN: (纸本)9798350337754

In the era of big data, mechanical intelligent fault diagnosis is more intelligent due to the large-scale collection of data. However, fault data is often difficult to obtain, resulting in small or even no samples. This article reviews the research on weakly supervised learning in the field of fault diagnosis. For few shot learning methods, it is divided into data augmentation method and feature extraction and transfer method. Metric-based zero-shot fault diagnosis using pre-trained word vectors to calculate similarity. Finally, this paper summarizes and prospects the future research of weakly supervised learning. © 2023 ieee.

关键词： Failure analysis

来源：评论

学校读者我要写书评

暂无评论

Sentiment analysis with Lipschitz Recurrent Neural Networks

Sentiment Analysis with Lipschitz Recurrent Neural Networks

引用

2023 International symposium on Networks, Computers and Communications, ISNCC 2023

作者： Hasan, Mahmudul Shetty, Sachin Old Dominion University Department of Electrical and Computer Engineering NorfolkVA United States

ISBN: (纸本)9798350335590

At present, sentiment analysis is used in almost every sector. In medical chatbots and other healthcare domains, government agencies and other essential industries use sentiment analysis for public safety, improving the service quality of products, and other necessary purposes. A large amount of data is required to analyze and briefly overview any product, movie, or comment. This is a challenging task to determine the sentiment from plain text. However, researchers use Natural Language Processing (NLP) to do this task accurately. Our model is based on the Lipschitz Recurrent Neural Networks (LRNN), which analyzed the IMBD dataset. The ultimate goal was to reduce the vanishing gradient problem and find the best accuracy number and a good number of evaluation metrics. The LRNN is based on hidden-to-hidden matrices, which update the hidden state, and this hidden state is used to predict the sentiment with reasonable accuracy. Finally, we got 86.75 % accuracy. © 2023 ieee.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

HLSFactory: A Framework Empowering High-Level Synthesis datasets for Machine Learning and Beyond 24

HLSFactory: A Framework Empowering High-Level Synthesis Data...

引用

6th International symposium on Machine Learning for CAD (MLCAD)

作者： Abi-Karam, Stefan Sarkar, Rishov Seigler, Allison Lowe, Sean Wei, Zhigang Chen, Hanqiu Rao, Nanditha John, Lizy Arora, Aman Hao, Cong Georgia Inst Technol Atlanta GA 30332 USA Georgia Tech Res Inst Atlanta GA 30332 USA Univ Texas Austin Austin TX USA Arizona State Univ Tempe AZ USA Int Inst Informat Technol Bangalore Bangalore Karnataka India

ISBN: (数字)9798350363562

ISBN: (纸本)9798400706998

Machine learning (ML) techniques have been applied to high-level synthesis (HLS) flows for quality-of-result (QoR) prediction and design space exploration (DSE). Nevertheless, the scarcity of accessible highquality HLS datasets and the complexity of building such datasets present great challenges to FPGA and ML researchers. Existing datasets either cover only a subset of previously published benchmarks, provide no way to enumerate optimization design spaces, are limited to a specific vendor, or have no reproducible and extensible software for dataset construction. Many works also lack user-friendly ways to add more designs to existing datasets, limiting wider adoption and sustainability of such datasets. In response to these challenges, we introduce HLSFactory, a comprehensive framework designed to facilitate the curation and generation of high-quality HLS design datasets. HLSFactory has three main stages: 1) a design space expansion stage to elaborate single HLS designs into large design spaces using various optimization directives across multiple vendor tools, 2) a design synthesis stage to execute HLS and FPGA tool flows concurrently across designs, and 3) a data aggregation stage for extracting standardized data into packaged datasets for ML usage. This tripartite architecture not only ensures broad coverage of data points via design space expansion but also supports interoperability with tools from multiple vendors. Users can contribute to each stage easily by submitting their own HLS designs or synthesis results via provided user APIs. The framework is also flexible, allowing extensions at every step via user APIs with custom frontends, synthesis tools, and scripts. To demonstrate the framework functionality, we include an initial set of built-in base designs from PolyBench, MachSuite, Rosetta, CHStone, Kastner et al.'s Parallel Programming for FPGAs, and curated kernels from existing open-source HLS designs. We report the statistical analyses and design space

关键词： Solid modeling Statistical analysis Buildings Machine learning Benchmark testing Software Space exploration Sustainable development Field programmable gate arrays Optimization

来源：评论

学校读者我要写书评

暂无评论

The Efficiency of Background data visualization of College Smart Education System Integrating Social Management Information Platform

The Efficiency of Background Data Visualization of College S...

引用

2022 International Conference on Edge Computing and Applications, ICECAA 2022

作者： Wang, Hailing Lijiang College Guangxi Normal University Guangxi Guilin541006 China

ISBN: (数字)9781665482325

ISBN: (纸本)9781665482325

Through the analysis of a large amount of data in the smart campus and related information on the Internet, educational resources and modern information technology are combined to form a networked, personalized and intelligent campus. The daily management of the campus, teacher and student activities, education and teaching and other campus management behaviors have been upgraded to a new level, reflecting the important influence and role of the visual management information system on the management and learning life of smart campuses. Using the Internet and big data, put forward corresponding countermeasures, create a new platform for campus construction, realize the visual management of smart campus, and improve the efficiency of resource utilization on campus. © 2022 ieee.

关键词： data visualization

来源：评论

学校读者我要写书评

暂无评论

Collection-CAM: A Faster Region-Based Saliency Method Using Collection-Wise Mask Over Pyramidal Features

引用

ieee ACCESS 2022年 10卷 112776-112788页

作者： Ha, Yungi Youn, Chan-Hyun Korea Adv Inst Sci & Technol Sch Elect Engn Daejeon 34141 South Korea

Due to the black-box nature of deep networks, making explanations of their decision-making is extremely challenging. A solution is using post-hoc attention mechanisms with the deep network to verify the decision basis. However, those methods have problems such as gradient noise and false confidence. In addition, existing saliency methods either have limited performance by using only the last convolution layer or suffer from large computational overhead. In this work, we propose the Collection-CAM, which generates an attention map with low computational overhead while utilizing multi-level feature maps. First, the Collection-CAM searches for the most appropriate form of the partition through bottom-up clustering and clustering validation process. Then the Collection-CAM applies different pre-processing procedures on the shallow feature map and final feature map to overcome the false positiveness when applied without distinction. By combining collection-wise masks according to their contribution to the confidence score, the Collection-CAM completes the attention map generation process. Experimental results on ImageNet1k, UC Merced, and CUB dataset and various deep network models demonstrate that the Collection-CAM not only can synthesize a saliency map with a better visual explanation but also requires significantly lower computational overhead compared to those of region-based saliency methods.

关键词： visualization Mathematical models data models Backpropagation Time measurement Deep learning Spatial resolution Clustering methods Visual explanation deep learning acceleration clustering analysis

来源：评论

学校读者我要写书评

暂无评论

Deep Object Tracking With Shrinkage Loss

引用

ieee TRANSACTIONS ON PATTERN analysis AND MACHINE INTELLIGENCE 2022年第5期44卷 2386-2401页

作者： Lu, Xiankai Ma, Chao Shen, Jianbing Yang, Xiaokang Reid, Ian Yang, Ming-Hsuan Shanghai Jiao Tong Univ AI Inst MoE Key Lab Artificial Intelligence Shanghai 200240 Peoples R China Shandong Univ Sch Software Jinan 250100 Peoples R China Incept Inst Artificial Intelligence Abu Dhabi 5151 U Arab Emirates Univ Adelaide Sch Comp Sci Adelaide SA 5005 Australia Univ Calif Merced Sch Elect Engn & Comp Sci Merced CA 95343 USA

In this paper, we address the issue of data imbalance in learning deep models for visual object tracking. Although it is well known that data distribution plays a crucial role in learning and inference models, considerably less attention has been paid to data imbalance in visual tracking. For the deep regression trackers that directly learn a dense mapping from input images of target objects to soft response maps, we identify their performance is limited by the extremely imbalanced pixel-to-pixel differences when computing regression loss. This prevents existing end-to-end learnable deep regression trackers from performing as well as discriminative correlation filters (DCFs) trackers. For the deep classification trackers that draw positive and negative samples to learn discriminative classifiers, there exists heavy class imbalance due to a limited number of positive samples when compared to the number of negative samples. To balance training data, we propose a novel shrinkage loss to penalize the importance of easy training data mostly coming from the background, which facilitates both deep regression and classification trackers to better distinguish target objects from the background. We extensively validate the proposed shrinkage loss function on six benchmark datasets, including the OTB-2013, OTB-2015, UAV-123, VOT-2016, VOT-2018 and LaSOT. Equipped with our shrinkage loss, the proposed one-stage deep regression tracker achieves favorable results against state-of-the-art methods, especially in comparison with DCFs trackers. Meanwhile, our shrinkage loss generalizes well to deep classification trackers. When replacing the original binary cross entropy loss with our shrinkage loss, three representative baseline trackers achieve large performance gains, even setting new state-of-the-art results.

关键词： Target tracking visualization Training Benchmark testing Object tracking data models Correlation data imbalance shrinkage loss regression tracking classification tracking Siamese tracking

来源：评论

学校读者我要写书评

暂无评论

Towards Multimodal Exploratory data analysis: SoniScope as a Prototypical Implementation

Towards Multimodal Exploratory Data Analysis: SoniScope as a...

引用

24th Eurographics Conference on visualization, EuroVis 2022

作者： Enge, K. Rind, A. Iber, M. Höldrich, R. Aigner, W. St. Pölten University of Applied Sciences Austria University of Music and Performing Arts Graz Austria

ISBN: (纸本)9783038681847

The metaphor of auscultating with a stethoscope can be an inspiration to combine visualization and sonification for exploratory data analysis. This paper presents SoniScope, a multimodal approach and its prototypical implementation based on this metaphor. It combines a scatterplot with an interactive parameter mapping sonification, thereby conveying additional information about items that were selected with a visual lens. SoniScope explores several design options for the shape of its lens and the sorting of the selected items for subsequent sonification. Furthermore, the open-source prototype serves as a blueprint framework for how to combine *** visualization and SuperCollider sonification in the Jupyter notebook environment. © 2022 The Author(s) © 2022 The Eurographics Association.

关键词： visualization

来源：评论

学校读者我要写书评

暂无评论

analysis Model for Decentralized Lending Protocols 22

Analysis Model for Decentralized Lending Protocols

引用

11th International symposium on Information and Communication Technology, SoICT 2022

作者： Pham, Viet-Bang Trinh, Tuan-Dat Hanoi University of Science and Technology Viet Nam

ISBN: (纸本)9781450397254

Decentralized Finance (DeFi)-An ecosystem of financial applications built on top of blockchain technology-has kept powering ahead and attracted a large amount of capital in just a few years. As the core part of the DeFi ecosystem, decentralized lending protocols allow users to become lenders or borrowers in a completely decentralized fashion, such that an individual has complete control over their funds at all times. Collecting, analyzing, and visualizing lending data are crucial as these lending protocols are growing steadily. This paper presents an analysis model for decentralized lending protocols. Our main contributions are as follows. (i) We analyze and compare the top lending protocols;(ii) We present a general model to collect and process a large amount of data from multiple lending protocols in an exact and efficient manner;(iii) We design and implement a lending visualization system with various charts divided into three categories, i.e., lending market overview, token analysis, and user analysis. © 2022 ACM.

关键词： Finance

来源：评论

学校读者我要写书评

暂无评论

Dimensionality Explorer for Single-Cell analysis

Dimensionality Explorer for Single-Cell Analysis

引用

Pacific (formerly Asia-Pacific APVIS) visualization symposium

作者： Haejin Jeong Hyoung-oh Jeong Semin Lee Won-Ki Jeong Korea University Ulsan National Institute of Science and Technology

Single-cell RNA sequencing (scRNA-seq) is becoming popular in studying the gene expression of cells at the single-cell level. ScRNA-seq enables analysts to characterize cell types, thereby providing a better understanding of dynamic biological processes. In scRNA-seq data analysis, principal component analysis (PCA) is commonly used to reduce at least thousands of dimensions in the raw data to a manageable size so that analysts can visualize and cluster cells to identify different cell types. The conventional process to determine the optimal dimensionality includes a laborious manual review of hundreds of different projection plots. To address this problem, we introduce a dimensionality explorer for single-cell analysis, which is a visualization system that helps analysts to effectively determine the optimal dimensionality of scRNA-seq data. It employs a hull heatmap, which provides a holistic view of overlaps among multiple cell types across various dimensionalities using a convex hull-embedded color map. The hull heatmap effectively reduces the burden of manually reviewing hundreds of projection plots to determine the optimal dimensionality. Our system also provides interactive gene expression level visualization and intuitive lasso selection, thereby allowing analysts to progressively refine the convex hulls of the hull heatmap. We demonstrate the usefulness of the proposed system through a user study and three case studies conducted by domain experts.

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：