检索结果-内蒙古大学图书馆

Negative Latency in computer Vision: A Key to Efficient Edge Offloading

学校读者我要写书评

暂无评论

Negative Latency in Computer Vision: A Key to Efficient Edge...

IEEE Conference on Global Communications (GLOBECOM)

作者： Jonas Schulz Susu Hu Stefanie Speidel Patrick Seeling Frank H. P. Fitzek Deutsche Telekom Chair of Communication Networks TU Dresden Germany Department of Translational Surgical Oncology NCT/UCC Dresden Faculty of Medicine University Hospital Carl Gustav Carus TUD Dresden University of Technology Dresden Germany Centre for Tactile Internet With Human-in-the-Loop (CeTI) Department of Computer Science Central Michigan University USA Joint Project 6G-Life

ISBN: (数字)9798350351255

ISBN: (纸本)9798350351262

Object recognition tasks are commonplace in industrial and medical applications, but resource-limited end devices common to the Internet of Things (IoT) are limited in capabilities and commonly require offloading of compute-intensive tasks. We evaluate the negative latency concept as a new approach to image detection and object recognition applications in low-bandwidth scenarios without a feedback channel. We employ image quality assessment and risk estimators associated with image quality degradation for progressively transmitted images to realize negative latency. Our comprehensive evaluation on image recognition and object detection tasks shows that image quality assessment over progressively transmitted images can be enabled by using risk estimators associated with image quality degradation during progressive transmission. In turn, our approach provides the framework for service latency and resource use reductions.

关键词： Image quality Measurement Degradation computer vision Computational modeling Transform coding Object detection Encoding Internet of Things Object recognition

Deep Learning: A Critical Analysis of its Effects on Organizational Performance

学校读者我要写书评

暂无评论

Deep Learning: A Critical Analysis of its Effects on Organiz...

International Conference on System Modeling & Advancement in Research Trends (SMART)

作者： Melanie Lourens Kamal Kishor Pandey Alok Upadhyay Mallika Tewari Shiv Kant Tiwari Surendra Kumar Shukla Department of Human Resources Management Durban University of Technology Berea Durban South Africa TMIMT-College of Management Teerthanker Mahaveer University Moradabad Uttar Pradesh School of Management Graphic Era Hill University Uttarakhand Institute of Business Management GLA University Mathura Uttar Pradesh Department of Computer Science & Engineering Graphic Era Deemed to be University Dehradun Uttarakhand India

ISBN: (纸本)9781665487351

With an emphasis on the enterprises' net wealth as learnt in the school changes, our study's primary goal is to assess the idea of learning algorithms and how it influences performance. The following steps have to be taken in order to finish the study. Following are some characteristics of supervised learning: Machine learning: What's it? Where does it function? What methods are employed? What are the problems and downfalls? What effects may reinforce learning have on the effectiveness of your organization? vii) Deep learning examples; and (vi) neural network training. Information technology skills have been used in our research to better determine how DL value proposition impacts organization performance. The technique of investigation (giving guidance based on research findings, responding to the research question, participating in discussions, finally developing and analyzing, and making recommendations). It incorporates several technological advancements, including chatbots, self-learning robots, and machine learning. All of these developments have the potential to improve people's comprehension of and responses to their surroundings. The process of reacting to or disrupting their environments while aiding in the development and expansion of competitive and strategic assets has been the driving force behind the implementation of artificial intelligence and machine understanding scientific developments by organizations. DL outperforms the competitor when it comes to enhancing the efficacy of present processes and enhancing the impact of automation, economic, and innovative advances because of its capacity to detect, predict, and engage with humans.

关键词： Deep learning Training Productivity Machine learning algorithms Supply chains Supervised learning Neural networks

New Interfaces for Musical Expression

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Poupyrev, Ivan Lyons, Michael J. Fels, Sidney Bean, Tina Blaine Interaction Lab Sony CSL 3-14-13 Higashi-Gotanda Tokyo141-0022 Japan ATR MIC 2-2 Hikafidai Seika Souraku-gun Kyoto619-02 Japan Department of ECE UBC VancouverBCV6T 1Z4 Canada School of Computer Science Carnegie Mellon University 5000 Forbes Avenue PittsburghPA15213 United States

The rapid evolution of electronics, digital media, advanced materials, and other areas of technology, is opening up unprecedented opportunities for musical interface inventors and designers. The possibilities afforded by these new technologies carry with them the challenges of a complex and often confusing array of choices for musical composers and performers. New musical technologies are at least partly responsible for the current explosion of new musical forms, some of which are controversial and challenge traditional definitions of music. Alternative musical controllers, currently the leading edge of the ongoing dialogue between technology and musical culture, involve many of the issues covered at past CHI meetings. This workshop brings together interface experts interested in musical controllers and musicians and composers involved in the development of new musical interfaces. © 2020, CC-BY.

关键词： Controllers

An Explainable Deep Neural Network Model Based on Multi-Granularity Co-Attention to Improve Prediction of Drug-Target Binding Affinity

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Lin, Yanmei Cao, Aiqing Deng, Lei Liao, Mingzhi Zhang, Ru Zhang, Hao Peng, Yuzhong Guangxi Key Lab of Human-machine Interaction and Intelligent Decision Nanning Normal University Nanning530001 China Guangxi Academy of Sciences Nanning530007 China School of Computer Science and Engineering Central South University Changsha410083 China Center of Bioinformatics College of Life Sciences Northwest A&F University Shaanxi Yangling712100 China Shenzhen Institute of Advanced Technology Chinese Academy of Sciences Shenzhen518000 China Center for Applied Mathematics of Guangxi Nanning Normal University Nanning530100 China

The prediction of drug-target interactions is pivotal in modern new drug development and drug repositioning. Nonetheless, current methods for predicting drug-target interactions often lose essential molecular structure information during the modeling process, leading to low prediction accuracy. This work proposes rMolGIN, a novel graph neural network with residual connectivity for molecular feature learning. rMolGIN not only takes into account the differences in chemical bonding properties and atomic neighborhood interactions, but also captures global and local structural information of drug molecules. Furthermore, we propose a novel drug-target affinity model based on rMolGIN and deep neural networks, coupled with a multi-granularity co-attention mechanism called HMCoDTA to capture comprehensive structural insights from both drugs and proteins. The proposed innovative model allows for comprehensive insight into the structural relationships between drugs and proteins, facilitating a more accurate prediction of binding affinities. To validate the effectiveness of our proposed methods, we conducted extensive experiments on three benchmark datasets of drug-target binding affinity, comparing HMCoDTA against 19 state-of-the-art models. Notably, the results show that HMCoDTA outperforms existing methods, demonstrating the effectiveness and advantages of our model in predicting drug-target binding affinity. © 2024, The Authors. All rights reserved.

关键词： Prediction models

Comparative Methodological Approaches for Analyzing Margin of Stability During Normal Walking

学校读者我要写书评

暂无评论

Comparative Methodological Approaches for Analyzing Margin o...

2024 International Symposium on 3D Analysis of human Movement, 3DAHM 2024

作者： Sganga, Magali Villagra, Federico Akanyeti, Otar Ravera, Emiliano Pablo Universidad Maimónides Universidad Nacional de La Matanza Laboratorio de Investigación del Movimiento Humano Departamento de Ciencias de la Salud Buenos Aires Argentina Penglais Campus Aberystwyth University Department of Life Sciences Carwyn James Building Wales AberystwythSY233DF United Kingdom Aberystwyth University Department of Computer Science Wales AberystwythSY23 3DB United Kingdom National University of Entre Rios Group of Analysis Modeling Processing and Clinician Implementation of Biomechanical Signals and Systems IBB CONICET-UNER Human Movement Research Laboratory Oro Verde Argentina

ISBN: (纸本)9798331521738

A comprehensive evaluation of pathological gait patterns is essential for understanding the underlying pathology and its progression to deliver personalized treatments. One quantitative indicator for assessing fall risks during walking is margin of stability (MoS), a measure of instantaneous mechanical stability derived from the dynamic relationship between the extrapolated center of mass (CoM) and the center of pressure. However, extrapolating CoM during movement is not straightforward, and there is an undesired variability induced by the selected measurement protocol, which often depends on the available technology. Up until now, the extent of this variability has not been investigated systematically. This study compares four different CoM estimation methods. Based on the Conventional Gait Model, the first two methods estimate CoM using virtual sacrum (1) and pelvis centroid (2). The third method estimates CoM from ground reaction force (3), and the last method uses the accelerometer and gyroscope data from a smartphone mounted on the lower back. 21 healthy subjects (10 female, 11 male, age: 31.7 ± 9.0 years;BMI: 24.7 ± 3.5 kg m-2) walked on an instrumented treadmill at a fixed speed at 1.3 m s -1 for 3 minutes. Descriptive statistical analysis was performed, including temporal and normalized gait cycle analysis, focusing on the minimum and maximum MoS, and its ranges during walking. SPMld with paired t-tests was employed for the analysis. Our results suggest that CoM estimation varies significantly depending on the selected measurement protocol and technology which should be considered while reporting and interpreting MoS results. © 2024 IEEE.

关键词： Extrapolation

Semi-supervised medical image classification with relation-driven self-ensembling model

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Liu, Quande Yu, Lequan Luo, Luyang Dou, Qi Heng, Pheng Ann Department of Computer Science and Engineering Chinese University of Hong Kong Department of Radiation Oncology Stanford University StanfordCA94305 United States T Stone Robotics Institute Chinese University of Hong Kong Guangdong-Hong Kong-Macao Joint Laboratory of Human-Machine Intelligence-Synergy Systems Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences Shenzhen518055 China

Training deep neural networks usually requires a large amount of labeled data to obtain good performance. However, in medical image analysis, obtaining high-quality labels for the data is laborious and expensive, as accurately annotating medical images demands expertise knowledge of the clinicians. In this paper, we present a novel relation-driven semi-supervised framework for medical image classification. It is a consistency-based method which exploits the unlabeled data by encouraging the prediction consistency of given input under perturbations, and leverages a self-ensembling model to produce high-quality consistency targets for the unlabeled data. Considering that human diagnosis often refers to previous analogous cases to make reliable decisions, we introduce a novel sample relation consistency (SRC) paradigm to effectively exploit unlabeled data by modeling the relationship information among different samples. Superior to existing consistency-based methods which simply enforce consistency of individual predictions, our framework explicitly enforces the consistency of semantic relation among different samples under perturbations, encouraging the model to explore extra semantic information from unlabeled data. We have conducted extensive experiments to evaluate our method on two public benchmark medical image classification datasets, i.e., skin lesion diagnosis with ISIC 2018 challenge and thorax disease classification with ChestX-ray14. Our method outperforms many state-of-the-art semi-supervised learning methods on both single-label and multi-label image classification scenarios. Copyright © 2020, The Authors. All rights reserved.

关键词： Image classification

One to many: Adaptive instrument segmentation via meta learning and dynamic online adaptation in robotic surgical video

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhao, Zixu Jin, Yueming Lu, Bo Ng, Chi-Fai Dou, Qi Liu, Yun-Hui Heng, Pheng-Ann The Department of Computer Science and Engineering The Chinese University of Hong Kong Hong Kong The T Stone Robotics Institute The Department of Mechanical and Automation Engineering The Chinese University of Hong Kong Hong Kong Guangdong-Hong Kong-Macao Joint Laboratory of Human-Machine Intelligence-Synergy Systems Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences China The Department of Surgery The Chinese University of Hong Kong Hong Kong

Surgical instrument segmentation in robot-assisted surgery (RAS) - especially that using learning-based models - relies on the assumption that training and testing videos are sampled from the same domain. However, it is impractical and expensive to collect and annotate sufficient data from every new domain. To greatly increase the label efficiency, we explore a new problem, i.e., adaptive instrument segmentation, which is to effectively adapt one source model to new robotic surgical videos from multiple target domains, only given the annotated instruments in the first frame. We propose MDAL, a meta-learning based dynamic online adaptive learning scheme with a two-stage framework to fast adapt the model parameters on the first frame and partial subsequent frames while predicting the results. MDAL learns the general knowledge of instruments and the fast adaptation ability through the video-specific meta-learning paradigm. The added gradient gate excludes the noisy supervision from pseudo masks for dynamic online adaptation on target videos. We demonstrate empirically that MDAL outperforms other state-of-the-art methods on two datasets (including a real-world RAS dataset). The promising performance on ex-vivo scenes also benefits the downstream tasks such as robot-assisted suturing and camera control. Copyright © 2021, The Authors. All rights reserved.

关键词： Surgical equipment

An Empirical Examination of Balancing Strategy for Counterfactual Estimation on Time Series

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Huang, Qiang Meng, Chuizheng Cao, Defu Huang, Biwei Chang, Yi Liu, Yan School of Artificial Intelligence Jilin University Jilin Changchun China International Center of Future Science Jilin University Jilin Changchun China Department of Computer Science University of Southern California California Los Angeles United States Halicioğlu Data Science Institute University of California San Diego San DiegoCA United States Engineering Research Center of Knowledge-Driven Human-Machine Intelligence MOE Jilin Changchun China

Counterfactual estimation from observations represents a critical endeavor in numerous application fields, such as healthcare and finance, with the primary challenge being the mitigation of treatment bias. The balancing strategy aimed at reducing covariate disparities between different treatment groups serves as a universal solution. However, when it comes to the time series data, the effectiveness of balancing strategies remains an open question, with a thorough analysis of the robustness and applicability of balancing strategies still lacking. This paper revisits counterfactual estimation in the temporal setting and provides a brief overview of recent advancements in balancing strategies. More importantly, we conduct a critical empirical examination for the effectiveness of the balancing strategies within the realm of temporal counterfactual estimation in various settings on multiple datasets. Our findings could be of significant interest to researchers and practitioners and call for a reexamination of the balancing strategy in time series settings. Copyright © 2024, The Authors. All rights reserved.

关键词： Time series analysis

Multimodal deep learning framework for image popularity prediction on social media

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Abousaleh, Fatma S. Cheng, Wen-Huang Yu, Neng-Hao Tsao, Yu Social Networks and Human-Centered Computing Program Taiwan International Graduate Program Institute of Information Science Research Center for Information Technology Innovation Academia Sinica Taipei11529 Taiwan Department of Computer Science National Chengchi University Taipei11605 Taiwan The Institute of Electronics Department of Electronics Engineering National Chiao Tung University Hsinchu30010 Taiwan The College of Design Department of Design National Taiwan University of Science and Technology Taipei10607 Taiwan The Research Center for Information Technology Innovation Academia Sinica Taipei11529 Taiwan

Billions of photos are uploaded to the web daily through various types of social networks. Some of these images receive millions of views and become popular, whereas others remain completely unnoticed. This raises the problem of predicting image popularity on social media. The popularity of an image can be affected by several factors, such as visual content, aesthetic quality, user, post metadata, and time. Thus, considering all these factors is essential for accurately predicting image popularity. In addition, the efficiency of the predictive model also plays a crucial role. In this study, motivated by multimodal learning, which uses information from various modalities, and the current success of convolutional neural networks (CNNs) in various fields, we propose a deep learning model, called visual-social convolutional neural network (VSCNN), which predicts the popularity of a posted image by incorporating various types of visual and social features into a unified network model. VSCNN first learns to extract high-level representations from the input visual and social features by utilizing two individual CNNs. The outputs of these two networks are then fused into a joint network to estimate the popularity score in the output layer. We assess the performance of the proposed method by conducting extensive experiments on a dataset of approximately 432K images posted on Flickr. The simulation results demonstrate that the proposed VSCNN model significantly outperforms state-of-the-art models, with a relative improvement of greater than 2.33%, 7.59%, and 14.16% in terms of Spearman’s Rho, mean absolute error, and mean squared error, respectively. Copyright © 2021, The Authors. All rights reserved.

关键词： Social networking (online)