检索结果-内蒙古大学图书馆

Discrepancy Matters: Learning from Inconsistent Decoder Features for Consistent Semi-supervised Medical Image Segmentation

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zeng, Qingjie Xie, Yutong Lu, Zilin Lu, Mengkang Xia, Yong The National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology School of Computer Science and Engineering Northwestern Polytechnical University Xi’an710072 China The Australian Institute for Machine Learning The University of Adelaide AdelaideSA5000 Australia

Semi-supervised learning (SSL) has been proven beneficial for mitigating the issue of limited labeled data especially on the task of volumetric medical image segmentation. Unlike previous SSL methods which focus on exploring highly confident pseudo-labels or developing consistency regularization schemes, our empirical findings suggest that inconsistent decoder features emerge naturally when two decoders strive to generate consistent predictions. Based on the observation, we first analyze the treasure of discrepancy in learning towards consistency, under both pseudo-labeling and consistency regularization settings, and subsequently propose a novel SSL method called LeFeD, which learns the feature-level discrepancy obtained from two decoders, by feeding the discrepancy as a feedback signal to the encoder. The core design of LeFeD is to enlarge the difference by training differentiated decoders, and then learn from the inconsistent information iteratively. We evaluate LeFeD against eight state-of-the-art (SOTA) methods on three public datasets. Experiments show LeFeD surpasses competitors without any bells and whistles such as uncertainty estimation and strong constraints, as well as setting a new state-of-the-art for semi-supervised medical image segmentation. Code is available at https://***/maxwell0027/LeFeD © 2023, CC BY.

关键词： Decoding

Modeling Annotator Preference and Stochastic Annotation Error for Medical Image Segmentation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Liao, Zehui Hu, Shishuai Xie, Yutong Xia, Yong National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology School of Computer Science and Engineering Northwestern Polytechnical University Xi’an710072 China Australian Institute for Machine Learning The University of Adelaide AdelaideSA5000 Australia

Manual annotation of medical images is highly subjective, leading to inevitable and huge annotation biases. Deep learning models may surpass human performance on a variety of tasks, but they may also mimic or amplify these biases. Although we can have multiple annotators and fuse their annotations to reduce stochastic errors, we cannot use this strategy to handle the bias caused by annotators’ preferences. In this paper, we highlight the issue of annotator-related biases on medical image segmentation tasks, and propose a Preference-involved Annotation Distribution Learning (PADL) framework to address it from the perspective of disentangling an annotator’s preference from stochastic errors using distribution learning so as to produce not only a meta segmentation but also the segmentation possibly made by each annotator. Under this framework, a stochastic error modeling (SEM) module estimates the meta segmentation map and average stochastic error map, and a series of human preference modeling (HPM) modules estimate each annotator’s segmentation and the corresponding stochastic error. We evaluated our PADL framework on two medical image benchmarks with different imaging modalities, which have been annotated by multiple medical professionals, and achieved promising performance on all five medical image segmentation tasks. © 2021, CC BY-NC-ND.

关键词： Stochastic systems

Momentum Recursive Darts

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Ma, Benteng Xia, Yong Zhang, Yanning National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology School of Computer Science and Engineering Northwestern Polytechnical University Xi’an710072 China Research and Development Institute of Northwestern Polytechnical University in Shenzhen Shenzhen518057 China

DARTS has emerged as a popular method for neural architecture search (NAS) owing to its efficiency and simplicity. It employs gradient-based bi-level optimization to iteratively optimize the upper-level architecture parameters and lower-level super-network weights. One of the key challenges in DARTS is the accurate estimation of gradients for both upper-level and lower-level object functions, which existing methods struggle with, leading to significant errors in gradient approximation. To address this issue, we propose a new approach called MR-DARTS that incorporates a momentum term and a recursive scheme to improve gradient estimation. Specifically, MR-DARTS leverages historical information by using a running average of past observed gradients to enhance the quality of current gradient estimation in both upper-level and lower-level functions. Our theoretical analysis shows that the variance of our estimated gradient decreases with each iteration. By utilizing momentum and a recursive scheme, MR-DARTS effectively controls the error in stochastic gradient updates that results from inaccurate gradient estimation. Furthermore, we utilize the Neumann series approximation and Hessian Vector Product scheme to reduce computational requirements and memory usage. We evaluate our proposed method on several search spaces, including NAS Bench-201 and four variants of the DARTS search space, and demonstrate its effectiveness through comprehensive experiments. © 2023, The Authors. All rights reserved.

关键词： Image recognition

A review of progress in coupled ocean-atmosphere model developments for ENSO studies in China

学校读者我要写书评

暂无评论

Journal of Oceanology and Limnology 2020年第4期38卷 930-961页

作者： ZHANG Rong-Hua YU Yongqiang SONG Zhenya REN Hong-Li TANG Youmin QIAO Fangli WU Tongwen GAO Chuan HU Junya TIAN Feng ZHU Yuchao CHEN Lin LIU Hailong LIN Pengfei WU Fanghua WANG Lin Key Laboratory of Ocean Circulation and Waves Institute of Oceanologyand Center for Ocean Mega-ScienceChinese Academy of SciencesQingdao 266071China Laboratory for Ocean and Climate Dynamics Pilot National Laboratory for Marine Science and TechnologyQingdao 266237China Center for Excellence in Quaternary Science and Global Change Chinese Academy of SciencesXi’an 710061China University of Chinese Academy of Sciences Beijing 100049China LASG Institute of Atmospheric PhysicsChinese Academy of SciencesBeijing 100049China First Institute of Oceanography Ministry of Natural ResourcesQingdao 266061China Laboratory for Regional Oceanography and Numerical Modeling Pilot National Laboratory for Marine Science and TechnologyQingdao 266237China Key Laboratory of Marine Science and Numerical Modeling(MASNUM) Ministry of Natural ResourcesQingdao 266061China National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology Qingdao 266061China State Key Laboratory of Severe Weather Chinese Academy of Meteorological SciencesBeijing 100081China Laboratory for Climate Studies and CMA-NJU Joint Laboratory for Climate Prediction Studies National Climate CenterChina Meteorological AdministrationBeijing 100081China Environmental Science and Engineering University of Northern British ColumbiaBritish Columbia V2N 4Z9Canada State Key Laboratory of Satellite Ocean Environment Dynamics Second Institute of OceanographyMinistry of Natural ResourcesHangzhou 310012China Key Laboratory of Meteorological Disaster Ministry of Education(KLME)and Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters(CIC-FEMD)Nanjing University of Information Science and TechnologyNanjing 210044China Beijing Climate Center China Meteorological AdministrationBeijing 100081China

El Nino-Southern Oscillation(ENSO) is the strongest interannual signal that is producedby basinscale processes in the tropical Pacific,with significant effects on weather and climate *** the past,extensive and intensive international efforts have been devoted to coupled model developments for ENSO studies.A hierarchy of coupled ocean-atmo sphere models has been formulated;in terms of their complexity,they can be categorized into intermediate coupled models(ICMs),hybrid coupled models(HCMs),and fully coupled general circulation models(CGCMs).ENSO modeling has made significant progress over the past decades,reaching a stage where coupled models can now be used to successfully predict ENSO events 6 months to one year in ***,ENSO exhibits great diversity and complexity as observed in nature,which still cannot be adequately captured by current state-of-the-art coupled models,presenting a challenge to ENSO *** primarily reviewed the long-term efforts in ENSO modeling continually and steadily made at different institutions in China;some selected representative examples are presented here to review the current status of ENSO model developments and applications,which have been actively pursued with noticeable progress being made *** ENSO simulations are very sensitive to model formulations and process representations etc.,dedicated efforts have been devoted to ENSO model developments and ***,different ocean-atmosphere coupled models have been available in China,which exhibit good model performances and have already had a variety of applications to climate modeling,including the Coupled Model Intercomparison Project Phase 6(CMIP6).Nevertheless,large biases and uncertainties still exist in ENSO simulations and predictions,and there are clear rooms for their improvements,which are still an active area of researches and ***,model performances of ENSO simulations are assessed in terms of advantages and disadvantage

关键词： El Niño-Southern Oscillation(ENSO) coupled ocean-atmosphere models simulations and predictions model biases and uncertainties

An Optimized Variational Bayesian Point Set Matching Approach using Coordinate Ascent and Simulated Annealing

学校读者我要写书评

暂无评论

An Optimized Variational Bayesian Point Set Matching Approac...

WRC Symposium on Advanced Robotics and Automation (WRC SARA)

作者： Fangjun Luan Shitong Liu Shuai Yuan Shenyang Branch of National Special Computer Engineering Technology Research Center Shenyang China School of Computer Science and Engineering Shenyang Jianzhu University Shenyang China Liaoning Province Big Data Management and Analysis Laboratory of Urban Construction Shenyang China

ISBN: (数字)9798331506100

ISBN: (纸本)9798331506117

Aiming at the interference problem of outliers in point set matching, we propose an optimized variational Bayesian point set matching algorithm using coordinate ascent and simulated annealing. For the complex terms in the objective function, we introduce the coordinate ascending strategy and use the Monte Carlo integral stochastic approximation and control variable method to directly optimize the parameters. To enhance the generalization ability of the algorithm and avoid falling into the local optimal solution, we use the simulated annealing scheme to optimize the spatial mapping of the point set. Finally, the affine transform point set matching experiment and the non-rigid point set matching experiment represent our algorithm have a good matching effect and high accuracy.

关键词： Monte Carlo methods Annealing Integral equations Simulated annealing Transforms Interference Approximation algorithms

Learning synergistic attention for light field salient object detection

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhang, Yi Chen, Geng Chen, Qian Sun, YuJia Xia, Yong Deforges, Olivier Hamidouche, Wassim Zhang, Lu France National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology School of Computer Science and Engineering Northwestern Polytechnical University China School of Information Science and Technology University of Science and Technology of China China School of Computer Science Inner Mongolia University China

In this work, we propose Synergistic Attention Network (SA-Net) to address the light field salient object detection by establishing a synergistic effect between multi-modal features with advanced attention mechanisms. Our SA-Net exploits the rich information of focal stacks via 3D convolutional neural networks, decodes the high-level features of multi-modal light field data with two cascaded synergistic attention modules, and predicts the saliency map using an effective feature fusion module in a progressive manner. Extensive experiments on three widely-used benchmark datasets show that our SA-Net outperforms 28 state-of-the-art models, sufficiently demonstrating its effectiveness and superiority. Our code is available at https://***/PanoAsh/ SA-Net. Copyright © 2021, The Authors. All rights reserved.

关键词： Convolutional neural networks

Research on Equipment State Correlation analysis Based on Digital Model of High-speed Railway Public Works Equipment

学校读者我要写书评

暂无评论

Research on Equipment State Correlation Analysis Based on Di...

Digital Twins and Parallel Intelligence (DTPI), IEEE International Conference on

作者： Yanhua Wu Ping Li Zhibo Cheng Wenya Xu Chenkun Jin Hairong Dong National Engineering Laboratory of Comprehensive Transportation Big Data Application Technology China Academy of Railway Sciences Corporation Limited Beijing China China Academy of Railway Sciences Corporation Limited Beijing China Beijing Jiaotong University Beijing China

One of the key factors in the safe operation of high-speed railways is the operational status of public works equipment. Constructing the digital model of public works equipment, fusing basic attribute information with detection and monitoring status information, and characterizing the current operating status of public works equipment, will provide more intuitive support for the change law of state change analysis. The correlation analysis between TQI and settlement based on digital model will provide auxiliary decision support for equipment maintenance.

关键词： Analytical models Correlation Digital twin Conferences Maintenance engineering Rail transportation Monitoring

Multi-Temporal Relationship Inference in Urban Areas

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Li, Shuangli Zhou, Jingbo Liu, Ji Xu, Tong Chen, Enhong Xiong, Hui School of Computer Science and Technology University of Science and Technology of China Baidu Research China Business Intelligence Lab Baidu Research China Big Data Lab Baidu Research China School of Computer Science and Technology University of Science and Technology of China State Key Laboratory of Cognitive Intelligence China Anhui Province Key Laboratory of Big Data Analysis and Application University of Science and Technology of China State Key Laboratory of Cognitive Intelligence China The Department of Computer Science and Engineering The Hong Kong University of Science and Technology China

Finding multiple temporal relationships among locations can benefit a bunch of urban applications, such as dynamic offline advertising and smart public transport planning. While some efforts have been made on finding static relationships among locations, little attention is focused on studying time-aware location relationships. Indeed, abundant location-based human activities are time-varying and the availability of these data enables a new paradigm for understanding the dynamic relationships in a period among connective locations. To this end, we propose to study a new problem, namely multi-Temporal relationship inference among locations (Trial for short), where the major challenge is how to integrate dynamic and geographical influence under the relationship sparsity constraint. Specifically, we propose a solution to Trial with a graph learning scheme, which includes a spatially evolving graph neural network (SEENet) with two collaborative components: spatially evolving graph convolution module (SEConv) and spatially evolving self-supervised learning strategy (SE-SSL). SEConv performs the intra-time aggregation and inter-time propagation to capture the multifaceted spatially evolving contexts from the view of location message passing. In addition, SE-SSL designs time-aware self-supervised learning tasks in a global-local manner with additional evolving constraint to enhance the location representation © 2023, CC BY.

关键词： Location

Unsupervised deep domain adaptation for hyperspectral image classification

学校读者我要写书评

暂无评论

Unsupervised deep domain adaptation for hyperspectral image ...

IEEE International Symposium on Geoscience and Remote Sensing (IGARSS)

作者： Wei Li Wei Wei Lei Zhang Cong Wang Yanning Zhang School of Computer Science Northwestern Polytechnical University Xi'an Shanxi China Inception Institute of Artificial Intelligence (IIAI) UAE The National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology Xi'an Shanxi China

Deep neural networks have been proven to be a promising way for hyperspectral image (HSI) classification. Their success depends on a premise that source domain (i.e., training) and target domain (i.e., test) samples are identically distributed. However, due to various imaging environments, in practice obvious distribution discrepancy often exists between these two domains, which can dramatically reduce the capacity of the classifier trained in source domain generalizing to target domain. To mitigate this problem, we present a novel deep unsupervised domain adaptation framework for HSI classification, which can simultaneously align the distributions of two domains and learn a classifier in source domain. Firstly, we employ two auto-encoder networks to separately project the samples from two domains into two low-dimensional feature spaces. Then, a multi-level maximum mean discrepancy (MMD) loss is imposed on the feature space to reduce the distribution discrepancy between two domains. Given the resultant features, a classification subnet is further learned to classify the labeled samples in source domain. Since the classifier is trained based on the domain-invariant features, it can well generalize to the target domain. Experimental results on one benchmark cross-domain HSI datasets prove the superior performance of the proposed method.

关键词：