26,473 Hits in 5.6 sec

Adaptive Label Smoothing To Regularize Large-Scale Graph Training [article]

Kaixiong Zhou, Ninghao Liu, Fan Yang, Zirui Liu, Rui Chen, Li Li, Soo-Hyun Choi, Xia Hu
2021 arXiv   pre-print
To handle large-scale graphs, most of the existing methods partition the input graph into multiple sub-graphs (e.g., through node clustering) and apply batch training to save memory cost.  ...  Specifically, ALS propagates node labels to aggregate the neighborhood label distribution in a pre-processing step, and then updates the optimal smoothed labels online to adapt to specific graph structure  ...  However, it is non-trivial to apply the label smoothing to regularize and adapt to the large-scale graph training from two structural levels: local node and global graph.  ... 
arXiv:2108.13555v1 fatcat:ujyhpronczgmxfi2rc3dw7srym

Interactive Multi-Label CNN Learning With Partial Labels

Dat Huynh, Ehsan Elhamifar
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
We introduce a new loss function that regularizes the cross-entropy loss with a cost function that measures the smoothness of labels and features of images on the data manifold.  ...  We address the problem of efficient end-to-end learning a multi-label Convolutional Neural Network (CNN) on training images with partial labels.  ...  While [30] learns an adaptive graph for label propagation, it cannot generalize to novel images due to its transductive nature and cannot scale to large datasets.  ... 
doi:10.1109/cvpr42600.2020.00944 dblp:conf/cvpr/HuynhE20b fatcat:ji733zx3uzdjjplk4vazuocjgu

Efficient Graph-Based Semi-Supervised Learning of Structured Tagging Models

Amarnag Subramanya, Slav Petrov, Fernando C. N. Pereira
2010 Conference on Empirical Methods in Natural Language Processing  
The similarity graph is used during training to smooth the state posteriors on the target domain. Standard inference can be used at test time.  ...  Our approach is able to scale to very large problems and yields significantly improved target domain accuracy.  ...  Here we use the graph as a smoothness regularizer to train CRFs in a semisupervised manner.  ... 
dblp:conf/emnlp/SubramanyaPP10 fatcat:imoqrbvfv5bk3nstevm2wvop4e

Multi-class regularization parameter learning for graph cut image segmentation

Sema Candemir, Kannappan Palaniappan, Yusuf Sinan Akgul
2013 2013 IEEE 10th International Symposium on Biomedical Imaging  
We demonstrate the performance of the approach within graph cut segmentation framework via qualitative results on chest x-rays.  ...  However, these algorithms depend on parameters which need to be tuned for a meaningful solution.  ...  The data term confines the segmentation labels to be close to the observed image. The smoothness term forces the algorithm to assign similar labels to the neighborhood pixels.  ... 
doi:10.1109/isbi.2013.6556813 dblp:conf/isbi/CandemirPA13 fatcat:buiwwsqp6re6pbfbs5p34uhrjq

Graph-Based Embedding Smoothing Network for Few-Shot Scene Classification of Remote Sensing Images

Zhengwu Yuan, Wendong Huang, Chan Tang, Aixia Yang, Xiaobo Luo
2022 Remote Sensing  
The most popular way to solve scene classification is to train a deep neural network with a large-scale remote sensing dataset.  ...  Specifically, GES-Net adopts an unsupervised non-parametric regularizer, called embedding smoothing, to regularize embedding features.  ...  In a wide range of realistic scenarios, collecting large-scale remote sensing images and labeling them are quite time-consuming and painstaking tasks.  ... 
doi:10.3390/rs14051161 fatcat:hvxzgzeg2bblvnpppdblgogd7y

Webly Supervised Image Classification with Self-Contained Confidence [article]

Jingkang Yang, Litong Feng, Weirong Chen, Xiaopeng Yan, Huabin Zheng, Ping Luo, Wayne Zhang
2020 arXiv   pre-print
A series of SCC-friendly regularization approaches are investigated, among which the proposed graph-enhanced mixup is the most effective method to provide high-quality confidence to enhance our framework  ...  The proposed WSL framework has achieved the state-of-the-art results on two large-scale WSL datasets, WebVision-1000 and Food101-N. Code is available at  ...  Introduction Large-scale human-labeled data plays a vital role in deep learning-based applications such as image classification [3] , scene recognition [41] , face recognition [30] , etc.  ... 
arXiv:2008.11894v1 fatcat:ntclraq52ba3dmsbguar2lnv4e

Graph Neural Network Based Attribute Auxiliary Structured Grouping for Person Re-Identification

Geyu Tang, Xingyu Gao, Zhenyu Chen, Huicai Zhong
2021 IEEE Access  
the hard one-hot label into "soft" label with smoothing regularization.  ...  Considering the "over-confidence" of inaccurate label may be harmful to the discriminative learning, we regularize the learning of the embeddig model with smoothed pseudo labels (SPL) when training with  ... 
doi:10.1109/access.2021.3069915 fatcat:xlukop5sdjfjbioycioaynzp6q

Robust and Scalable Graph-Based Semisupervised Learning

Wei Liu, Jun Wang, Shih-Fu Chang
2012 Proceedings of the IEEE  
Second, to support scalability to the gigantic scale (millions or billions of samples), recent solutions based on anchor graphs are reviewed.  ...  It has been shown effective in propagating a limited amount of initial labels to a large amount of unlabeled data, matching the needs of many emerging applications such as image annotation and information  ...  This web-scale experimental design corroborates that AGR can be well adapted to cope with web-scale data through training anchor graph models over million-scale data sets and then inductively applying  ... 
doi:10.1109/jproc.2012.2197809 fatcat:fk66s5zl75d35pn3rjwmhver7a

GraphHop: An Enhanced Label Propagation Method for Node Classification [article]

Tian Xie, Bin Wang, C.-C. Jay Kuo
2021 arXiv   pre-print
The LP method is not effective in modeling node attributes and labels jointly or facing a slow convergence rate on large-scale graphs. GraphHop is proposed to its shortcoming.  ...  This iterative procedure exploits the neighborhood information and enables GraphHop to perform well in an extremely small label rate setting and scale well for very large graphs.  ...  For GraphHop, we apply it to three large-scale graph datasets.  ... 
arXiv:2101.02326v1 fatcat:b35wqynunrbxjhs2hpuspxgoqi

LEReg: Empower Graph Neural Networks with Local Energy Regularization [article]

Xiaojun Ma, Hanyue Chen, Guojie Song
2022 arXiv   pre-print
With the proposed two regularization terms, GNNs are able to filter the most useful information adaptively, learn more robustly and gain higher expressiveness.  ...  Existing GNNs treat all parts of the graph uniformly, which makes it difficult to adaptively pass the most informative message for each unique part.  ...  It provides a simplified graph convolution framework for large-scale graph datasets with high speed.  ... 
arXiv:2203.10565v1 fatcat:uj5kgklmajdqpopvkl3dqyt72i

Domain adaptive semantic diffusion for large scale context-based video annotation

Yu-Gang Jiang, Jun Wang, Shih-Fu Chang, Chong-Wah Ngo
2009 2009 IEEE 12th International Conference on Computer Vision  
This paper proposes a novel and efficient approach, named domain adaptive semantic diffusion (DASD), to exploit semantic context while considering the domain-shift-ofcontext for large scale video concept  ...  The adaptation provides a means to handle domain change between training and test data, which occurs very often in video annotation task.  ...  For large scale video annotation which could involve simultaneous labeling of hundreds of concepts, the problem becomes worse when the unlabeled videos are from a domain different from that of the training  ... 
doi:10.1109/iccv.2009.5459295 dblp:conf/iccv/JiangWCN09 fatcat:vrhzwcuggzeqtcgfpixqpvbhpe

Graph Neural Networks With Lifting-based Adaptive Graph Wavelets [article]

Mingxing Xu, Wenrui Dai, Chenglin Li, Junni Zou, Hongkai Xiong, Pascal Frossard
2022 arXiv   pre-print
However, existing SGNNs are limited in implementing graph filters with rigid transforms (e.g., graph Fourier or predefined graph wavelet transforms) and cannot adapt to signals residing on graphs and tasks  ...  In this paper, we propose a novel class of graph neural networks that realizes graph filters with adaptive graph wavelets.  ...  Thus, they can not scale to large and varying-size graphs. C.  ... 
arXiv:2108.01660v3 fatcat:liunq2ozw5dxlps7s3wcc52vry

A Unified View on Graph Neural Networks as Graph Signal Denoising [article]

Yao Ma, Xiaorui Liu, Tong Zhao, Yozen Liu, Jiliang Tang, Neil Shah
2021 arXiv   pre-print
To demonstrate its promising potential, we instantiate a novel GNN model, ADA-UGNN, derived from UGNN, to handle graphs with adaptive smoothness across nodes.  ...  Graph Neural Networks (GNNs) have risen to prominence in learning representations for graph structured data.  ...  Ideally, for nodes with high local label smoothness, we expect the learned C to be larger, such that a higher-level local smoothness is enforced to this node during model training.  ... 
arXiv:2010.01777v2 fatcat:7pygu2lukbb55cgcmvahh3kaba

Exploring Object Relation in Mean Teacher for Cross-Domain Detection [article]

Qi Cai, Yingwei Pan, Chong-Wah Ngo, Xinmei Tian, Lingyu Duan, Ting Yao
2019 arXiv   pre-print
The whole architecture is then optimized with three consistency regularizations: 1) region-level consistency to align the region-level predictions between teacher and student, 2) inter-graph consistency  ...  for matching the graph structures between teacher and student, and 3) intra-graph consistency to enhance the similarity between regions of same class within the graph of student.  ...  Mean teacher aims for learning a more smooth domain-invariant function than the model trained with no regularization (Figure 2 (a) ) or only augmented labeled source data (Figure 2 (b) ).  ... 
arXiv:1904.11245v2 fatcat:ymy6aajapvfb7ohytndmtiue7e

Harmonic Unpaired Image-to-image Translation [article]

Rui Zhang, Tomas Pfister, Jia Li
2019 arXiv   pre-print
In this paper, we take a manifold view of the problem by introducing a smoothness term over the sample graph to attain harmonic functions to enforce consistent mappings during the translation.  ...  The recent direction of unpaired image-to-image translation is on one hand very exciting as it alleviates the big burden in obtaining label-intensive pixel-to-pixel supervision, but it is on the other  ...  We introduce smooth regularization over the graph for unpaired image-to-image translation to attain harmonic translations. 2.  ... 
arXiv:1902.09727v1 fatcat:aqgsrimnbzfyjc3vazfyghwd5i
« Previous Showing results 1 — 15 out of 26,473 results