Filters








53,275 Hits in 3.8 sec

Robust Graph Learning From Noisy Data

Zhao Kang, Haiqi Pan, Steven C. H. Hoi, Zenglin Xu
2019 IEEE Transactions on Cybernetics  
In this paper, we propose a novel robust graph learning scheme to learn reliable graphs from the real-world noisy data by adaptively removing noise and errors in the raw data.  ...  Learning graphs from data automatically have shown encouraging performance on clustering and semisupervised learning tasks.  ...  In this work, we propose a novel robust graph learning scheme to learn robust and reliable graphs from real data.  ... 
doi:10.1109/tcyb.2018.2887094 pmid:30629527 fatcat:qkvy5szpa5ajhj7ugigsg5oqgq

Robust Deep Graph Based Learning for Binary Classification [article]

Minxiang Ye, Vladimir Stankovic, Lina Stankovic, Gene Cheung
2019 arXiv   pre-print
In this paper, we propose a robust binary classifier, based on CNNs, to learn deep metric functions, which are then used to construct an optimal underlying graph structure used to clean noisy labels via  ...  However, feature learning becomes more difficult if some training labels are noisy.  ...  Robust Graph-based Learning A label propagation method is proposed in [29] to evenly spread, throughout the graph, label distributions from selected labeled nodes, which are usually noisy and with heuristic  ... 
arXiv:1912.03321v1 fatcat:sxoqsqn35bfzdehitoddaoetri

How Does a Neural Network's Architecture Impact Its Robustness to Noisy Labels? [article]

Jingling Li, Mozhi Zhang, Keyulu Xu, John P. Dickerson, Jimmy Ba
2021 arXiv   pre-print
Noisy labels are inevitable in large real-world datasets. In this work, we explore an area understudied by previous works -- how the network's architecture impacts its robustness to noisy labels.  ...  Our framework measures a network's robustness via the predictive power in its representations -- the test performance of a linear model trained on the learned representations using a small set of clean  ...  Module (1) can learn good representations from noisy labels.  ... 
arXiv:2012.12896v2 fatcat:hborktrhdbdablmb4ulp6hrlay

Sparse Graph Attention Networks [article]

Yang Ye, Shihao Ji
2021 arXiv   pre-print
Graph Neural Networks (GNNs) have proved to be an effective representation learning framework for graph-structured data, and have achieved state-of-the-art performance on many practical predictive tasks  ...  In particular, SGATs can remove about 50%-80% edges from large assortative graphs, while retaining similar classification accuracies.  ...  The robustness of SGAT is of practical importance as real-world graph-structured data are often very noisy, and a robust graph learning algorithm that can learn from both assortative and disassortative  ... 
arXiv:1912.00552v2 fatcat:vqvo7aty6beyphlgtgi5qsg7fq

Bayesian Robust Graph Contrastive Learning [article]

Yancheng Wang, Yingzhen Yang
2022 arXiv   pre-print
In this work, we propose a novel and robust method, Bayesian Robust Graph Contrastive Learning (BRGCL), which trains a GNN encoder to learn robust node representations.  ...  However, noise, which inevitably exists in real-world graph data, would considerably degrade the performance of GNNs as the noise is easily propagated via the graph structure.  ...  First, we propose Bayesian Robust Graph Contrastive Learning (BRGCL) where a fully unsupervised encoder is trained on noisy graph data.  ... 
arXiv:2205.14109v3 fatcat:ue34cs4rxnarzequnzx3di2ns4

Self-Guided Learning to Denoise for Robust Recommendation [article]

Yunjun Gao, Yuntao Du, Yujia Hu, Lu Chen, Xinjun Zhu, Ziquan Fang, Baihua Zheng
2022 arXiv   pre-print
Besides, our method can automatically switch its learning phase at the memorization point from memorization to self-guided learning, and select clean and informative memorized data via a novel adaptive  ...  The experimental results on three benchmark datasets demonstrate the effectiveness of SGDL over the state-of-the-art denoising methods like T-CE, IR, DeCA, and even state-of-the-art robust graph-based  ...  ., NGCF and LightGCN in our experiments), since they regard the noisy interactions as noisy edges, and devise enhanced graph learning methods for robust recommendation. • SGCN [5] is the state-of-the-art  ... 
arXiv:2204.06832v1 fatcat:tajlofkk4vazjevt7jelwo2a7u

Learning to Denoise Unreliable Interactions for Graph Collaborative Filtering

Changxin Tian, Yuexiang Xie, Yaliang Li, Nan Yang, Wayne Xin Zhao
2022 Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval  
However, existing GNN-based CF models suffer from noisy user-item interaction data, which seriously affects the effectiveness and robustness in real-world applications.  ...  The graph denoising module is designed for reducing the impact of noisy interactions on the representation learning of GNN, by adopting both a hard denoising strategy (i.e., discarding interactions that  ...  However, since we add unobserved interactions in G, the learned embeddings from G are likely to deviate from those learned from the denoised interaction graph G (Section 3.2.3) according to [40, 54] .  ... 
doi:10.1145/3477495.3531889 fatcat:le2y3wxuine4hovgubxhb66kbq

Decoupling Representation and Classifier for Noisy Label Learning [article]

Hui Zhang, Quanming Yao
2020 arXiv   pre-print
Thus, we are motivated to design a new method, i.e., REED, to leverage above discoveries to learn from noisy labels robustly.  ...  ., obtaining the representation by self-supervised learning without any labels, transferring the noisy label learning problem into a semisupervised one by the classifier directly and reliably trained with  ...  information from other data.  ... 
arXiv:2011.08145v1 fatcat:7fokk3rdobbh7efkuwpo4f3zqq

NRGNN: Learning a Label Noise-Resistant Graph Neural Network on Sparsely and Noisily Labeled Graphs [article]

Enyan Dai, Charu Aggarwal, Suhang Wang
2021 arXiv   pre-print
Thus, we investigate a novel problem of learning a robust GNN with noisy and limited labels.  ...  Though extensive studies have been conducted to learn neural networks with noisy labels, they mostly focus on independent and identically distributed data and assume a large number of noisy labels are  ...  The work on learning a robust GNN with noisy and limited labels is rather limited [8, 43] . Therefore, it is important to develop a robust GNN that could deal with noisy and limited labels.  ... 
arXiv:2106.04714v1 fatcat:vesmcf4q2fay7ewycv5aqiaomm

Noise-robust classification with hypergraph neural network [article]

Nguyen Trinh Vu Dang, Loc Tran, Linh Tran
2022 arXiv   pre-print
This method is utilized to solve the noisy label learning problem.  ...  neural network are employed to solve the noisy label learning problem.  ...  However, filtering all the samples that are misclassified by the classification filtering system is too inflexible and risky since the classification filtering system learned from data with noisy labels  ... 
arXiv:2102.01934v3 fatcat:cva5g7xzjbh5lao4ijusk42xxa

A Regularized Attention Mechanism for Graph Attention Networks [article]

Uday Shankar Shanthamallu, Jayaraman J. Thiagarajan, Andreas Spanias
2020 arXiv   pre-print
In particular, there is a surge in deep learning solutions for graph-structured data, due to its wide-spread applicability in several fields.  ...  Machine learning models that can exploit the inherent structure in data have gained prominence.  ...  Consequently, machine learning formalisms for graph-structured data [5, 6] have become prominent, and are regularly being adopted for information extraction and analysis.  ... 
arXiv:1811.00181v2 fatcat:2li2pqrwpfcmjnjeioe7jun5c4

Robust Spectral Learning for Unsupervised Feature Selection

Lei Shi, Liang Du, Yi-Dong Shen
2014 2014 IEEE International Conference on Data Mining  
Compared with existing methods which are sensitive to noisy features, our proposed method utilizes a robust local learning method to construct the graph Laplacian and a robust spectral regression method  ...  However, existing spectral feature selection algorithms suffer from two major problems: 1) since the graph Laplacian is constructed from the original feature space, noisy and irrelevant features may have  ...  Our goal is to design a method which can both utilize the local structure of data and handle noisy features and outliers for robust graph embedding.  ... 
doi:10.1109/icdm.2014.58 dblp:conf/icdm/ShiDS14 fatcat:6nr5yqyranemjpkwjqlsy2ihoe

Noise-robust classification with hypergraph neural network

Nguyen Trinh Vu Dang, Loc Tran, Linh Tran
2021 Indonesian Journal of Electrical Engineering and Computer Science  
This method is utilized to solve the noisy label learning problem.  ...  neural network are employed to solve the noisy label learning problem.  ...  However, filtering all the samples that are misclassified by the classification filtering system is too inflexible and risky since the classification filtering system learned from data with noisy labels  ... 
doi:10.11591/ijeecs.v21.i3.pp1465-1473 fatcat:iz7g63by3vgofnrj6gocz5b4zy

Semantic Sparse Recoding of Visual Content for Image Applications

Zhiwu Lu, Peng Han, Liwei Wang, Ji-Rong Wen
2015 IEEE Transactions on Image Processing  
This paper presents a novel L1-norm semi-supervised learning algorithm for robust image analysis by giving new L1-norm formulation of Laplacian regularization which is the key step of graph-based semi-supervised  ...  Due to the sparsity induced by sparse coding, the proposed algorithm can deal with the noise in the data to some extent and thus has important applications to robust image analysis, such as noise-robust  ...  In this paper, we focus on proposing a novel noise-robust graph-based semi-supervised learning method, rather than the well-studied graph construction.  ... 
doi:10.1109/tip.2014.2375641 pmid:25438314 fatcat:hx7psd56d5d5dh2epajtblgsam

PI-GNN: A Novel Perspective on Semi-Supervised Node Classification against Noisy Labels [article]

Xuefeng Du, Tian Bian, Yu Rong, Bo Han, Tongliang Liu, Tingyang Xu, Wenbing Huang, Junzhou Huang
2021 arXiv   pre-print
In this paper, we propose a novel robust learning objective dubbed pairwise interactions (PI) for the model, such as Graph Neural Network (GNN) to combat noisy labels.  ...  Semi-supervised node classification, as a fundamental problem in graph learning, leverages unlabeled nodes along with a small portion of labeled nodes for training.  ...  From Table 2 , we made several observations: Firstly, the GNN trained with the PI learning objective is more robust to the noisy labels, where both PI-GNN wo/ ue and PI-GNN perform much better than a  ... 
arXiv:2106.07451v1 fatcat:rugtqiw2q5e6jjhtvdhr7qzfwq
« Previous Showing results 1 — 15 out of 53,275 results