9,589 Hits in 1.5 sec

Multi-label learning with missing and completely unobserved labels

Jun Huang, Linchuan Xu, Kun Qian, Jing Wang, Kenji Yamanishi
2021 Data mining and knowledge discovery  
In this paper, we propose a new approach named MCUL to solve multi-label learning with Missing and Completely Unobserved Labels.  ...  We refer to the problem as multi-label learning with missing and completely unobserved labels, and argue that it is necessary to discover these completely unobserved labels in order to mine useful knowledge  ...  as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.  ... 
doi:10.1007/s10618-021-00743-x fatcat:dbsdr34lgrg3tm3gxw5fuf4m5i

To Avoid the Pitfall of Missing Labels in Feature Selection: A Generative Model Gives the Answer

Yuanyuan Xu, Jun Wang, Jinmao Wei
In multi-label learning, instances have a large number of noisy and irrelevant features, and each instance is associated with a set of class labels wherein label information is generally incomplete.  ...  Existing approaches either superficially consider the missing labels as negative or indiscreetly impute them with some predicted values, which may either overestimate unobserved labels or introduce new  ...  Acknowledgments This work was partially supported by the National Natural Science Foundation of China(61772288), the Natural Science Foundation of Tianjin City(18JCZDJC30900) and the Shandong Provincial  ... 
doi:10.1609/aaai.v34i04.6127 fatcat:c7mwfflsozckdl7wtwpm3pqmha

Multi-Label Learning from Single Positive Labels [article]

Elijah Cole, Oisin Mac Aodha, Titouan Lorieul, Pietro Perona, Dan Morris, Nebojsa Jojic
2021 arXiv   pre-print
We explore this special case of learning from missing labels across four different multi-label image classification datasets for both linear classifiers and end-to-end fine-tuned deep networks.  ...  We extend existing multi-label losses to this setting and propose novel variants that constrain the number of expected positive labels during training.  ...  DGE1745301) and the Microsoft AI for Earth program. We would also like to thank Jennifer J. Sun, Matteo Ruggero Ronchi, and Joseph Marino for helpful feedback.  ... 
arXiv:2106.09708v2 fatcat:xao37rffsbcepdrvtsh4yiidzm

An Efficient Large-scale Semi-supervised Multi-label Classifier Capable of Handling Missing labels [article]

Amirhossein Akbarnejad, Mahdieh Soleymani Baghshah
2016 arXiv   pre-print
Multi-label classifiers have to address many problems including: handling large-scale datasets with many instances and a large set of labels, compensating missing label assignments in the training set,  ...  Moreover, the proposed method have excellent mechanisms for handling missing labels, dealing with large-scale datasets, as well as exploiting unlabeled data.  ...  To exploit the unlabeled data, as in SLRM [14] , we seek to learn a smooth mapping to transform the feature vectors to the complete unobserved label vectors.  ... 
arXiv:1606.05725v1 fatcat:ndn5busqpffwlcjw5vcj6fheuy

Semi-supervised Learning with Missing Values Imputation [article]

Buliao Huang and Yunhui Zhu and Muhammad Usman and Huanhuan Chen
2021 arXiv   pre-print
Missing values imputation methods are often employed to replace the missing values with substitute values.  ...  SSCFlow explicitly utilizes the label information to facilitate the imputation and classification simultaneously by estimating the conditional distribution of incomplete instances with a novel semi-supervised  ...  The task of this paper is to assign each missing value in the incomplete dataset X with its substituted value and predict the unobserved labels Y te .  ... 
arXiv:2106.01708v2 fatcat:4fr74ttuljgmxiexjrva2too7e

Active Refinement for Multi-Label Learning: A Pseudo-Label Approach [article]

Cheng-Yu Hsieh, Wei-I Lin, Miao Xu, Gang Niu, Hsuan-Tien Lin, Masashi Sugiyama
2021 arXiv   pre-print
The goal of multi-label learning (MLL) is to associate a given instance with its relevant labels from a set of concepts.  ...  The key idea within our approach is to learn to assign pseudo-labels to the unlabeled entries, and in turn leverage the pseudo-labels to train the underlying classifier and to inform a better query strategy  ...  We outline the complete learning algorithm with pseudo-labels in Algorithm 1.  ... 
arXiv:2109.14676v1 fatcat:jfvo2an43rc5djxempnav2bdmy

Variational Selective Autoencoder: Learning from Partially-Observed Heterogeneous Data [article]

Yu Gong and Hossein Hajimirsadeghi and Jiawei He and Thibaut Durand and Greg Mori
2021 arXiv   pre-print
VSAE learns the latent dependencies in heterogeneous data by modeling the joint distribution of observed data, unobserved data, and the imputation mask which represents how the data are missing.  ...  Meanwhile, heterogeneous data are often associated with missingness in real-world applications due to heterogeneity and noise of input sources.  ...  Missingness is completely independent of data, p(x o , x u , m) = p(x o , x u )p(m) (1) Missing At Random (MAR).  ... 
arXiv:2102.12679v1 fatcat:qmwywecuwnbbtm3b3hjldgcsca

Learning a Deep ConvNet for Multi-Label Classification With Partial Labels

Thibaut Durand, Nazanin Mehrasa, Greg Mori
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Our approach allows the use of the same training settings as when learning with all the annotations. We further explore several curriculum learning based strategies to predict missing labels.  ...  Experiments are performed on three large-scale multi-label datasets: MS COCO, NUS-WIDE and Open Images.  ...  Multi-label tasks often involve incomplete training data, hence several methods have been proposed to solve the problem of multi-label learning with missing labels (MLML).  ... 
doi:10.1109/cvpr.2019.00074 dblp:conf/cvpr/DurandMM19 fatcat:pmhwu26hcnaf7iz2gzfrwi2bzy

CEMENT: Incomplete Multi-View Weak-Label Learning with Long-Tailed Labels [article]

Zhiwei Li, Lu Sun, Mineichi Kudo, Kego Kimura
2022 arXiv   pre-print
A variety of modern applications exhibit multi-view multi-label learning, where each sample has multi-view features, and multiple labels are correlated via common views.  ...  In recent years, several methods have been proposed to cope with it and achieve much success, but still suffer from two key problems: 1) lack the ability to deal with the incomplete multi-view weak-label  ...  In contrast, the proposed CEMENT completes the unobserved labels by embedding both incomplete views and weak labels into multiple subspaces with adaptive weights, and captures tail labels by exploiting  ... 
arXiv:2201.01079v3 fatcat:jppxuc2mx5bmvatjja3ejyiqj4

Incomplete Attribute Learning with auxiliary labels

Kongming Liang, Yuhong Guo, Hong Chang, Xilin Chen
2017 Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence  
By jointly predicting the attributes from the input images and modeling the relationship of attributes and auxiliary labels, the missing attributes can be recovered effectively.  ...  Visual attribute learning is a fundamental and challenging problem for image understanding.  ...  In this case, our label indicator matrix Y for the labeled training images is not completely observed. Hence both Y u and part of Y contain missing labels or unobserved entries.  ... 
doi:10.24963/ijcai.2017/313 dblp:conf/ijcai/LiangGCC17 fatcat:pzdzmo5we5cntall42nz6hbjyq

Learning a Deep ConvNet for Multi-label Classification with Partial Labels [article]

Thibaut Durand, Nazanin Mehrasa, Greg Mori
2019 arXiv   pre-print
Our approach allows the use of the same training settings as when learning with all the annotations. We further explore several curriculum learning based strategies to predict missing labels.  ...  Experiments are performed on three large-scale multi-label datasets: MS COCO, NUS-WIDE and Open Images.  ...  Multi-label tasks often involve incomplete training data, hence several methods have been proposed to solve the problem of multi-label learning with missing labels (MLML).  ... 
arXiv:1902.09720v1 fatcat:tqv7lexguzhlzemmcujme7htde

Learning Priors for Semantic 3D Reconstruction [chapter]

Ian Cherabier, Johannes L. Schönberger, Martin R. Oswald, Marc Pollefeys, Andreas Geiger
2018 Lecture Notes in Computer Science  
Our network performs a fixed number of unrolled multi-scale optimization iterations with shared interaction weights.  ...  In contrast to existing variational methods for semantic 3D reconstruction, our model is end-to-end trainable and captures more complex dependencies between the semantic labels and the 3D geometry.  ...  Disclaimer: The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied  ... 
doi:10.1007/978-3-030-01258-8_20 fatcat:ot4nhsuvzzb7ja5dtpj6e7yyce

Improving Temporal Interpolation of Head and Body Pose using Gaussian Process Regression in a Matrix Completion Setting [article]

Stephanie Tan, Hayley Hung
2018 arXiv   pre-print
The current state-of-the-art multimodal approach to HBPE utilizes the matrix completion method in a transductive setting to predict pose labels for unobserved samples.  ...  As well as fitting a more flexible model to missing labels in time, we posit that our approach also loosens the head and body coupling constraint, allowing for a more expressive model of the head and body  ...  Matrix completion is an iterative method that a empts to ll in missing entries in a matrix, which in our context correspond to unobserved pose labels.  ... 
arXiv:1808.01837v1 fatcat:uygt3s7pqjfwxp4xhjozm7kqna

Im2Pano3D: Extrapolating 360° Structure and Semantics Beyond the Field of View

Shuran Song, Andy Zeng, Angel X. Chang, Manolis Savva, Silvio Savarese, Thomas Funkhouser
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
Experiments demonstrate that Im2Pano3D is able to predict the semantics and 3D structure of the unobserved scene with more than 56% pixel accuracy and less than 0.52m average distance error, which is significantly  ...  To make this possible, Im2Pano3D leverages strong contextual priors learned from large-scale synthetic and realworld indoor scenes.  ...  Acknowledgment This work is supported by Google, Intel, and the NSF(VEC 1539014/ 1539099). It makes use of data from Matterport3D and Planner5D, and hardware donated by NVIDIA and Intel.  ... 
doi:10.1109/cvpr.2018.00405 dblp:conf/cvpr/SongZCSSF18 fatcat:tqgptzdezbdsllikoixksbewfa

Matrix Co-completion for Multi-label Classification with Missing Features and Labels [article]

Miao Xu, Gang Niu, Bo Han, Ivor W. Tsang, Zhi-Hua Zhou, Masashi Sugiyama
2018 arXiv   pre-print
We consider a challenging multi-label classification problem where both feature matrix and label matrix have missing entries.  ...  An existing method concatenated and as [; ] and applied a matrix completion (MC) method to fill the missing entries, under the assumption that [; ] is of low-rank.  ...  Acknowledgments We want to thank Bo-Jian Hou for discussion and polishing of the paper.  ... 
arXiv:1805.09156v1 fatcat:s2tjqgeny5bh3daxm5y5adf25a
« Previous Showing results 1 — 15 out of 9,589 results