Filters








15,537 Hits in 3.6 sec

Bridging Ordinary-Label Learning and Complementary-Label Learning [article]

Yasuhiro Katsura, Masato Uchida
2020 arXiv   pre-print
In this paper, we focus on the fact that the loss functions for one-versus-all and pairwise classification corresponding to ordinary-label learning and complementary-label learning satisfy certain additivity  ...  and duality, and provide a framework which directly bridge those existing supervised learning frameworks.  ...  Focusing on equivalency of ordinary label and complementary label, we found that additivity and duality satisfied by loss functions plays a significant role in bridging learning from single ordinary/complementary  ... 
arXiv:2002.02158v4 fatcat:zq2sfhfwyvazdcqsu63udo5iie

Candidate-Label Learning: A Generalization of Ordinary-Label Learning and Complementary-Label Learning

Yasuhiro Katsura, Masato Uchida
2021 SN Computer Science  
In this paper, we focus on the fact that the loss functions for one-versus-all and pairwise classifications corresponding to ordinary-label learning and complementary-label learning satisfy additivity  ...  and duality, and provide a framework that directly bridges the existing supervised learning frameworks.  ...  This mixture representation bridges ordinary-label learning and complementary-label learning from the perspective of the labeled data-generation probability model.  ... 
doi:10.1007/s42979-021-00681-x dblp:journals/sncs/KatsuraU21 fatcat:iprtcbtcqvfhrkv3ticoeggwnq

Learning from a Complementary-label Source Domain: Theory and Algorithms [article]

Yiyang Zhang, Feng Liu, Zhen Fang, Bo Yuan, Guangquan Zhang, Jie Lu
2020 arXiv   pre-print
source domain has plenty of complementary-label data and a small amount of true-label data (partly complementary unsupervised domain adaptation, PC-UDA).  ...  To this end, a complementary label adversarial network} (CLARINET) is proposed to solve CC-UDA and PC-UDA problems.  ...  Previous domain adaptation methods in the shallow regime either try to bridge the source and target by learning invariant feature representations or estimating instance importances using labeled source  ... 
arXiv:2008.01454v1 fatcat:24mffvn2gfd2nf3hy6oxzazsiu

A Circuit-Based Neural Network with Hybrid Learning of Backpropagation and Random Weight Change Algorithms

Changju Yang, Hyongsuk Kim, Shyam Adhikari, Leon Chua
2016 Sensors  
and Circuit-Based Complementary Learning with RWC Hardware-Friendly Error Backpropagation Algorithm The weight updating rule of the ordinary backpropagation learning algorithm [14, 15] is: where µ  ...  Proposed Hybrid Learning: Hardware-Friendly Error-Backpropagation and Circuit-Based Complementary Learning with RWC Hardware-Friendly Error Backpropagation Algorithm The weight updating rule of the ordinary  ...  Author Contributions: Changju Yang conducted the circuit design, experiments and data analysis. Hyognsuk Kim created the research subject and directed the research.  ... 
doi:10.3390/s17010016 pmid:28025566 pmcid:PMC5298589 fatcat:uexzu7mrtnbpdn5jkw4zy7nzyq

What can machine learning do? Implications for citizen scientists

Anna Jia Gander, Marisa Ponti, Alena Seredko
2021 Zenodo  
Human-machine integration in citizen science can harness the contributions of many human observers and use machine learning (ML) to process their contributed data.  ...  The results indicate that experts are involved in every aspect of the loop, from annotating or labeling data to giving them to algorithms to train and make decisions from such predictions.  ...  ., by providing labels and features.  ... 
doi:10.5281/zenodo.4731162 fatcat:hwk7bkjcyja5jlwpari2zjdg3i

Towards Few-shot Entity Recognition in Document Images: A Label-aware Sequence-to-Sequence Framework [article]

Zilong Wang, Jingbo Shang
2022 arXiv   pre-print
Traditional sequence labeling frameworks treat the entity types as class IDs and rely on extensive data and high-quality annotations to learn semantics which are typically expensive in practice.  ...  to capture the spatial correspondence between regions and labels.  ...  However, there are also drops due to improper labels. Overall, we conclude that the semantic meanings of the label surface names are useful to bridge the gap between the labels and entities.  ... 
arXiv:2204.05819v1 fatcat:g723bnhpnrcelavvweijxj6c2y

Better and Faster: Knowledge Transfer from Multiple Self-supervised Learning Tasks via Graph Distillation for Video Classification

Chenrui Zhang, Yuxin Peng
2018 Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence  
Video representation learning is a vital problem for classification task.  ...  Second, high computational and memory cost hinders their application in real-world scenarios.  ...  Acknowledgments This work was supported by National Natural Science Foundation of China under Grant 61771025 and Grant 61532005.  ... 
doi:10.24963/ijcai.2018/158 dblp:conf/ijcai/ZhangP18a fatcat:56vujzi3ajhfdhmnxubck7relu

Balanced MSE for Imbalanced Visual Regression [article]

Jiawei Ren, Mingyuan Zhang, Cunjun Yu, Ziwei Liu
2022 arXiv   pre-print
Compared to imbalanced classification, imbalanced regression focuses on continuous labels, which can be boundless and high-dimensional and hence more challenging.  ...  We revisit MSE from a statistical view and propose a novel loss function, Balanced MSE, to accommodate the imbalanced training label distribution.  ...  Note that we do not include Feature Distribution Smoothing (FDS) in the comparison since it works on feature learning and should be complementary to our method.  ... 
arXiv:2203.16427v1 fatcat:u6esuf5darftnkyilqv5frl4fa

SemiCCA: Efficient Semi-supervised Learning of Canonical Correlations

Akisato Kimura, Hirokazu Kameoka, Masashi Sugiyama, Takuho Nakano, Eisaku Maeda, Hitoshi Sakano, Katsuhiko Ishiguro
2010 2010 20th International Conference on Pattern Recognition  
The proposed method smoothly bridges the eigenvalue problems of CCA and principal component analysis (PCA), and thus its solution can be computed efficiently just by solving a single (generalized) eigenvalue  ...  Preliminary experiments with artificially generated samples and PASCAL VOC data sets demonstrate the effectiveness of the proposed method.  ...  Namely, SemiCCA smoothly bridges CCA with paired samples and PCA with paired and unpaired samples by a trade-off parameter.  ... 
doi:10.1109/icpr.2010.719 dblp:conf/icpr/KimuraKSNMSI10 fatcat:6qgw3rw7zbfhdejazt3q7hmrzu

SemiCCA: Efficient Semi-supervised Learning of Canonical Correlations

Akisato Kimura, Masashi Sugiyama, Takuho Nakano, Hirokazu Kameoka, Hitoshi Sakano, Eisaku Maeda, Katsuhiko Ishiguro
2013 IPSJ Online Transactions  
The proposed method smoothly bridges the eigenvalue problems of CCA and principal component analysis (PCA), and thus its solution can be computed efficiently just by solving a single (generalized) eigenvalue  ...  Preliminary experiments with artificially generated samples and PASCAL VOC data sets demonstrate the effectiveness of the proposed method.  ...  Namely, SemiCCA smoothly bridges CCA with paired samples and PCA with paired and unpaired samples by a trade-off parameter.  ... 
doi:10.2197/ipsjtrans.6.37 fatcat:hofm3mf26rbaheus54yazjlzjq

Better and Faster: Knowledge Transfer from Multiple Self-supervised Learning Tasks via Graph Distillation for Video Classification [article]

Chenrui Zhang, Yuxin Peng
2018 arXiv   pre-print
Video representation learning is a vital problem for classification task.  ...  Recently, a promising unsupervised paradigm termed self-supervised learning has emerged, which explores inherent supervisory signals implied in massive data for feature learning via solving auxiliary tasks  ...  Acknowledgments This work was supported by National Natural Science Foundation of China under Grant 61771025 and Grant 61532005.  ... 
arXiv:1804.10069v1 fatcat:27xmwll4vzec3aatscu6zmwriu

Walk and Learn: Facial Attribute Representation Learning from Egocentric Video and Contextual Data [article]

Jing Wang, Yu Cheng, Rogerio Schmidt Feris
2016 arXiv   pre-print
Moreover, even without using manually annotated identity labels for pre-training as in previous methods, our approach achieves results that are better than the state of the art.  ...  This work explores, for the first time, the use of this contextual information, as people with wearable cameras walk across different neighborhoods of a city, in order to learn a rich feature representation  ...  Italy, Brooklyn Bridge, Chinatown, Flushing, and others.  ... 
arXiv:1604.06433v3 fatcat:6txunxggp5b6nievr7obm7du6m

Vast Darkness: The Nature of Explanation

CLYDE KLUCKHOHN
1959 Contemporary Psychology  
And the key is a satisfactory theory of learn- ing. Symmetrical formation is schismogenesis leads to more intense rivalry; complementary schismogenesis produces increasing role differentiation.  ...  These same labels identify which aspects may be found in each behavior.  ... 
doi:10.1037/005987 fatcat:mw4uwlwww5aajmf7r646xrmfcu

Walk and Learn: Facial Attribute Representation Learning from Egocentric Video and Contextual Data

Jing Wang, Yu Cheng, Rogerio Schmidt Feris
2016 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
Moreover, even without using manually annotated identity labels for pre-training as in previous methods, our approach achieves results that are better than the state of the art.  ...  This work explores, for the first time, the use of this contextual information, as people with wearable cameras walk across different neighborhoods of a city, in order to learn a rich feature representation  ...  Italy, Brooklyn Bridge, Chinatown, Flushing, and others.  ... 
doi:10.1109/cvpr.2016.252 dblp:conf/cvpr/WangCF16 fatcat:mh266al54fem7k3tpjngw5jv6m

Effective Model Integration Algorithm for Improving Link and Sign Prediction in Complex Networks [article]

Chuang Liu, Shimin Yu, Ying Huang, Zi-Ke Zhang
2021 arXiv   pre-print
Link and sign prediction in complex networks bring great help to decision-making and recommender systems, such as in predicting potential relationships or relative status levels.  ...  In this work, we propose an effective model integration algorithm consisting of network embedding, network feature engineering, and an integrated classifier, which can perform the link and sign prediction  ...  (A) and (B) show an ordinary undirected unsigned network and a directed signed network as examples of link and sign prediction, where orange and gray respectively represent typical bridge edge structures  ... 
arXiv:2108.01532v1 fatcat:k5qqni5fzvcgxisxdpwfcocu7q
« Previous Showing results 1 — 15 out of 15,537 results