Filters








1,186 Hits in 4.5 sec

Unimodal probability distributions for deep ordinal classification [article]

Christopher Beckham, Christopher Pal
2017 arXiv   pre-print
Probability distributions produced by the cross-entropy loss for ordinal classification problems can possess undesired properties.  ...  We propose a straightforward technique to constrain discrete ordinal probability distributions to be unimodal via the use of the Poisson and binomial probability distributions.  ...  arXiv:1705.05278v2 [stat.ML] 22 Jun 2017 Unimodal Probability Distributions for Deep Ordinal Classification (a) An adult woman baby kid schooler teen adult senior elder distribution A p(y  ... 
arXiv:1705.05278v2 fatcat:t2ymgim5mzh5zccbezpefiu45m

Deep Ordinal Regression using Optimal Transport Loss and Unimodal Output Probabilities [article]

Uri Shaham, Igal Zaidman, Jonathan Svirsky
2021 arXiv   pre-print
In this work, we propose a framework for deep ordinal regression, based on unimodal output distribution and optimal transport loss.  ...  ordinal regression with unimodal output probabilities, while having guarantee on the output unimodality.  ...  Conclusion In this manuscript we presented an approach for deep ordinal regression, inspired by the proportional odds model, utilizing an architectural mechanism for generation of unimodal output probabilities  ... 
arXiv:2011.07607v2 fatcat:fcjo2chbqbbilluei6tynfq7wu

Quasi-Unimodal Distributions for Ordinal Classification

Tomé Albuquerque, Ricardo Cruz, Jaime S. Cardoso
2022 Mathematics  
For that reason, many losses have been proposed in the literature, which model the output probabilities as following a unimodal distribution.  ...  Ordinal classification tasks are present in a large number of different domains.  ...  Contributions: For neural networks, a novel non-parametric ordinal loss is presented that induces output probabilities to follow a quasi-unimodal distribution.  ... 
doi:10.3390/math10060980 fatcat:t7zf3qulz5h35m2rbpz27bs2dm

Ordinal losses for classification of cervical cancer risk

Tomé Albuquerque, Ricardo Cruz, Jaime S Cardoso
2021 PeerJ Computer Science  
A non-parametric ordinal loss for neuronal networks is proposed that promotes the output probabilities to follow a unimodal distribution.  ...  Current approaches to ordinal inference for neural networks are found to not sufficiently take advantage of the ordinal problem or to be too uncompromising.  ...  It would be preferable for output probabilities to follow a unimodal distribution, as depicted by Fig. 1 .  ... 
doi:10.7717/peerj-cs.457 pmid:33981833 pmcid:PMC8080423 fatcat:ago3kcbqbng3fny6pjp6eqjq2y

Unimodal-Concentrated Loss: Fully Adaptive Label Distribution Learning for Ordinal Regression [article]

Qiang Li, Jingjing Wang, Zhaoliang Yao, Yachun Li, Pengju Yang, Jingwei Yan, Chunmao Wang, Shiliang Pu
2022 arXiv   pre-print
Second, the probabilities of neighboring labels should decrease with the increase of distance away from the ground-truth, i.e., the distribution is unimodal.  ...  Under the premise of these principles, we propose a novel loss function for fully adaptive label distribution learning, namely unimodal-concentrated loss.  ...  Unimodal loss Based on the principles we have summarized previously, it is crucial to output a unimodal distribution for ordinal regression tasks.  ... 
arXiv:2204.00309v1 fatcat:px6azhv7qbd4fme5eqptoxfi6e

Unimodal regularisation based on beta distribution for deep ordinal regression

Víctor Manuel Vargas, Pedro Antonio Gutiérrez, César Hervás-Martínez
2021 Pattern Recognition  
Currently, the use of deep learning for solving ordinal classification problems, where categories follow a natural order, has not received much attention.  ...  This regularisation encourages the distribution of the labels to be a soft unimodal distribution, more appropriate for ordinal problems.  ...  ordinal classification problems where the output distribution should be unimodal.  ... 
doi:10.1016/j.patcog.2021.108310 fatcat:gmblepvzenh3zfil45frtjlk4e

COLD Fusion: Calibrated and Ordinal Latent Distribution Fusion for Uncertainty-Aware Multimodal Emotion Recognition [article]

Mani Kumar Tellamekala, Shahin Amiriparian, Björn W. Schuller, Elisabeth André, Timo Giesbrecht, Michel Valstar
2022 arXiv   pre-print
In particular, we impose Calibration and Ordinal Ranking constraints on the variance vectors of audiovisual latent distributions.  ...  To this end, we propose a novel fusion framework in which we first learn latent distributions over audiovisual temporal context vectors separately, and then constrain the variance vectors of unimodal latent  ...  COLD: Calibrated and Ordinal Latent Distributions To effectively learn the unimodal latent distributions for uncertainty-aware fusion, we propose to condition their variance values by applying optimisation  ... 
arXiv:2206.05833v1 fatcat:7skw5owwpndkdgwrbmlymwwexu

Unimodal-uniform Constrained Wasserstein Training for Medical Diagnosis [article]

Xiaofeng Liu, Xu Han, Yukai Qiao, Yi Ge, Lu Jun
2019 arXiv   pre-print
Meanwhile, this paper also proposes of constructing the smoothed target labels that model the inlier and outlier noises by using a unimodal-uniform mixture distribution.  ...  This labeling system is common for medical disease. Previous methods usually construct a multi-binary-classification task or propose some re-parameter schemes in the output unit.  ...  a unimodal-uniform mixture distribution, we also implicitly encourage the probabilities to distribute on the neighbor classes of j * .  ... 
arXiv:1911.02475v1 fatcat:vsiqaverwrbytb4abgbu4wsgpe

Non-parametric Uni-modality Constraints for Deep Ordinal Classification [article]

Soufiane Belharbi, Ismail Ben Ayed, Luke McCaffrey, Eric Granger
2020 arXiv   pre-print
(https://github.com/sbelharbi/unimodal-prob-deep-oc-free-distribution)  ...  We propose a new constrained-optimization formulation for deep ordinal classification, in which uni-modality of the label distribution is enforced implicitly via a set of inequality constraints over all  ...  solution for deep ordinal classification.  ... 
arXiv:1911.10720v3 fatcat:c2tvsha6vnfxdiiuxxxgbtlkpm

Cumulative link models for deep ordinal classification [article]

Víctor-Manuel Vargas and Pedro-Antonio Gutiérrez and César Hervás-Martínez
2019 arXiv   pre-print
This paper proposes a deep convolutional neural network model for ordinal regression by considering a family of probabilistic ordinal link functions in the output layer.  ...  The experiments run over two different ordinal classification problems and the statistical tests confirm that these models improve the results of a nominal model and outperform other robust proposals considered  ...  Unimodal probability distributions Beckham and Pal [26] proposed a straightforward technique to constrain discrete ordinal probability distributions to be unimodal, via the use of the Poisson and binomial  ... 
arXiv:1905.13392v2 fatcat:7hkl64srvfggjdujfzc6hw2vdi

Class Distance Weighted Cross-Entropy Loss for Ulcerative Colitis Severity Estimation [article]

Gorkem Polat, Ilkay Ergenc, Haluk Tarik Kani, Yesim Ozen Alahdab, Ozlen Atug, Alptekin Temizel
2022 arXiv   pre-print
On the other hand, most studies use categorical cross-entropy loss function to train deep learning models, which is not optimal for the ordinal regression problem.  ...  ordinal regression problems.  ...  This approach enforces unimodality by punishing inconsistencies in the posterior probability distribution among adjacent labels.  ... 
arXiv:2202.05167v2 fatcat:t2vaw6u3wndt7avp556rj4s2f4

Multi stain graph fusion for multimodal integration in pathology [article]

Chaitanya Dwivedi, Shima Nofallah, Maryam Pouryahya, Janani Iyer, Kenneth Leidal, Chuhan Chung, Timothy Watkins, Andrew Billin, Robert Myers, John Abel, Ali Behrooz
2022 arXiv   pre-print
Broadly, this paper demonstrates the value of leveraging diverse pathology images for improved ML-powered histologic assessment.  ...  Deep learning methods for WSI analysis Deep learning has achieved unparalleled success in histopathology image analysis and enabled the fast, accurate, and robust classification of complex cell and tissue  ...  This combined representation is passed to a feed-forward network that performs the downstream ordinal classification.  ... 
arXiv:2204.12541v1 fatcat:wkulvrymevgflhhrbyue74cvra

Meta Ordinal Regression Forest for Medical Image Classification with Ordinal Labels [article]

Yiming Lei, Haiping Zhu, Junping Zhang, Hongming Shan
2022 arXiv   pre-print
To improve model generalization with ordinal information, we propose a novel meta ordinal regression forest (MORF) method for medical image classification with ordinal labels, which learns the ordinal  ...  The performance of medical image classification has been enhanced by deep convolutional neural networks (CNNs), which are typically trained with cross-entropy (CE) loss.  ...  However, CE loss is inferior for fitting the ordinal distribution of labels.  ... 
arXiv:2203.07725v1 fatcat:euv64kmbmjbujbkxhlz4lyt6n4

Conservative Wasserstein Training for Pose Estimation

Xiaofeng Liu, Yang Zou, Tong Che, Ping Jia, Peng Ding, Jane You, B. V. K. Vijaya Kumar
2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
We also propose to construct the conservative target labels which model the inlier and outlier noises using a wrapped unimodal-uniform mixture distribution.  ...  ., pose/orientation estimation) in the context of deep learning.  ...  We can sample K +1probabilities (i.e., 0 ≤ k ≤ K) on this PMF and followed by normalization for discrete unimodal probability distributions.  ... 
doi:10.1109/iccv.2019.00835 dblp:conf/iccv/LiuZCJDYK19 fatcat:xfp33xjnfrc2llx2udkpwbwole

Conservative Wasserstein Training for Pose Estimation [article]

Xiaofeng Liu, Yang Zou, Tong Che, Peng Ding, Ping Jia, Jane You, Kumar B.V.K
2019 arXiv   pre-print
We also propose to construct the conservative target labels which model the inlier and outlier noises using a wrapped unimodal-uniform mixture distribution.  ...  ., pose/orientation estimation) in the context of deep learning.  ...  We can sample K + 1 probabilities (i.e., 0 ≤ k ≤ K) on this PMF and followed by normalization for discrete unimodal probability distributions.  ... 
arXiv:1911.00962v1 fatcat:ypcbqpadtvhkpjsuqk7or3dkgi
« Previous Showing results 1 — 15 out of 1,186 results