Filters








72,953 Hits in 4.5 sec

Large-Scale Distance Metric Learning with Uncertainty

Qi Qian, Jiasheng Tang, Hao Li, Shenghuo Zhu, Rong Jin
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
Most of the existing methods propose to learn a distance metric with pairwise or triplet constraints.  ...  Distance metric learning (DML) has been studied extensively in the past decades for its superior performance with distance-based algorithms.  ...  It indicates that latent examples with low uncertainty are more appropriate for the large-scale data set as the reference points.  ... 
doi:10.1109/cvpr.2018.00891 dblp:conf/cvpr/QianTLZJ18 fatcat:fbpgqfyyqja7njxvdcuyz6kfqq

Large-scale Distance Metric Learning with Uncertainty [article]

Qi Qian, Jiasheng Tang, Hao Li, Shenghuo Zhu, Rong Jin
2018 arXiv   pre-print
Most of the existing methods propose to learn a distance metric with pairwise or triplet constraints.  ...  Distance metric learning (DML) has been studied extensively in the past decades for its superior performance with distance-based algorithms.  ...  It indicates that latent examples with low uncertainty are more appropriate for the large-scale data set as the reference points.  ... 
arXiv:1805.10384v1 fatcat:msmn5l4bxfhahdny7knpjjjime

Meta-Learned Confidence for Few-shot Learning [article]

Seong Min Kye, Hae Beom Lee, Hoirin Kim, Sung Ju Hwang
2020 arXiv   pre-print
We validate our few-shot learning model with meta-learned confidence on four benchmark datasets, on which it largely outperforms strong recent baselines and obtains new state-of-the-art results.  ...  We achieve this by meta-learning an input-adaptive distance metric over a task distribution under various model and data perturbations, which will enforce consistency on the model predictions under diverse  ...  We see that the PN with metric scaling underperforms the plain PN with Euclidean distance.  ... 
arXiv:2002.12017v2 fatcat:iizg5agw2zexhaxfs3vip4ackq

MetaCVR: Conversion Rate Prediction via Meta Learning in Small-Scale Recommendation Scenarios [article]

Xiaofeng Pan, Ming Li, Jing Zhang, Keren Yu, Luping Wang, Hong Wen, Chengjun Mao, Bo Cao
2022 arXiv   pre-print
Subsequently, a Distance Metric Network (DMN) is devised to calculate the distance metrics between each sample and all prototypes to facilitate mitigating the distribution uncertainty.  ...  Different from large-scale platforms such as Taobao and Amazon, CVR modeling in small-scale recommendation scenarios is more challenging due to the severe Data Distribution Fluctuation (DDF) issue.  ...  Metric Network In metric-based meta learning, the choice of distance metric is crucial.  ... 
arXiv:2112.13753v4 fatcat:u3st3d3rhrc2xm7arvz5d3qwfu

Mitigating Uncertainty in Document Classification

Xuchao Zhang, Fanglan Chen, Chang-Tien Lu, Naren Ramakrishnan
2019 Proceedings of the 2019 Conference of the North  
We also design a metric learning method on feature representations, which can boost the performance of dropout-based uncertainty methods with smaller prediction variance in accurate prediction trials.  ...  returned by machine learning models.  ...  Most existing uncertainty models are based on Bayesian models, which are not only timeconsuming but also unable to handle large-scale data sets.  ... 
doi:10.18653/v1/n19-1316 dblp:conf/naacl/ZhangCLR19 fatcat:ylkundijqzdffbvzua7yxayj5u

Perception Score, A Learned Metric for Open-ended Text Generation Evaluation [article]

Jing Gu, Qingyang Wu, Zhou Yu
2020 arXiv   pre-print
Existing metrics such as BLEU show a low correlation with human judgment. We propose a novel and powerful learning-based evaluation metric: Perception Score.  ...  Moreover, it also shows the amount of uncertainty about its evaluation result. By connecting the uncertainty, Perception Score gives a more accurate evaluation for the generation system.  ...  Uncertainty comes with judgment.  ... 
arXiv:2008.03082v2 fatcat:sgjbrr6w4vanjc6x7yifde42s4

Active learning for large multi-class problems

Prateek Jain, Ashish Kapoor
2009 2009 IEEE Conference on Computer Vision and Pattern Recognition  
Scarcity and infeasibility of human supervision for large scale multi-class classification problems necessitates active learning.  ...  Unfortunately, existing active learning methods for multi-class problems are inherently binary methods and do not scale up to a large number of classes.  ...  Recent methods in metric learning learns a distance metric parameterized by a Mahalanobis metric that is consistent with the training data [1, 5, 20, 8] .  ... 
doi:10.1109/cvpr.2009.5206651 dblp:conf/cvpr/JainK09 fatcat:sn5s2wvyxjb2bamvnkti7emngi

Only Bayes should learn a manifold (on the estimation of differential geometric structure from data) [article]

Søren Hauberg
2019 arXiv   pre-print
To properly learn the differential geometric structure, non-probabilistic methods must apply regularizations that enforce large gradients, which go against common wisdom.  ...  We investigate learning of the differential geometric structure of a data manifold embedded in a high-dimensional Euclidean space.  ...  Also note that scaling Σ also scales the expected distance: very uncertain metrics imply large expected distances.  ... 
arXiv:1806.04994v3 fatcat:qhysomkjbjgnfdr6tqmfkjh47e

Active learning for large multi-class problems

P. Jain, A. Kapoor
2009 2009 IEEE Conference on Computer Vision and Pattern Recognition  
Scarcity and infeasibility of human supervision for large scale multi-class classification problems necessitates active learning.  ...  Unfortunately, existing active learning methods for multi-class problems are inherently binary methods and do not scale up to a large number of classes.  ...  Recent methods in metric learning learns a distance metric parameterized by a Mahalanobis metric that is consistent with the training data [1, 5, 20, 8] .  ... 
doi:10.1109/cvprw.2009.5206651 fatcat:uw25r5zxqvgsjjnpnqbug5fm6a

Randomised Gaussian Process Upper Confidence Bound for Bayesian Optimisation

Julian Berk, Sunil Gupta, Santu Rana, Svetha Venkatesh
2020 Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence  
with large incremental batches.  ...  Deep metric learning with 3 standard AL heuristics (uncertainty (US), expected gradient length (EGL), model output change (MOC)) all outperform random sampling (RND) when triplets are added one at a time  ... 
doi:10.24963/ijcai.2020/312 dblp:conf/ijcai/KumariGCC20 fatcat:t6l6srkcufe5bihmudcgs4fazu

Parcel-Based Active Learning for Large Extent Cultivated Area Mapping

Ines Ben Slimene Ben Amor, Nesrine Chehata, Jean-Stephane Bailly, Imed Riadh Farah, Philippe Lagacherie
2018 IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing  
This paper focuses on agricultural land cover mapping at a high resolution scale and over large areas from an operational point of view and from a high-resolution monodate image.  ...  processed at parcel scale.  ...  MAO is not effective with a multi-class problem. Thus the kept distance-based metric is Mahalanobis distance [26] although is it parametric.  ... 
doi:10.1109/jstars.2017.2751148 fatcat:ykkwbtlg5bfjjokfzuvmg6xsli

Absolute geo-localization thanks to Hidden Markov Model and exemplar-based metric learning

Cedric Le Barz, Nicolas Thome, Matthieu Cord, Stephane Herbin, Martial Sanfourche
2015 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
To achieve this goal, we define some constraints so that the distance between a database image and a query image representing the same scene is smaller than the distance between this query image and other  ...  This paper addresses the problem of absolute visual ego-localization of an autonomous vehicle equipped with a monocular camera that has to navigate in an urban environment.  ...  leading to an exemplar-based metric learning scheme.  ... 
doi:10.1109/cvprw.2015.7301394 dblp:conf/cvpr/BarzTCHS15 fatcat:iqaxko2jkvcdjkwo6vang64bte

STUN: Self-Teaching Uncertainty Estimation for Place Recognition [article]

Kaiwen Cai, Chris Xiaoxuan Lu, Xiaowei Huang
2022 arXiv   pre-print
To this end, we first train a teacher net using a standard metric learning pipeline to produce embedding priors.  ...  Our experimental results on the large-scale Pittsburgh30k dataset demonstrate that STUN outperforms the state-of-the-art methods in both recognition accuracy and the quality of uncertainty estimation.  ...  This limitation makes classification based place recognition impractical for real-world large scale applications. Thus, in this study, we only focus on metric learning based place recognition.  ... 
arXiv:2203.01851v1 fatcat:7kwmog5f3fffrads64cvx6bf7a

Elliptical Ordinal Embedding [article]

Aïssatou Diallo, Johannes Fürnkranz
2021 arXiv   pre-print
Typically, each object is mapped onto a point vector in a low dimensional metric space.  ...  In addition, we illustrate the merit of modelling uncertainty, which enriches the visual perception of the mapped objects in the space.  ...  Moreover, these methods rely on expensive gradient projections and do not easily scale to large datasets.  ... 
arXiv:2105.10457v2 fatcat:b62rf5kgu5fjzdxubhc5lq5npi

Batch Decorrelation for Active Metric Learning [article]

Priyadarshini K, Ritesh Goru, Siddhartha Chaudhuri, Subhasis Chaudhuri
2020 arXiv   pre-print
We present an active learning strategy for training parametric models of distance metrics, given triplet-based similarity assessments: object x_i is more similar to object x_j than to x_k.  ...  In contrast to prior work on class-based learning, where the fundamental goal is classification and any implicit or explicit metric is binary, we focus on perceptual metrics that express the degree of  ...  with large incremental batches.  ... 
arXiv:2005.10008v2 fatcat:5dacudmo25didelhmaav2peo3m
« Previous Showing results 1 — 15 out of 72,953 results