168 Hits in 4.7 sec

Bandit Multiclass Linear Classification: Efficient Algorithms for the Separable Case [article]

Alina Beygelzimer, Dávid Pál, Balázs Szörényi, Devanathan Thiruvenkatachari, Chen-Yu Wei, Chicheng Zhang
2019 arXiv   pre-print
We study the problem of efficient online multiclass linear classification with bandit feedback, where all examples belong to one of K classes and lie in the d-dimensional Euclidean space.  ...  Under the strong linear separability condition, we design an efficient algorithm that achieves a near-optimal mistake bound of O( K/γ^2 ). 2.  ...  notions of linear separability for multiclass classification.  ... 
arXiv:1902.02244v2 fatcat:tlwtx7ojs5hw5j46jm6olfehwu

The price of bandit information in multiclass online classification [article]

Amit Daniely, Tom Helbertal
2013 arXiv   pre-print
. - Efficient bandit algorithms for online multiclass prediction).  ...  We apply these results to the class of γ-margin multiclass linear classifiers in ^d.  ...  Theorem 2.5 We shall use the following Lemma. Consider the following game: A r.v., U , is sampled uniformly from Y . Then the player, that does not observe U , try to guess U .  ... 
arXiv:1302.1043v2 fatcat:q5zqmjm3rjg35o3splzmfbabjq

Multiclass Classification using dilute bandit feedback [article]

Gaurav Batra, Naresh Manwani
2021 arXiv   pre-print
We propose an algorithm for multiclass classification using dilute bandit feedback (MC-DBF), which uses the exploration-exploitation strategy to predict the candidate set in each trial.  ...  This paper introduces a new online learning framework for multiclass classification called learning with diluted bandit feedback.  ...  Online algorithms for multiclass classification using partial labels.  ... 
arXiv:2105.08093v1 fatcat:lkbjtlhazbe5fm2gqxh2zmv45q

Exact Passive-Aggressive Algorithms for Multiclass Classification Using Bandit Feedbacks

Maanik Arora, Naresh Manwani
2020 Asian Conference on Machine Learning  
This paper proposes exact passive-aggressive online algorithms for multiclass classification under bandit feedback (EPABF).  ...  In many real-life classification problems, we may not get exact class labels for training samples. One such example is bandit feedback in multiclass classification.  ...  Authors proposed efficient online multiclass linear classification algorithms with bandit feedback when the data is linearly separable by a γ margin.  ... 
dblp:conf/acml/AroraM20 fatcat:q5q33ygl2fdhzmljkiypf3uy4i

Exploiting the Surrogate Gap in Online Multiclass Classification [article]

Dirk van der Hoeven
2021 arXiv   pre-print
We present Gaptron, a randomized first-order algorithm for online multiclass classification.  ...  In the bandit classification setting we show that Gaptron is the first linear time algorithm with O(K√(T)) expected regret, where K is the number of classes.  ...  The author was supported by the Netherlands Organization for Scientific Research (NWO grant TOP2EW.15.211).  ... 
arXiv:2007.12618v2 fatcat:mspcjq2kbjdche77ayleywdtzm

Applying Multiclass Bandit algorithms to call-type classification

Liva Ralaivola, Benoit Favre, Pierre Gotab, Frederic Bechet, Geraldine Damnati
2011 2011 IEEE Workshop on Automatic Speech Recognition & Understanding  
We prove that it is possible to learn linear classifiers in this setting, by estimating adequate expectations inspired by the Multiclass Bandit paradgim.  ...  We analyze the problem of call-type classification using data that is weakly labelled.  ...  More formally, this new algorithm is a linear multiclass perceptron-based classification algorithm that is capable of learning from lazily labelled data.  ... 
doi:10.1109/asru.2011.6163970 dblp:conf/asru/RalaivolaFGBD11 fatcat:dmwgptvjpna6doivx364z5gljq

Multiclass classification with bandit feedback using adaptive regularization

Koby Crammer, Claudio Gentile
2012 Machine Learning  
We present a new multiclass algorithm in the bandit framework, where after making a prediction, the learning algorithm receives only partial feedback, i.e., a single bit indicating whether the predicted  ...  We evaluate our algorithm on nine real-world text classification problems and on four vowel recognition tasks, often obtaining state-of-the-art results, even compared with non-bandit online algorithms,  ...  Acknowledgements We thank the anonymous reviewer for useful comments that helped us to significantly improve the presentation of this paper.  ... 
doi:10.1007/s10994-012-5321-8 fatcat:jgwgenzlcrfn7ivfjwrwoe7siy

A Hierarchical Spectral Method for Extreme Classification [article]

Paul Mineiro, Nikos Karampatziakis
2016 arXiv   pre-print
The end result is a computationally efficient algorithm that provides good statistical performance on several extreme data sets.  ...  Extreme classification problems are multiclass and multilabel classification problems where the number of outputs is so large that straightforward strategies are neither statistically nor computationally  ...  Thus while equation (2) was motivated by analysis for the multiclass case, it is also plausible for multilabel problems.  ... 
arXiv:1511.03260v4 fatcat:gwb366dq3vbudbaduq36e4d3yq

Active Learning for Cost-Sensitive Classification [article]

Akshay Krishnamurthy, Alekh Agarwal, Tzu-Kuo Huang, Hal Daume III, John Langford
2021 arXiv   pre-print
We design an active learning algorithm for cost-sensitive multiclass classification: problems where different errors have different costs.  ...  Our algorithm, COAL, makes predictions by regressing to each label's cost and predicting the smallest.  ...  AK thanks Chicheng Zhang for insightful conversations. AK is supported in part by NSF Award IIS-1763618.  ... 
arXiv:1703.01014v4 fatcat:pnn2cehrijbh7jrgf7jxmaux6u

On multilabel classification and ranking with bandit feedback

Claudio Gentile, Francesco Orabona
2014 Journal of machine learning research  
We analyze this algorithm in a partial adversarial setting, where covariates can be adversarial, but multilabel probabilities are ruled by (generalized) linear models.  ...  The algorithm is based on 2nd-order descent methods, and relies on upper-confidence bounds to trade-off exploration and exploitation.  ...  Acknowledgments We would like to thank the anonymous reviewers for their constructive comments that helped us to improve the presentation of this paper.  ... 
dblp:journals/jmlr/GentileO14 fatcat:ujgtvvoot5cifhcevdkfxxxc2a

Efficient bandit algorithms for online multiclass prediction

Sham M. Kakade, Shai Shalev-Shwartz, Ambuj Tewari
2008 Proceedings of the 25th international conference on Machine learning - ICML '08  
This paper introduces the Banditron, a variant of the Perceptron [Rosenblatt, 1958] , for the multiclass bandit setting.  ...  The Banditron has the ability to learn in a multiclass classification setting with the "bandit" feedback which only reveals whether or not the prediction made by the algorithm was correct or not (but does  ...  This paper provides an efficient bandit algorithm, the Banditron, for multiclass prediction using linear hypothesis spaces, which enjoys a favorable mistake bound.  ... 
doi:10.1145/1390156.1390212 dblp:conf/icml/KakadeST08 fatcat:bjulmu26lvfffd7un4fp7bh3hm

Learning Reductions that Really Work [article]

Alina Beygelzimer, Hal Daumé III, John Langford, Paul Mineiro
2015 arXiv   pre-print
We provide a summary of the mathematical and computational techniques that have enabled learning reductions to effectively address a wide class of problems, and show that this approach to solving machine  ...  Although linear separability with a positive margin implies weak learnability [49], linear separability is still a strong assumption.  ...  Efficient non-reduction techniques exist only for special cases of this problem [35].  ... 
arXiv:1502.02704v1 fatcat:6je6nyymifh47ca455ip3ryy5u

Boosting with Online Binary Learners for the Multiclass Bandit Problem

Shang-Tse Chen, Hsuan-Tien Lin, Chi-Jen Lu
2014 International Conference on Machine Learning  
We consider the problem of online multiclass prediction in the bandit setting.  ...  The approach matches the idea of boosting, which has been shown to be powerful for batch learning as well as online learning.  ...  For the case of multiclass classification, we assume Y = {1, . . . , K}, and for simplicity we adopt the one-versusrest approach to reduce the multiclass problem to a binary one.  ... 
dblp:conf/icml/ChenLL14 fatcat:somhrvb7hvaxfm77l4mca3zyjq

Efficient Online Bandit Multiclass Learning with Õ(√(T)) Regret [article]

Alina Beygelzimer, Francesco Orabona, Chicheng Zhang
2018 arXiv   pre-print
We present an efficient second-order algorithm with Õ(1/η√(T)) regret for the bandit online multiclass problem.  ...  An efficient bandit algorithm for √(T)-regret in online multiclass prediction? In COLT, 2009). We test our algorithm experimentally, showing that it also performs favorably against earlier algorithms.  ...  We also thank the anonymous reviewers for thoughtful comments.  ... 
arXiv:1702.07958v3 fatcat:rfgm4x6zo5hwtoixt4uegr2jke

Learning Multiclass Classifier Under Noisy Bandit Feedback [article]

Mudit Agarwal, Naresh Manwani
2021 arXiv   pre-print
This paper addresses the problem of multiclass classification with corrupted or noisy bandit feedback. In this setting, the learner may not receive true feedback.  ...  The proposed algorithm enjoys a mistake bound of the order of O(√(T)) in the high noise case and of the order of O(T^2/3) in the worst case.  ...  [4] proposed efficient algorithms under bandit feedback when the data is linearly separable by a margin of γ.  ... 
arXiv:2006.03545v2 fatcat:noufr56y45coxlpcl7qg2yxspe
« Previous Showing results 1 — 15 out of 168 results