Filters








736 Hits in 4.9 sec

Kernelized Online Imbalanced Learning with Fixed Budgets

Junjie Hu, Haiqin Yang, Irwin King, Michael Lyu, Anthony Man-Cho So
2015 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
To tackle this problem, we propose a kernelized online imbalanced learning (KOIL) algorithm to directly maximize the area under the ROC curve (AUC).  ...  To this end, we introduce two buffers with fixed budgets (buffer sizes) for positive class and negative class, respectively, to store the learned support vectors, which can allow us to capture the global  ...  Acknowledgments The  ... 
doi:10.1609/aaai.v29i1.9587 fatcat:saktkygiq5gulldbg5bkgggrkm

Online Kernel Selection via Incremental Sketched Kernel Alignment

Xiao Zhang, Shizhong Liao
2018 Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence  
We prove that the ISKA criterion is an unbiased estimate of the maximum mean discrepancy, enjoys the optimal logarithmic regret bound for online kernel learning, and has a constant maintenance time complexity  ...  a sublinear regret bound for online kernel learning, and requiring a constant maintenance time complexity at each round and an efficient overall time complexity integrated with online kernel learning.  ...  Acknowledgments This work was supported in part by the National Natural Science Foundation of China (No. 61673293 and No. 61703396), and the State Key Development Program of China (No. 2017YFE0111900)  ... 
doi:10.24963/ijcai.2018/433 dblp:conf/ijcai/ZhangL18 fatcat:sgay6fqzmjcmliedc45vgzw3f4

Dynamic Weights in Multi-Objective Deep Reinforcement Learning [article]

Axel Abels, Diederik M. Roijers, Tom Lenaerts, Ann Nowé, Denis Steckelmacher
2019 arXiv   pre-print
However, this earlier work is not feasible for RL settings that necessitate the use of function approximators.  ...  with DER dominates these adapted algorithms across weight change scenarios and problem domains.  ...  ., 2016) reduces Qvalue overestimation by using the online network for action selection in the training phase.  ... 
arXiv:1809.07803v2 fatcat:3omaz4akevecdelvqepofg5cfi

Fairness-Aware Online Meta-learning [article]

Chen Zhao, Feng Chen, Bhavani Thuraisingham
2021 arXiv   pre-print
To overcome such issues and bridge the gap, in this paper for the first time we proposed a novel online meta-learning algorithm, namely FFML, which is under the setting of unfairness prevention.  ...  Theoretic analysis provides sub-linear upper bounds for loss regret and for violation of cumulative fairness constraints.  ...  A major drawback of aforementioned methods is that it immerses in minimizing objective functions but ignores the fairness of prediction.  ... 
arXiv:2108.09435v1 fatcat:iuwg7ihb7ngv5dleluisquagqa

Online Learning: A Comprehensive Survey [article]

Steven C.H. Hoi, Doyen Sahoo, Jing Lu, Peilin Zhao
2018 arXiv   pre-print
Generally speaking, according to the learning type and the forms of feedback information, the existing online learning works can be classified into three major categories: (i) supervised online learning  ...  This survey aims to provide a comprehensive survey of the online machine learning literatures through a systematic review of basic ideas and key principles and a proper categorization of different algorithms  ...  Weighted Majority Algorithms The weighted majority algorithm (WM) is a simple but widely studied algorithm that makes a binary prediction based on a series of expert advices Warmuth, 1994, 1989) .  ... 
arXiv:1802.02871v2 fatcat:mqorsb4gknhfhjfb4jcsvbrtwm

Online Variance Reduction with Mixtures [article]

Zalán Borsos, Sebastian Curi, Kfir Y. Levy, Andreas Krause
2019 arXiv   pre-print
In this work, we propose a new framework for variance reduction that enables the use of mixtures over predefined sampling distributions, which can naturally encode prior knowledge about the data.  ...  While these sampling distributions are fixed, the mixture weights are adapted during the optimization process.  ...  K.Y.L. is supported by the ETH Zurich Postdoctoral Fellowship and Marie Curie Actions for People COFUND program.  ... 
arXiv:1903.12416v1 fatcat:avpyaq5v2fa3xavqdx65c7o4je

Predictor-Corrector Policy Optimization [article]

Ching-An Cheng, Xinyan Yan, Nathan Ratliff, Byron Boots
2019 arXiv   pre-print
Unlike previous algorithms, PicCoLO corrects for the mistakes of using imperfect predicted gradients and hence does not suffer from model bias.  ...  to achieve the optimal regret with respect to predictable information.  ...  Particularly, adapt may depend also on whether the problem is weighted, as different base algorithms may handle weighted problems differently.  ... 
arXiv:1810.06509v2 fatcat:ybj3walnzjgzhc2jreoxyf6rbq

Competitive ratio versus regret minimization: achieving the best of both worlds [article]

Amit Daniely, Yishay Mansour
2019 arXiv   pre-print
We then show how to use the extended regret minimization algorithm to combine multiple online algorithms.  ...  For the paging problem we further show an efficient online algorithm (polynomial in the number of pages) with this guarantee.  ...  Using our regret minimization algorithm we can show that for any metrical task system (MTS) there is an online algorithm that guarantees the minimum between: (1) the competitive ratio times the offline  ... 
arXiv:1904.03602v1 fatcat:qm24hvltvvfcfmitr7wao6dnh4

Random-Walk Perturbations for Online Combinatorial Optimization

Luc Devroye, Gabor Lugosi, Gergely Neu
2015 IEEE Transactions on Information Theory  
Further examples of such problems include the online buffering problem described by Geulen, Voecking and Winkler [22] and the online lossy source coding problem of György and Neu [23] .  ...  Perhaps the most popular one is the exponentially weighted average forecaster (a variant of weighted majority algorithm of Littlestone and Warmuth [14] , and aggregating strategies of Vovk [15] , also  ...  Thus, we are left with the problem of bounding Similarly to [20] , we do this by introducing for all u ∈ S and studying the relationship between the distributions p t andp t .  ... 
doi:10.1109/tit.2015.2428253 fatcat:p4rcrr6i4beohmqmccjteff4lm

Online Convex Optimization Against Adversaries with Memory and Application to Statistical Arbitrage [article]

Oren Anava, Elad Hazan, Shie Mannor
2014 arXiv   pre-print
The framework of online learning with memory naturally captures learning problems with temporal constraints, and was previously studied for the experts setting.  ...  In this work we extend the notion of learning with memory to the general Online Convex Optimization (OCO) framework, and present two algorithms that attain low regret.  ...  Regret minimization for online buffering problems using the weighted majority algorithm. In COLT, pages 132-143, 2010. [HAK07] Elad Hazan, Amit Agarwal, and Satyen Kale.  ... 
arXiv:1302.6937v2 fatcat:gw55doyq6jbktnj54fkds72ozq

Competitive ratio vs regret minimization: achieving the best of both worlds

Amit Daniely, Yishay Mansour
2019 International Conference on Algorithmic Learning Theory  
We then show how to use the extended regret minimization algorithm to combine multiple online algorithms.  ...  For the paging problem we further show an efficient online algorithm (polynomial in the number of pages) with this guarantee.  ...  Acknowledgments We thank Eyal Gofer, Alon Gonen and Ohad Shamir for valuable discussions.  ... 
dblp:conf/alt/DanielyM19 fatcat:ge4x7y7vebb3rdyzmy4xozjsj4

CBA: Contextual Quality Adaptation for Adaptive Bitrate Video Streaming (Extended Version) [article]

Bastian Alt, Trevor Ballard, Ralf Steinmetz, Heinz Koeppl, Amr Rizk
2019 arXiv   pre-print
The fundamental problem here is to determine how valuable either information is for the adaptation decision.  ...  We perform an extensive evaluation of our adaptation algorithm in the particularly challenging setting of NDN, where we use an emulation testbed to demonstrate the efficacy of CBA compared to state-of-the-art  ...  (8) The evidence lower bound (ELBO) L(q) is used for solving the optimization problem over the KL divergence (7), since maximizing L(q) is equivalent to minimizing the KL divergence.  ... 
arXiv:1901.05712v1 fatcat:qjcqsmsqunfbbdflybaf4ousea

Privacy-Aware Online Task Offloading for Mobile-Edge Computing [chapter]

Ting Li, Haitao Liu, Jie Liang, Hangsheng Zhang, Liru Geng, Yinlong Liu
2020 Lecture Notes in Computer Science  
Then, we propose a privacy-aware online task offloading (PAOTO) algorithm based on the transformed Thompson sampling (TS) architecture, through which we can (1) receive the best possible delay and energy  ...  Simulation results demonstrate that the proposed scheme outperforms the existing methods in terms of minimizing the system cost and preserving the privacy of users.  ...  Algorithm Design In this section, we focus on the privacy preserving task offloading problem in the MEC-enabled network and propose a device-level privacy-aware online learning scheme to minimize the objective  ... 
doi:10.1007/978-3-030-59016-1_21 fatcat:wmjd24uhurgrdl4xq4udbv6zla

Prediction by Random-Walk Perturbation [article]

Luc Devroye, Gábor Lugosi, Gergely Neu
2013 arXiv   pre-print
We propose a version of the follow-the-perturbed-leader online prediction algorithm in which the cumulative losses are perturbed by independent symmetric random walks.  ...  We also extend the analysis to online combinatorial optimization and show that even in this more general setting, the forecaster rarely switches between experts while having a regret of near-optimal order  ...  Minimizing the regret defined above is a well-studied problem.  ... 
arXiv:1302.5797v1 fatcat:qbzmetu3yzfo5oeaaejyz7g3eu

Memory Bounds for the Experts Problem [article]

Vaidehi Srinivas, David P. Woodruff, Ziyu Xu, Samson Zhou
2022 arXiv   pre-print
An algorithm is judged by how well it does compared to the best expert in the set. The classical algorithm for this problem is the multiplicative weights algorithm.  ...  Our lower bound for i.i.d., random order, and adversarial order streams uses a reduction to a custom-built problem using a novel masking technique, to show a smooth trade-off for regret versus memory.  ...  Acknowledgements We thank Santosh Vempala for pointing out the connection to follow the perturbed leader. David P.  ... 
arXiv:2204.09837v1 fatcat:6mssboorbbcghglgzfp3nfmlji
« Previous Showing results 1 — 15 out of 736 results