Online aggregation of unbounded losses using shifting experts with confidence

DOI10.1007/S10994-018-5751-ZzbMATH Open1483.68324arXiv1808.00741OpenAlexW3106524968WikidataQ129375653 ScholiaQ129375653MaRDI QIDQ669288FDOQ669288

Vladimir V. V'yugin, Vladimir Trunov

Publication date: 15 March 2019

Published in: Machine Learning (Search for Journal in Brave)

Abstract: We develop the setting of sequential prediction based on shifting experts and on a "smooth" version of the method of specialized experts. To aggregate experts predictions, we use the AdaHedge algorithm, which is a version of the Hedge algorithm with adaptive learning rate, and extend it by the meta-algorithm Fixed Share. Due to this, we combine the advantages of both algorithms: (1) we use the shifting regret which is a more optimal characteristic of the algorithm; (2) regret bounds are valid in the case of signed unbounded losses of the experts. Also, (3) we incorporate in this scheme a "smooth" version of the method of specialized experts which allows us to make more flexible and accurate predictions. All results are obtained in the adversarial setting -- no assumptions are made about the nature of data source. We present results of numerical experiments for short-term forecasting of electricity consumption based on a real data.

Full work available at URL: https://arxiv.org/abs/1808.00741

Recommendations

• Forecasting electricity consumption by aggregating specialized experts
• Efficient tracking of a growing number of experts
• How to better use expert advice
• A closer look at adaptive regret
• A closer look at adaptive regret

zbMATH Keywords

online learningconfidence leveladaptive learning ratealgorithm hedgemethod of mixing past posteriorsprediction with expert adviceshifting expertsshort-term prediction of electricity consumptionspecialized expertsunbounded losses

Mathematics Subject Classification ID

Applications of statistics in engineering and industry; control charts (62P30) Learning and adaptive systems in artificial intelligence (68T05) Online algorithms; streaming algorithms (68W27)

Cites Work

• A decision-theoretic generalization of on-line learning and an application to boosting
• Forecasting electricity consumption by aggregating specialized experts
• Prediction, Learning, and Games
• Learning Theory
• The weighted majority algorithm
• Tracking the best expert
• 10.1162/153244303321897654
• Title not available (Why is that?)
• A game of prediction with expert advice
• Title not available (Why is that?)
• Title not available (Why is that?)
• A Closer Look at Adaptive Regret
• Derandomizing stochastic prediction strategies
• Prediction with expert evaluators' advice

Uses Software

• AdaBoost.MH

This page was built for publication: Online aggregation of unbounded losses using shifting experts with confidence

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q669288)