Stochastic Online Convex Optimization; Application to probabilistic time series forecasting
Preprint
Abstract
Stochastic regret bounds for online algorithms are usually derived from an ''online to batch'' conversion. Inverting the reasoning, we start our analyze by a ''batch to online'' conversion that applies in any Stochastic Online Convex Optimization problem under stochastic exp-concavity condition. We obtain fast rate stochastic regret bounds with high probability for non-convex loss functions. Based on this approach, we provide prediction and probabilistic forecasting methods for non-stationary unbounded time series.Keywords
All Related Versions
This publication has 12 references indexed in Scilit:
- Kalman recursions Aggregated OnlineStatistical Papers, 2023
- AdaVol: An Adaptive Recursive Volatility Prediction MethodEconometrics and Statistics, 2022
- Online learning with the Continuous Ranked Probability Score for ensemble forecastingQuarterly Journal of the Royal Meteorological Society, 2016
- The Generalization Ability of Online Algorithms for Dependent DataIEEE Transactions on Information Theory, 2012
- Volatility forecast comparison using imperfect volatility proxiesJournal of Econometrics, 2011
- Inconsistency of the MLE and inference based on weighted LS for LARCH modelsJournal of Econometrics, 2010
- Exponential inequalities for self-normalized martingales with applicationsThe Annals of Applied Probability, 2008
- Learning by mirror averagingThe Annals of Statistics, 2008
- Strictly Proper Scoring Rules, Prediction, and EstimationJournal of the American Statistical Association, 2007
- Tracking the Best ExpertMachine Learning, 1998