English

Adaptive Online Estimation of Piecewise Polynomial Trends

Machine Learning 2020-10-02 v1 Optimization and Control Machine Learning

Abstract

We consider the framework of non-stationary stochastic optimization [Besbes et al, 2015] with squared error losses and noisy gradient feedback where the dynamic regret of an online learner against a time varying comparator sequence is studied. Motivated from the theory of non-parametric regression, we introduce a new variational constraint that enforces the comparator sequence to belong to a discrete kthk^{th} order Total Variation ball of radius CnC_n. This variational constraint models comparators that have piece-wise polynomial structure which has many relevant practical applications [Tibshirani, 2014]. By establishing connections to the theory of wavelet based non-parametric regression, we design a polynomial time algorithm that achieves the nearly optimal dynamic regret of O~(n12k+3Cn22k+3)\tilde{O}(n^{\frac{1}{2k+3}}C_n^{\frac{2}{2k+3}}). The proposed policy is adaptive to the unknown radius CnC_n. Further, we show that the same policy is minimax optimal for several other non-parametric families of interest.

Keywords

Cite

@article{arxiv.2010.00073,
  title  = {Adaptive Online Estimation of Piecewise Polynomial Trends},
  author = {Dheeraj Baby and Yu-Xiang Wang},
  journal= {arXiv preprint arXiv:2010.00073},
  year   = {2020}
}