Learning Theory and Algorithms for Forecasting Non-Stationary Time Series

Abstract

We present data-dependent learning bounds for the general scenario of non-stationary non-mixing stochastic processes. Our learning guarantees are expressed in terms of a data-dependent measure of sequential complexity and a discrepancy measure that can be estimated from data under some mild assumptions. We use our learning bounds to devise new algorithms for non-stationary time series forecasting for which we report some preliminary experimental results.

Cite

Text

Kuznetsov and Mohri. "Learning Theory and Algorithms for Forecasting Non-Stationary Time Series." Neural Information Processing Systems, 2015.

Markdown

[Kuznetsov and Mohri. "Learning Theory and Algorithms for Forecasting Non-Stationary Time Series." Neural Information Processing Systems, 2015.](https://mlanthology.org/neurips/2015/kuznetsov2015neurips-learning/)

BibTeX

@inproceedings{kuznetsov2015neurips-learning,
  title     = {{Learning Theory and Algorithms for Forecasting Non-Stationary Time Series}},
  author    = {Kuznetsov, Vitaly and Mohri, Mehryar},
  booktitle = {Neural Information Processing Systems},
  year      = {2015},
  pages     = {541-549},
  url       = {https://mlanthology.org/neurips/2015/kuznetsov2015neurips-learning/}
}