{
  "id": "1803.05814",
  "version": "v1",
  "published": "2018-03-15T15:37:40.000Z",
  "updated": "2018-03-15T15:37:40.000Z",
  "title": "Theory and Algorithms for Forecasting Time Series",
  "authors": [
    "Vitaly Kuznetsov",
    "Mehryar Mohri"
  ],
  "comment": "An extended abstract has appeared in (Kuznetsov and Mohri, 2015)",
  "categories": [
    "cs.LG"
  ],
  "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 also also provide novel analysis of stable time series forecasting algorithm using this new notion of discrepancy that we introduce. We use our learning bounds to devise new algorithms for non-stationary time series forecasting for which we report some preliminary experimental results.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2018-03-15T15:37:40.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "forecasting time series",
      "preliminary experimental results",
      "non-stationary non-mixing stochastic processes",
      "stable time series forecasting algorithm",
      "learning bounds"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}