{
  "id": "1406.6812",
  "version": "v1",
  "published": "2014-06-26T08:57:05.000Z",
  "updated": "2014-06-26T08:57:05.000Z",
  "title": "Online learning in MDPs with side information",
  "authors": [
    "Yasin Abbasi-Yadkori",
    "Gergely Neu"
  ],
  "categories": [
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "We study online learning of finite Markov decision process (MDP) problems when a side information vector is available. The problem is motivated by applications such as clinical trials, recommendation systems, etc. Such applications have an episodic structure, where each episode corresponds to a patient/customer. Our objective is to compete with the optimal dynamic policy that can take side information into account. We propose a computationally efficient algorithm and show that its regret is at most $O(\\sqrt{T})$, where $T$ is the number of rounds. To best of our knowledge, this is the first regret bound for this setting.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2014-06-26T08:57:05.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "online learning",
      "finite markov decision process",
      "first regret bound",
      "side information vector",
      "optimal dynamic policy"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable",
      "adsabs": "2014arXiv1406.6812A"
    }
  }
}