{
  "id": "1802.04350",
  "version": "v1",
  "published": "2018-02-12T20:31:58.000Z",
  "updated": "2018-02-12T20:31:58.000Z",
  "title": "On the Sample Complexity of Learning from a Sequence of Experiments",
  "authors": [
    "Longyun Guo",
    "Jean Honorio",
    "John Morgan"
  ],
  "comment": "9 pages, 2 figures",
  "categories": [
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "We analyze the sample complexity of a new problem: learning from a sequence of experiments. In this problem, the learner should choose a hypothesis that performs well with respect to an infinite sequence of experiments, and their related data distributions. In practice, the learner can only perform m experiments with a total of N samples drawn from those data distributions. By using a Rademacher complexity approach, we show that the gap between the training and generation error is O((m/N)^0.5). We also provide some examples for linear prediction, two-layer neural networks and kernel methods.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2018-02-12T20:31:58.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "sample complexity",
      "experiments",
      "rademacher complexity approach",
      "two-layer neural networks",
      "kernel methods"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 9,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}