{
  "id": "2007.00939",
  "version": "v1",
  "published": "2020-07-02T07:35:49.000Z",
  "updated": "2020-07-02T07:35:49.000Z",
  "title": "BOSH: Bayesian Optimization by Sampling Hierarchically",
  "authors": [
    "Henry B. Moss",
    "David S. Leslie",
    "Paul Rayson"
  ],
  "categories": [
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "Deployments of Bayesian Optimization (BO) for functions with stochastic evaluations, such as parameter tuning via cross validation and simulation optimization, typically optimize an average of a fixed set of noisy realizations of the objective function. However, disregarding the true objective function in this manner finds a high-precision optimum of the wrong function. To solve this problem, we propose Bayesian Optimization by Sampling Hierarchically (BOSH), a novel BO routine pairing a hierarchical Gaussian process with an information-theoretic framework to generate a growing pool of realizations as the optimization progresses. We demonstrate that BOSH provides more efficient and higher-precision optimization than standard BO across synthetic benchmarks, simulation optimization, reinforcement learning and hyper-parameter tuning tasks.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2020-07-02T07:35:49.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "bayesian optimization",
      "simulation optimization",
      "stochastic evaluations",
      "cross validation",
      "true objective function"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}