{
  "id": "2112.07457",
  "version": "v2",
  "published": "2021-12-14T15:13:31.000Z",
  "updated": "2022-05-20T15:40:15.000Z",
  "title": "Triangulation candidates for Bayesian optimization",
  "authors": [
    "Robert B. Gramacy",
    "Annie Sauer",
    "Nathan Wycoff"
  ],
  "comment": "10 pages, 5 figures",
  "categories": [
    "stat.CO",
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "Bayesian optimization involves \"inner optimization\" over a new-data acquisition criterion which is non-convex/highly multi-modal, may be non-differentiable, or may otherwise thwart local numerical optimizers. In such cases it is common to replace continuous search with a discrete one over random candidates. Here we propose using candidates based on a Delaunay triangulation of the existing input design. We detail the construction of these \"tricands\" and demonstrate empirically how they outperform both numerically optimized acquisitions and random candidate-based alternatives, and are well-suited for hybrid schemes, on benchmark synthetic and real simulation experiments.",
  "revisions": [
    {
      "version": "v2",
      "updated": "2022-05-20T15:40:15.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "bayesian optimization",
      "triangulation candidates",
      "new-data acquisition criterion",
      "thwart local numerical optimizers",
      "real simulation experiments"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 10,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}