{
  "id": "2202.04513",
  "version": "v1",
  "published": "2022-02-09T15:24:30.000Z",
  "updated": "2022-02-09T15:24:30.000Z",
  "title": "The no-free-lunch theorems of supervised learning",
  "authors": [
    "Tom F. Sterkenburg",
    "Peter D. Grünwald"
  ],
  "journal": "Synthese 199:9979-10015 (2021)",
  "doi": "10.1007/s11229-021-03233-1",
  "categories": [
    "cs.LG"
  ],
  "abstract": "The no-free-lunch theorems promote a skeptical conclusion that all possible machine learning algorithms equally lack justification. But how could this leave room for a learning theory, that shows that some algorithms are better than others? Drawing parallels to the philosophy of induction, we point out that the no-free-lunch results presuppose a conception of learning algorithms as purely data-driven. On this conception, every algorithm must have an inherent inductive bias, that wants justification. We argue that many standard learning algorithms should rather be understood as model-dependent: in each application they also require for input a model, representing a bias. Generic algorithms themselves, they can be given a model-relative justification.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2022-02-09T15:24:30.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "supervised learning",
      "no-free-lunch results presuppose",
      "no-free-lunch theorems promote",
      "machine learning algorithms equally lack",
      "learning algorithms equally lack justification"
    ],
    "tags": [
      "journal article"
    ],
    "publication": {
      "publisher": "Springer"
    },
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}