{
  "id": "1901.00774",
  "version": "v1",
  "published": "2019-01-02T04:48:19.000Z",
  "updated": "2019-01-02T04:48:19.000Z",
  "title": "A New Strategy in Applying the Learning Machine to Study Phase Transitions",
  "authors": [
    "Rongxing Xu",
    "Weicheng Fu",
    "Hong Zhao"
  ],
  "comment": "5 pages, 5 figures",
  "categories": [
    "cond-mat.stat-mech",
    "physics.class-ph"
  ],
  "abstract": "In this Letter, we present a new strategy for applying the learning machine to study phase transitions. We train the learning machine with samples only obtained at a non-critical parameter point, aiming to establish intrinsic correlations between the learning machine and the target system. Then, we find that the accuracy of the learning machine, which is the most important performance index in conventional learning machines, is no longer a key goal of the training in our approach. Instead, relatively low accuracy of identifying unlabeled data category can help to determine the critical point with greater precision, manifesting the singularity around the critical point. It thus provides a robust tool to study the phase transition. The classical ferromagnetic and percolation phase transitions are employed as illustrative examples.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2019-01-02T04:48:19.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "study phase transitions",
      "percolation phase transitions",
      "critical point",
      "important performance index",
      "parameter point"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 5,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}