{
  "id": "1412.6564",
  "version": "v2",
  "published": "2014-12-20T00:31:30.000Z",
  "updated": "2015-04-10T19:03:34.000Z",
  "title": "Move Evaluation in Go Using Deep Convolutional Neural Networks",
  "authors": [
    "Chris J. Maddison",
    "Aja Huang",
    "Ilya Sutskever",
    "David Silver"
  ],
  "comment": "Minor edits and included captures in Figure 2",
  "categories": [
    "cs.LG",
    "cs.NE"
  ],
  "abstract": "The game of Go is more challenging than other board games, due to the difficulty of constructing a position or move evaluation function. In this paper we investigate whether deep convolutional networks can be used to directly represent and learn this knowledge. We train a large 12-layer convolutional neural network by supervised learning from a database of human professional games. The network correctly predicts the expert move in 55% of positions, equalling the accuracy of a 6 dan human player. When the trained convolutional network was used directly to play games of Go, without any search, it beat the traditional search program GnuGo in 97% of games, and matched the performance of a state-of-the-art Monte-Carlo tree search that simulates a million positions per move.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2014-12-20T00:31:30.000Z",
      "comment": null,
      "journal": null,
      "doi": null
    },
    {
      "version": "v2",
      "updated": "2015-04-10T19:03:34.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "deep convolutional neural networks",
      "traditional search program gnugo",
      "state-of-the-art monte-carlo tree search",
      "deep convolutional networks",
      "move evaluation function"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable",
      "adsabs": "2014arXiv1412.6564M"
    }
  }
}