{
  "id": 82675,
  "title": "use different cost function train vs test",
  "url": "/competitions/vsb-power-line-fault-detection/discussion/82675",
  "author_name": "Nicu",
  "post_date": "2019-03-03T14:30:12.729000",
  "votes": 0,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi everyone,</p>\n\n<p>I am curious if anyone is considering or even succeeded to use a different cost function (other than matthews correlation) during training ? </p>\n\n<p>It seems matthews correlation causes a lot of problems in this competition. </p>",
  "messages": [
    {
      "id": 482851,
      "postDate": "2019-03-03T18:35:38.443Z",
      "content": "<p>That is why they picked it as a target metric, so it would not be that easy. ;)</p>",
      "rawMarkdown": "That is why they picked it as a target metric, so it would not be that easy. ;)",
      "replies": [
        {
          "id": 483221,
          "postDate": "2019-03-04T10:35:17.253Z",
          "rawMarkdown": "",
          "votes": -1,
          "isDeleted": true
        }
      ]
    },
    {
      "id": 482696,
      "postDate": "2019-03-03T14:30:12.730Z",
      "content": "<p>Hi everyone,</p>\n\n<p>I am curious if anyone is considering or even succeeded to use a different cost function (other than matthews correlation) during training ? </p>\n\n<p>It seems matthews correlation causes a lot of problems in this competition. </p>",
      "rawMarkdown": "Hi everyone,\n\nI am curious if anyone is considering or even succeeded to use a different cost function (other than matthews correlation) during training ? \n\nIt seems matthews correlation causes a lot of problems in this competition. "
    }
  ],
  "comments": [
    {
      "id": 482851,
      "author_name": "Marcin Pękalski",
      "author_url": "",
      "post_date": "2019-03-03T18:35:38.443000",
      "content": "<p>That is why they picked it as a target metric, so it would not be that easy. ;)</p>",
      "votes": 0,
      "replies": [
        {
          "id": 483221,
          "author_name": "",
          "author_url": "",
          "post_date": "2019-03-04T10:35:17.253000",
          "content": "",
          "votes": -1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "482851": "That is why they picked it as a target metric, so it would not be that easy. ;)",
    "482696": "Hi everyone,\n\nI am curious if anyone is considering or even succeeded to use a different cost function (other than matthews correlation) during training ? \n\nIt seems matthews correlation causes a lot of problems in this competition. "
  }
}