{
  "id": 94319,
  "title": "self-defeating blending",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/94319",
  "author_name": "",
  "post_date": "2019-06-04T00:59:17.490305500Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>My model's public lb = 1.47957, so I merged my prediction with <a href=\"https://www.kaggle.com/tocha4/lanl-master-s-approach\">the great kernel</a>, and my public lb goes from 1.47957 to 1.42195, but private lb goes from 2.48248 to 2.51561. Unfortunately i use the latter for Final Score.\nMy best model's public lb = 1.57065, not a high score, but its private lb = 2.46863.  I should trust my own val_mae.</p>",
  "messages": [
    {
      "id": "542532",
      "postDate": "06/04/2019 00:59:17",
      "content": "<p>My model's public lb = 1.47957, so I merged my prediction with <a href=\"https://www.kaggle.com/tocha4/lanl-master-s-approach\">the great kernel</a>, and my public lb goes from 1.47957 to 1.42195, but private lb goes from 2.48248 to 2.51561. Unfortunately i use the latter for Final Score.\nMy best model's public lb = 1.57065, not a high score, but its private lb = 2.46863.  I should trust my own val_mae.</p>",
      "rawMarkdown": "My model's public lb = 1.47957, so I merged my prediction with [the great kernel](https://www.kaggle.com/tocha4/lanl-master-s-approach), and my public lb goes from 1.47957 to 1.42195, but private lb goes from 2.48248 to 2.51561. Unfortunately i use the latter for Final Score.\nMy best model's public lb = 1.57065, not a high score, but its private lb = 2.46863.  I should trust my own val_mae.",
      "votes": null
    },
    {
      "id": "542576",
      "postDate": "06/04/2019 01:41:43",
      "content": "<p>Never trust completely in the Public LB. Build a good local validation strategy and trust on it.</p>",
      "rawMarkdown": "Never trust completely in the Public LB. Build a good local validation strategy and trust on it.",
      "votes": null
    },
    {
      "id": "542617",
      "postDate": "06/04/2019 02:11:27",
      "content": "<p>Thank you. Yes, that's a bloody lesson. I just couldn't resist the public score. </p>",
      "rawMarkdown": "Thank you. Yes, that's a bloody lesson. I just couldn't resist the public score.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 542576,
      "author_name": "titericz",
      "author_url": "",
      "post_date": "06/04/2019 01:41:43",
      "content": "<p>Never trust completely in the Public LB. Build a good local validation strategy and trust on it.</p>",
      "votes": null,
      "replies": [
        {
          "id": 542617,
          "author_name": "jiangjixiang",
          "author_url": "",
          "post_date": "06/04/2019 02:11:27",
          "content": "<p>Thank you. Yes, that's a bloody lesson. I just couldn't resist the public score. </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "542532": "My model's public lb = 1.47957, so I merged my prediction with [the great kernel](https://www.kaggle.com/tocha4/lanl-master-s-approach), and my public lb goes from 1.47957 to 1.42195, but private lb goes from 2.48248 to 2.51561. Unfortunately i use the latter for Final Score.\nMy best model's public lb = 1.57065, not a high score, but its private lb = 2.46863.  I should trust my own val_mae.",
    "542576": "Never trust completely in the Public LB. Build a good local validation strategy and trust on it.",
    "542617": "Thank you. Yes, that's a bloody lesson. I just couldn't resist the public score."
  },
  "source": "meta"
}