{
  "id": 177235,
  "title": "Gentle Reminder on 85% private test set",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/177235",
  "author_name": "",
  "post_date": "2020-08-25T08:21:27.323584800Z",
  "votes": 30,
  "comment_count": 9,
  "views": 0,
  "content": "<p>Developing a robust CV is crucial in this competition, especially including the fact that there are 85% unknown private test set. Try not to overfit and good luck guys 😃</p>",
  "messages": [
    {
      "id": "984689",
      "postDate": "08/25/2020 08:21:27",
      "content": "<p>Developing a robust CV is crucial in this competition, especially including the fact that there are 85% unknown private test set. Try not to overfit and good luck guys 😃</p>",
      "rawMarkdown": "Developing a robust CV is crucial in this competition, especially including the fact that there are 85% unknown private test set. Try not to overfit and good luck guys 😃",
      "votes": null
    },
    {
      "id": "984946",
      "postDate": "08/25/2020 11:38:12",
      "content": "<p>I do agree !!!</p>",
      "rawMarkdown": "I do agree !!!",
      "votes": null
    },
    {
      "id": "985178",
      "postDate": "08/25/2020 14:53:40",
      "content": "<p><a href=\"https://www.kaggle.com/reighns\" target=\"_blank\">@reighns</a>, you are absolutely right. Are you overfitting, or is your best model (making you 6th place currently) strongly CV-backed ;)? I am just curious! </p>",
      "rawMarkdown": "reighns, you are absolutely right. Are you overfitting, or is your best model (making you 6th place currently) strongly CV-backed ;)? I am just curious!",
      "votes": null
    },
    {
      "id": "985771",
      "postDate": "08/26/2020 02:14:45",
      "content": "<p><a href=\"https://www.kaggle.com/chrisden\" target=\"_blank\">@chrisden</a> we might be overfitting slightly, so our aim now is to develop a solid cv. To be honest it’s very scary that we only have 15% of the data. Might have a shakeup worse than melanoma imo. </p>",
      "rawMarkdown": "chrisden we might be overfitting slightly, so our aim now is to develop a solid cv. To be honest it’s very scary that we only have 15% of the data. Might have a shakeup worse than melanoma imo.",
      "votes": null
    },
    {
      "id": "987429",
      "postDate": "08/27/2020 08:47:48",
      "content": "<p>Are your best models able to utilize the scan data or are they based on tabular data. </p>",
      "rawMarkdown": "Are your best models able to utilize the scan data or are they based on tabular data.",
      "votes": null
    },
    {
      "id": "988589",
      "postDate": "08/28/2020 06:22:51",
      "content": "<p>Yes,trust your intuition and experience.</p>",
      "rawMarkdown": "Yes,trust your intuition and experience.",
      "votes": null
    },
    {
      "id": "989427",
      "postDate": "08/28/2020 19:15:58",
      "content": "<p>Not sure where I got this from but are there 200 patients in the test set? Making it only 30 for the private split?</p>",
      "rawMarkdown": "Not sure where I got this from but are there 200 patients in the test set? Making it only 30 for the private split?",
      "votes": null
    },
    {
      "id": "989508",
      "postDate": "08/28/2020 21:20:59",
      "content": "<p>Around 200 patients in the test set, 15% of which are used in the public leaderboard, and the rest for the private one.</p>",
      "rawMarkdown": "Around 200 patients in the test set, 15% of which are used in the public leaderboard, and the rest for the private one.",
      "votes": null
    },
    {
      "id": "992972",
      "postDate": "08/31/2020 14:53:41",
      "content": "<p>How much Dicom files for these 200 patients ??</p>",
      "rawMarkdown": "How much Dicom files for these 200 patients ??",
      "votes": null
    },
    {
      "id": "993776",
      "postDate": "09/01/2020 06:14:36",
      "content": "<p>There is one CT for each patient</p>",
      "rawMarkdown": "There is one CT for each patient",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 984946,
      "author_name": "ulrich07",
      "author_url": "",
      "post_date": "08/25/2020 11:38:12",
      "content": "<p>I do agree !!!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 985178,
      "author_name": "ChristianDenich",
      "author_url": "",
      "post_date": "08/25/2020 14:53:40",
      "content": "<p><a href=\"https://www.kaggle.com/reighns\" target=\"_blank\">@reighns</a>, you are absolutely right. Are you overfitting, or is your best model (making you 6th place currently) strongly CV-backed ;)? I am just curious! </p>",
      "votes": null,
      "replies": [
        {
          "id": 985771,
          "author_name": "reighns",
          "author_url": "",
          "post_date": "08/26/2020 02:14:45",
          "content": "<p><a href=\"https://www.kaggle.com/chrisden\" target=\"_blank\">@chrisden</a> we might be overfitting slightly, so our aim now is to develop a solid cv. To be honest it’s very scary that we only have 15% of the data. Might have a shakeup worse than melanoma imo. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 987429,
          "author_name": "twinklesebastian",
          "author_url": "",
          "post_date": "08/27/2020 08:47:48",
          "content": "<p>Are your best models able to utilize the scan data or are they based on tabular data. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 988589,
      "author_name": "hfutybx",
      "author_url": "",
      "post_date": "08/28/2020 06:22:51",
      "content": "<p>Yes,trust your intuition and experience.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 989427,
      "author_name": "alexandersoare",
      "author_url": "",
      "post_date": "08/28/2020 19:15:58",
      "content": "<p>Not sure where I got this from but are there 200 patients in the test set? Making it only 30 for the private split?</p>",
      "votes": null,
      "replies": [
        {
          "id": 989508,
          "author_name": "ahmedhshahin",
          "author_url": "",
          "post_date": "08/28/2020 21:20:59",
          "content": "<p>Around 200 patients in the test set, 15% of which are used in the public leaderboard, and the rest for the private one.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 992972,
          "author_name": "servietsky",
          "author_url": "",
          "post_date": "08/31/2020 14:53:41",
          "content": "<p>How much Dicom files for these 200 patients ??</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 993776,
          "author_name": "dmytroilin",
          "author_url": "",
          "post_date": "09/01/2020 06:14:36",
          "content": "<p>There is one CT for each patient</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "984689": "Developing a robust CV is crucial in this competition, especially including the fact that there are 85% unknown private test set. Try not to overfit and good luck guys 😃",
    "984946": "I do agree !!!",
    "985178": "reighns, you are absolutely right. Are you overfitting, or is your best model (making you 6th place currently) strongly CV-backed ;)? I am just curious!",
    "985771": "chrisden we might be overfitting slightly, so our aim now is to develop a solid cv. To be honest it’s very scary that we only have 15% of the data. Might have a shakeup worse than melanoma imo.",
    "987429": "Are your best models able to utilize the scan data or are they based on tabular data.",
    "988589": "Yes,trust your intuition and experience.",
    "989427": "Not sure where I got this from but are there 200 patients in the test set? Making it only 30 for the private split?",
    "989508": "Around 200 patients in the test set, 15% of which are used in the public leaderboard, and the rest for the private one.",
    "992972": "How much Dicom files for these 200 patients ??",
    "993776": "There is one CT for each patient"
  },
  "source": "meta"
}