{
  "id": 255323,
  "title": "Preventing Model Underfitting on 78% of test set",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/255323",
  "author_name": "",
  "post_date": "2021-07-27T03:43:42.660168500Z",
  "votes": 7,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi fellow participants,<br>\nI find myself troubled with the fact that private submission is based of 22% of the private test set…</p>\n<p>Does this stand a chance of causing a massive shakeup on private LB on the remaining 78%?</p>\n<p>How can I guard against that… <br>\nany advice on this would be helpful..</p>",
  "messages": [
    {
      "id": "1401146",
      "postDate": "07/27/2021 03:43:42",
      "content": "<p>Hi fellow participants,<br>\nI find myself troubled with the fact that private submission is based of 22% of the private test set…</p>\n<p>Does this stand a chance of causing a massive shakeup on private LB on the remaining 78%?</p>\n<p>How can I guard against that… <br>\nany advice on this would be helpful..</p>",
      "rawMarkdown": "Hi fellow participants,\nI find myself troubled with the fact that private submission is based of 22% of the private test set...\n\nDoes this stand a chance of causing a massive shakeup on private LB on the remaining 78%?\n\nHow can I guard against that... \nany advice on this would be helpful..",
      "votes": null
    },
    {
      "id": "1401282",
      "postDate": "07/27/2021 07:15:11",
      "content": "<p>This type of splitting of test data in public and private is common in kaggle.<br>\nThe only thing that most of the kagglers from saving them from a big shakeup is believing in CV and use a variety of models trained with different strategies and ensemble them.</p>",
      "rawMarkdown": "This type of splitting of test data in public and private is common in kaggle.\nThe only thing that most of the kagglers from saving them from a big shakeup is believing in CV and use a variety of models trained with different strategies and ensemble them.",
      "votes": null
    },
    {
      "id": "1401343",
      "postDate": "07/27/2021 08:30:39",
      "content": "<p>So when CV begins to fail or decline but public goes up, and ensembling doesn't seem to work..<br>\nIs there anything else I can try to have a definitive idea on how models will perform on private LB ?. .</p>",
      "rawMarkdown": "So when CV begins to fail or decline but public goes up, and ensembling doesn't seem to work..\nIs there anything else I can try to have a definitive idea on how models will perform on private LB ?. .",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1401282,
      "author_name": "mrinath",
      "author_url": "",
      "post_date": "07/27/2021 07:15:11",
      "content": "<p>This type of splitting of test data in public and private is common in kaggle.<br>\nThe only thing that most of the kagglers from saving them from a big shakeup is believing in CV and use a variety of models trained with different strategies and ensemble them.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1401343,
          "author_name": "troublem1",
          "author_url": "",
          "post_date": "07/27/2021 08:30:39",
          "content": "<p>So when CV begins to fail or decline but public goes up, and ensembling doesn't seem to work..<br>\nIs there anything else I can try to have a definitive idea on how models will perform on private LB ?. .</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1401146": "Hi fellow participants,\nI find myself troubled with the fact that private submission is based of 22% of the private test set...\n\nDoes this stand a chance of causing a massive shakeup on private LB on the remaining 78%?\n\nHow can I guard against that... \nany advice on this would be helpful..",
    "1401282": "This type of splitting of test data in public and private is common in kaggle.\nThe only thing that most of the kagglers from saving them from a big shakeup is believing in CV and use a variety of models trained with different strategies and ensemble them.",
    "1401343": "So when CV begins to fail or decline but public goes up, and ensembling doesn't seem to work..\nIs there anything else I can try to have a definitive idea on how models will perform on private LB ?. ."
  },
  "source": "meta"
}