{
  "id": 116334,
  "title": "Why does model giving good score in stage 1 gives poor ",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/116334",
  "author_name": "Jaideep",
  "post_date": "2019-11-08T12:43:54.252000",
  "votes": -1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I get extremly low result from model which was giving decent score on stage1 without any further training. if some one can explain this.</p>",
  "messages": [
    {
      "id": 668513,
      "postDate": "2019-11-08T14:09:24.430Z",
      "content": "<p>Many, many threads on this and Kaggle team have commented many times also. </p>\n\n<p>The public leader board on Stage2 has &lt;1% of the data, it is basically just there so that you can see you have submitted and probably because Kaggle does not have any mechanism to have NO public leader board at all. Stage2 submissions should have no scientific changes from Stage1 submissions so there is no reason for anyone to take meaning from the leaderboard score at all in this stage. </p>\n\n<p>Check out this thread, should make you feel better that your model hasn't completely broken for Stage2 data:</p>\n\n<p><a href=\"https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/116240#latest-668508\">https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/116240#latest-668508</a></p>",
      "rawMarkdown": "Many, many threads on this and Kaggle team have commented many times also. \n\nThe public leader board on Stage2 has &lt;1% of the data, it is basically just there so that you can see you have submitted and probably because Kaggle does not have any mechanism to have NO public leader board at all. Stage2 submissions should have no scientific changes from Stage1 submissions so there is no reason for anyone to take meaning from the leaderboard score at all in this stage. \n\nCheck out this thread, should make you feel better that your model hasn't completely broken for Stage2 data:\n\nhttps://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/116240#latest-668508",
      "votes": 1,
      "replies": [
        {
          "id": 668552,
          "postDate": "2019-11-08T14:51:48.727Z",
          "content": "<p>thanks.. so ranking here starts from bottom  hahahah :)...\nif we are able to see some score submitted that means our new code is able to create right submission</p>",
          "rawMarkdown": "thanks.. so ranking here starts from bottom  hahahah :)...\nif we are able to see some score submitted that means our new code is able to create right submission"
        }
      ]
    },
    {
      "id": 668441,
      "postDate": "2019-11-08T12:43:54.253Z",
      "content": "<p>I get extremly low result from model which was giving decent score on stage1 without any further training. if some one can explain this.</p>",
      "rawMarkdown": "I get extremly low result from model which was giving decent score on stage1 without any further training. if some one can explain this.",
      "votes": -1
    }
  ],
  "comments": [
    {
      "id": 668513,
      "author_name": "cherring",
      "author_url": "",
      "post_date": "2019-11-08T14:09:24.430000",
      "content": "<p>Many, many threads on this and Kaggle team have commented many times also. </p>\n\n<p>The public leader board on Stage2 has &lt;1% of the data, it is basically just there so that you can see you have submitted and probably because Kaggle does not have any mechanism to have NO public leader board at all. Stage2 submissions should have no scientific changes from Stage1 submissions so there is no reason for anyone to take meaning from the leaderboard score at all in this stage. </p>\n\n<p>Check out this thread, should make you feel better that your model hasn't completely broken for Stage2 data:</p>\n\n<p><a href=\"https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/116240#latest-668508\">https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/116240#latest-668508</a></p>",
      "votes": 1,
      "replies": [
        {
          "id": 668552,
          "author_name": "Jaideep",
          "author_url": "",
          "post_date": "2019-11-08T14:51:48.727000",
          "content": "<p>thanks.. so ranking here starts from bottom  hahahah :)...\nif we are able to see some score submitted that means our new code is able to create right submission</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "668513": "Many, many threads on this and Kaggle team have commented many times also. \n\nThe public leader board on Stage2 has &lt;1% of the data, it is basically just there so that you can see you have submitted and probably because Kaggle does not have any mechanism to have NO public leader board at all. Stage2 submissions should have no scientific changes from Stage1 submissions so there is no reason for anyone to take meaning from the leaderboard score at all in this stage. \n\nCheck out this thread, should make you feel better that your model hasn't completely broken for Stage2 data:\n\nhttps://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/116240#latest-668508",
    "668441": "I get extremly low result from model which was giving decent score on stage1 without any further training. if some one can explain this."
  }
}