{
  "id": 447481,
  "title": "Question on the validity/usefulness of the train mean/weighted mean submissions here",
  "url": "/competitions/rsna-2023-abdominal-trauma-detection/discussion/447481",
  "author_name": "something4kag",
  "post_date": "2023-10-16T05:11:35.153000",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Early on in the competition there were a number of notebooks that submitted the mean or a weighted mean based on the train data (see Code for examples) and these scored around 0.65 on the public LB.  <br>\nInterestingly (or perhaps not) they seem to score 0.66 on the private LB and seems in the Bronze zone. </p>\n<p>Is this approach really a valid submission especially for private LB? or even ideal in a competition like this?<br>\nMight it have been better not to have made train csvs available or at least empty/dummy files when notebooks are run against hidden test?</p>\n<p>These seem to put a handbrake on any notebooks or ideas that would be using the Dicom images which is  hopefully how anyone would try to determine Trauma injuries.  </p>",
  "messages": [
    {
      "id": 2483914,
      "postDate": "2023-10-16T05:11:35.153Z",
      "content": "<p>Early on in the competition there were a number of notebooks that submitted the mean or a weighted mean based on the train data (see Code for examples) and these scored around 0.65 on the public LB.  <br>\nInterestingly (or perhaps not) they seem to score 0.66 on the private LB and seems in the Bronze zone. </p>\n<p>Is this approach really a valid submission especially for private LB? or even ideal in a competition like this?<br>\nMight it have been better not to have made train csvs available or at least empty/dummy files when notebooks are run against hidden test?</p>\n<p>These seem to put a handbrake on any notebooks or ideas that would be using the Dicom images which is  hopefully how anyone would try to determine Trauma injuries.  </p>",
      "rawMarkdown": "Early on in the competition there were a number of notebooks that submitted the mean or a weighted mean based on the train data (see Code for examples) and these scored around 0.65 on the public LB.  \nInterestingly (or perhaps not) they seem to score 0.66 on the private LB and seems in the Bronze zone. \n\nIs this approach really a valid submission especially for private LB? or even ideal in a competition like this?\nMight it have been better not to have made train csvs available or at least empty/dummy files when notebooks are run against hidden test?\n\nThese seem to put a handbrake on any notebooks or ideas that would be using the Dicom images which is  hopefully how anyone would try to determine Trauma injuries.  \n",
      "votes": 3
    },
    {
      "id": 2483962,
      "postDate": "2023-10-16T06:06:13.953Z",
      "content": "<p>\"Bronze zone\" doesn't make it valid for the clinical perspective. I don't think they are useful at all.</p>",
      "rawMarkdown": "\"Bronze zone\" doesn't make it valid for the clinical perspective. I don't think they are useful at all.",
      "votes": 2
    }
  ],
  "comments": [
    {
      "id": 2483962,
      "author_name": "Gunes Evitan",
      "author_url": "",
      "post_date": "2023-10-16T06:06:13.953000",
      "content": "<p>\"Bronze zone\" doesn't make it valid for the clinical perspective. I don't think they are useful at all.</p>",
      "votes": 2,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2483914": "Early on in the competition there were a number of notebooks that submitted the mean or a weighted mean based on the train data (see Code for examples) and these scored around 0.65 on the public LB.  \nInterestingly (or perhaps not) they seem to score 0.66 on the private LB and seems in the Bronze zone. \n\nIs this approach really a valid submission especially for private LB? or even ideal in a competition like this?\nMight it have been better not to have made train csvs available or at least empty/dummy files when notebooks are run against hidden test?\n\nThese seem to put a handbrake on any notebooks or ideas that would be using the Dicom images which is  hopefully how anyone would try to determine Trauma injuries.  \n",
    "2483962": "\"Bronze zone\" doesn't make it valid for the clinical perspective. I don't think they are useful at all."
  }
}