{
  "id": 383629,
  "title": "Unbalanced LB score",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/383629",
  "author_name": "Hyunsoo Lee 1010",
  "post_date": "2023-02-04T11:52:12",
  "votes": 0,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I'm currently using the ConvNeXtV2 model. When I go through the articles on this competition, I see that there are some unbalanced results about the LB score, which was created by the model.<br>\nHow others are different? Your model for LB score is stable? (I heard that EfficientNet is stable for the score)</p>",
  "messages": [
    {
      "id": 2131548,
      "postDate": "2023-02-06T07:46:32.743Z",
      "content": "<p>For the training result of single model, I used different fold data for the same parameter training, and the result was from 0.33 to 0.45. I observed in the discussion board that there were many deviations larger than mine.</p>",
      "rawMarkdown": "For the training result of single model, I used different fold data for the same parameter training, and the result was from 0.33 to 0.45. I observed in the discussion board that there were many deviations larger than mine.",
      "votes": 1
    },
    {
      "id": 2129184,
      "postDate": "2023-02-04T11:52:12Z",
      "content": "<p>I'm currently using the ConvNeXtV2 model. When I go through the articles on this competition, I see that there are some unbalanced results about the LB score, which was created by the model.<br>\nHow others are different? Your model for LB score is stable? (I heard that EfficientNet is stable for the score)</p>",
      "rawMarkdown": "I'm currently using the ConvNeXtV2 model. When I go through the articles on this competition, I see that there are some unbalanced results about the LB score, which was created by the model.\nHow others are different? Your model for LB score is stable? (I heard that EfficientNet is stable for the score)"
    }
  ],
  "comments": [
    {
      "id": 2131548,
      "author_name": "Jijie Li",
      "author_url": "",
      "post_date": "2023-02-06T07:46:32.743000",
      "content": "<p>For the training result of single model, I used different fold data for the same parameter training, and the result was from 0.33 to 0.45. I observed in the discussion board that there were many deviations larger than mine.</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2131548": "For the training result of single model, I used different fold data for the same parameter training, and the result was from 0.33 to 0.45. I observed in the discussion board that there were many deviations larger than mine.",
    "2129184": "I'm currently using the ConvNeXtV2 model. When I go through the articles on this competition, I see that there are some unbalanced results about the LB score, which was created by the model.\nHow others are different? Your model for LB score is stable? (I heard that EfficientNet is stable for the score)"
  }
}