{
  "id": 533630,
  "title": "Repair NaN data",
  "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/533630",
  "author_name": "",
  "post_date": "2024-09-12T07:31:25.318735600Z",
  "votes": null,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I am attempt to fill NaN labels with previous model's preds. Is this meaningful for the metric? Was there anybody trying this way?</p>",
  "messages": [
    {
      "id": "2986934",
      "postDate": "09/12/2024 07:31:25",
      "content": "<p>I am attempt to fill NaN labels with previous model's preds. Is this meaningful for the metric? Was there anybody trying this way?</p>",
      "rawMarkdown": "I am attempt to fill NaN labels with previous model's preds. Is this meaningful for the metric? Was there anybody trying this way?",
      "votes": null
    },
    {
      "id": "2986977",
      "postDate": "09/12/2024 08:53:20",
      "content": "<p>There is not many NaNs in my opinion. I didn't tryed, but I don't spect a big difference.</p>",
      "rawMarkdown": "There is not many NaNs in my opinion. I didn't tryed, but I don't spect a big difference.",
      "votes": null
    },
    {
      "id": "2989719",
      "postDate": "09/15/2024 13:04:44",
      "content": "<p>I'm not sure but I think they are not annotated because the condition was not visible. In that case, it would be better to ignore them on loss calculation.</p>",
      "rawMarkdown": "I'm not sure but I think they are not annotated because the condition was not visible. In that case, it would be better to ignore them on loss calculation.",
      "votes": null
    },
    {
      "id": "2993269",
      "postDate": "09/19/2024 15:38:54",
      "content": "<p>I have tried with ignoring nans and considering nans as normal_mild.<br>\nDidn't see any difference in CV and LB score for both.<br>\nI am using single stage approach.</p>",
      "rawMarkdown": "I have tried with ignoring nans and considering nans as normal_mild.\nDidn't see any difference in CV and LB score for both.\nI am using single stage approach.",
      "votes": null
    },
    {
      "id": "2996072",
      "postDate": "09/23/2024 03:20:15",
      "content": "<p>I exprimented that adding NaNs is a bad idea😂</p>",
      "rawMarkdown": "I exprimented that adding NaNs is a bad idea😂",
      "votes": null
    },
    {
      "id": "2996073",
      "postDate": "09/23/2024 03:21:16",
      "content": "<p>My result shows you are right</p>",
      "rawMarkdown": "My result shows you are right",
      "votes": null
    },
    {
      "id": "2996074",
      "postDate": "09/23/2024 03:22:19",
      "content": "<p>I tried both too and found it better to ignore nans</p>",
      "rawMarkdown": "I tried both too and found it better to ignore nans",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2986977,
      "author_name": "sacuscreed",
      "author_url": "",
      "post_date": "09/12/2024 08:53:20",
      "content": "<p>There is not many NaNs in my opinion. I didn't tryed, but I don't spect a big difference.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2996072,
          "author_name": "xukongji",
          "author_url": "",
          "post_date": "09/23/2024 03:20:15",
          "content": "<p>I exprimented that adding NaNs is a bad idea😂</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2989719,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "09/15/2024 13:04:44",
      "content": "<p>I'm not sure but I think they are not annotated because the condition was not visible. In that case, it would be better to ignore them on loss calculation.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2996073,
          "author_name": "xukongji",
          "author_url": "",
          "post_date": "09/23/2024 03:21:16",
          "content": "<p>My result shows you are right</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2993269,
      "author_name": "rohitchaudhari25",
      "author_url": "",
      "post_date": "09/19/2024 15:38:54",
      "content": "<p>I have tried with ignoring nans and considering nans as normal_mild.<br>\nDidn't see any difference in CV and LB score for both.<br>\nI am using single stage approach.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2996074,
          "author_name": "xukongji",
          "author_url": "",
          "post_date": "09/23/2024 03:22:19",
          "content": "<p>I tried both too and found it better to ignore nans</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2986934": "I am attempt to fill NaN labels with previous model's preds. Is this meaningful for the metric? Was there anybody trying this way?",
    "2986977": "There is not many NaNs in my opinion. I didn't tryed, but I don't spect a big difference.",
    "2989719": "I'm not sure but I think they are not annotated because the condition was not visible. In that case, it would be better to ignore them on loss calculation.",
    "2993269": "I have tried with ignoring nans and considering nans as normal_mild.\nDidn't see any difference in CV and LB score for both.\nI am using single stage approach.",
    "2996072": "I exprimented that adding NaNs is a bad idea😂",
    "2996073": "My result shows you are right",
    "2996074": "I tried both too and found it better to ignore nans"
  },
  "source": "meta"
}