{
  "id": 603679,
  "title": "CV vs LB thread",
  "url": "/competitions/rsna-intracranial-aneurysm-detection/discussion/603679",
  "author_name": "Satwik",
  "post_date": "2025-09-03T16:47:44.983000",
  "votes": 16,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hi all ,  just wanted to see how local CV is comparing to LB for everyone! Here's mine so far:</p>\n<table>\n<thead>\n<tr>\n<th>Model</th>\n<th>CV</th>\n<th>LB</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.64 (avg)</td>\n<td>0.67</td>\n</tr>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.63 (avg)</td>\n<td>0.69</td>\n</tr>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.66(avg)</td>\n<td>0.70</td>\n</tr>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.68 (avg)</td>\n<td>0.68</td>\n</tr>\n</tbody>\n</table>\n<p>will update as I do more experiments!</p>\n<p>Mine does not seem to be very stable. I am calculating the weighted AUC as given in the competition metric code.</p>",
  "messages": [
    {
      "id": 3281117,
      "postDate": "2025-09-03T16:47:44.983Z",
      "content": "<p>Hi all ,  just wanted to see how local CV is comparing to LB for everyone! Here's mine so far:</p>\n<table>\n<thead>\n<tr>\n<th>Model</th>\n<th>CV</th>\n<th>LB</th>\n</tr>\n</thead>\n<tbody>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.64 (avg)</td>\n<td>0.67</td>\n</tr>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.63 (avg)</td>\n<td>0.69</td>\n</tr>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.66(avg)</td>\n<td>0.70</td>\n</tr>\n<tr>\n<td>EffNetV2 (2D) (5-fold)</td>\n<td>0.68 (avg)</td>\n<td>0.68</td>\n</tr>\n</tbody>\n</table>\n<p>will update as I do more experiments!</p>\n<p>Mine does not seem to be very stable. I am calculating the weighted AUC as given in the competition metric code.</p>",
      "rawMarkdown": "Hi all ,  just wanted to see how local CV is comparing to LB for everyone! Here's mine so far:\n\n|Model  | CV  | LB\n| --- | --- | --- |\n| EffNetV2 (2D) (5-fold) | 0.64 (avg) | 0.67 |\n| EffNetV2 (2D) (5-fold) | 0.63 (avg) | 0.69 |\n| EffNetV2 (2D) (5-fold) | 0.66(avg) | 0.70 |\n| EffNetV2 (2D) (5-fold) | 0.68 (avg) | 0.68 |\n\nwill update as I do more experiments!\n\nMine does not seem to be very stable. I am calculating the weighted AUC as given in the competition metric code.\n\n",
      "votes": 16
    },
    {
      "id": 3297290,
      "postDate": "2025-10-02T19:32:30.843Z",
      "content": "<p>Hi satwik, May I ask how u done the train and val split. I have a serious data-leak problem with Effnet<br>\nThanks a lot!</p>",
      "rawMarkdown": "Hi satwik, May I ask how u done the train and val split. I have a serious data-leak problem with Effnet\nThanks a lot!",
      "replies": [
        {
          "id": 3297360,
          "postDate": "2025-10-03T01:09:46.943Z",
          "content": "<p>Hey, I am doing Stratified KFold on Aneurysm Present class.</p>",
          "rawMarkdown": "Hey, I am doing Stratified KFold on Aneurysm Present class.",
          "votes": 1,
          "replies": [
            {
              "id": 3297363,
              "postDate": "2025-10-03T01:31:03.447Z",
              "content": "<p>Thx a lot!</p>",
              "rawMarkdown": "Thx a lot!"
            }
          ]
        }
      ]
    },
    {
      "id": 3281748,
      "postDate": "2025-09-04T20:18:36.473Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 3297290,
      "author_name": "sam_the_rice_cake",
      "author_url": "",
      "post_date": "2025-10-02T19:32:30.843000",
      "content": "<p>Hi satwik, May I ask how u done the train and val split. I have a serious data-leak problem with Effnet<br>\nThanks a lot!</p>",
      "votes": 0,
      "replies": [
        {
          "id": 3297360,
          "author_name": "Satwik",
          "author_url": "",
          "post_date": "2025-10-03T01:09:46.943000",
          "content": "<p>Hey, I am doing Stratified KFold on Aneurysm Present class.</p>",
          "votes": 1,
          "replies": [
            {
              "id": 3297363,
              "author_name": "sam_the_rice_cake",
              "author_url": "",
              "post_date": "2025-10-03T01:31:03.447000",
              "content": "<p>Thx a lot!</p>",
              "votes": 0,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 3281748,
      "author_name": "",
      "author_url": "",
      "post_date": "2025-09-04T20:18:36.473000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3281117": "Hi all ,  just wanted to see how local CV is comparing to LB for everyone! Here's mine so far:\n\n|Model  | CV  | LB\n| --- | --- | --- |\n| EffNetV2 (2D) (5-fold) | 0.64 (avg) | 0.67 |\n| EffNetV2 (2D) (5-fold) | 0.63 (avg) | 0.69 |\n| EffNetV2 (2D) (5-fold) | 0.66(avg) | 0.70 |\n| EffNetV2 (2D) (5-fold) | 0.68 (avg) | 0.68 |\n\nwill update as I do more experiments!\n\nMine does not seem to be very stable. I am calculating the weighted AUC as given in the competition metric code.\n\n",
    "3297290": "Hi satwik, May I ask how u done the train and val split. I have a serious data-leak problem with Effnet\nThanks a lot!",
    "3281748": ""
  }
}