{
  "id": 536835,
  "title": "Not sure if is important or not...",
  "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/536835",
  "author_name": "Ángel Jacinto Sánchez Ruiz",
  "post_date": "2024-09-30T08:11:48.776000",
  "votes": 8,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Cause fast last minut editing to use last daily submission in time I've accidentally submitted a notebook with flipped left/right foraminal predictions. It scored .44 on public, just .02 below not flipped score…</p>\n<p>EDIT: Actually was only .02 under my previous best score. But once fixed the other changes to model and code scored .4, so flip foraminal predictions scored .04 below. Is still not much difference.</p>",
  "messages": [
    {
      "id": 3002601,
      "postDate": "2024-09-30T08:11:48.777Z",
      "content": "<p>Cause fast last minut editing to use last daily submission in time I've accidentally submitted a notebook with flipped left/right foraminal predictions. It scored .44 on public, just .02 below not flipped score…</p>\n<p>EDIT: Actually was only .02 under my previous best score. But once fixed the other changes to model and code scored .4, so flip foraminal predictions scored .04 below. Is still not much difference.</p>",
      "rawMarkdown": "Cause fast last minut editing to use last daily submission in time I've accidentally submitted a notebook with flipped left/right foraminal predictions. It scored .44 on public, just .02 below not flipped score...\n\nEDIT: Actually was only .02 under my previous best score. But once fixed the other changes to model and code scored .4, so flip foraminal predictions scored .04 below. Is still not much difference.",
      "votes": 8
    },
    {
      "id": 3004236,
      "postDate": "2024-10-01T16:37:29.347Z",
      "content": "<p><a href=\"https://www.kaggle.com/sacuscreed\" target=\"_blank\">@sacuscreed</a> </p>\n<p>This is true.<br>\nIn the beginning I did exactly the same and even trained the model with the flipped left and right, the cv did not change much.</p>",
      "rawMarkdown": "@sacuscreed \n\nThis is true.\nIn the beginning I did exactly the same and even trained the model with the flipped left and right, the cv did not change much.",
      "votes": 1
    },
    {
      "id": 3009891,
      "postDate": "2024-10-08T13:40:03.030Z",
      "content": "<p>At one point I thought to create single output for left and right sides for nfn and ss but didn't follow through.<br>\nI think itwill be good regularization as we don't see many instances of single sided case.</p>",
      "rawMarkdown": "At one point I thought to create single output for left and right sides for nfn and ss but didn't follow through.\nI think itwill be good regularization as we don't see many instances of single sided case."
    },
    {
      "id": 3002608,
      "postDate": "2024-09-30T08:29:57.033Z",
      "content": "<p>what is flipped left/right foraminal predictions.</p>",
      "rawMarkdown": "what is flipped left/right foraminal predictions.",
      "replies": [
        {
          "id": 3002648,
          "postDate": "2024-09-30T08:53:08.397Z",
          "content": "<p>If my model sayd:</p>\n<p>study_id_left_neural_foraminal_narrowing_level = A</p>\n<p>study_id_right_neural_foraminal_narrowing_level = B</p>\n<p>I've submitted:</p>\n<p>study_id_left_neural_foraminal_narrowing_level = B</p>\n<p>study_id_right_neural_foraminal_narrowing_level = A</p>",
          "rawMarkdown": "If my model sayd:\n\nstudy_id_left_neural_foraminal_narrowing_level = A\n\nstudy_id_right_neural_foraminal_narrowing_level = B\n\nI've submitted:\n\nstudy_id_left_neural_foraminal_narrowing_level = B\n\nstudy_id_right_neural_foraminal_narrowing_level = A\n",
          "replies": [
            {
              "id": 3002650,
              "postDate": "2024-09-30T08:54:52.387Z",
              "content": "<p>HAHA maybe Overfitting</p>",
              "rawMarkdown": "HAHA maybe Overfitting"
            },
            {
              "id": 3002653,
              "postDate": "2024-09-30T08:58:05.610Z",
              "content": "<p>I was thinking more about at least for public seems like left and right foraminal are strongly correlated since flip his predictions doesn't decrease score much. Not sure about possible overfitthing implication.</p>",
              "rawMarkdown": "I was thinking more about at least for public seems like left and right foraminal are strongly correlated since flip his predictions doesn't decrease score much. Not sure about possible overfitthing implication."
            }
          ]
        }
      ]
    },
    {
      "id": 3002761,
      "postDate": "2024-09-30T11:23:57.757Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 3004236,
      "author_name": "SSS",
      "author_url": "",
      "post_date": "2024-10-01T16:37:29.347000",
      "content": "<p><a href=\"https://www.kaggle.com/sacuscreed\" target=\"_blank\">@sacuscreed</a> </p>\n<p>This is true.<br>\nIn the beginning I did exactly the same and even trained the model with the flipped left and right, the cv did not change much.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 3009891,
      "author_name": "Rohit Chaudhari",
      "author_url": "",
      "post_date": "2024-10-08T13:40:03.030000",
      "content": "<p>At one point I thought to create single output for left and right sides for nfn and ss but didn't follow through.<br>\nI think itwill be good regularization as we don't see many instances of single sided case.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 3002608,
      "author_name": "Corleone",
      "author_url": "",
      "post_date": "2024-09-30T08:29:57.033000",
      "content": "<p>what is flipped left/right foraminal predictions.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 3002648,
          "author_name": "Ángel Jacinto Sánchez Ruiz",
          "author_url": "",
          "post_date": "2024-09-30T08:53:08.397000",
          "content": "<p>If my model sayd:</p>\n<p>study_id_left_neural_foraminal_narrowing_level = A</p>\n<p>study_id_right_neural_foraminal_narrowing_level = B</p>\n<p>I've submitted:</p>\n<p>study_id_left_neural_foraminal_narrowing_level = B</p>\n<p>study_id_right_neural_foraminal_narrowing_level = A</p>",
          "votes": 0,
          "replies": [
            {
              "id": 3002650,
              "author_name": "Corleone",
              "author_url": "",
              "post_date": "2024-09-30T08:54:52.387000",
              "content": "<p>HAHA maybe Overfitting</p>",
              "votes": 0,
              "replies": []
            },
            {
              "id": 3002653,
              "author_name": "Ángel Jacinto Sánchez Ruiz",
              "author_url": "",
              "post_date": "2024-09-30T08:58:05.610000",
              "content": "<p>I was thinking more about at least for public seems like left and right foraminal are strongly correlated since flip his predictions doesn't decrease score much. Not sure about possible overfitthing implication.</p>",
              "votes": 0,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 3002761,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-09-30T11:23:57.757000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3002601": "Cause fast last minut editing to use last daily submission in time I've accidentally submitted a notebook with flipped left/right foraminal predictions. It scored .44 on public, just .02 below not flipped score...\n\nEDIT: Actually was only .02 under my previous best score. But once fixed the other changes to model and code scored .4, so flip foraminal predictions scored .04 below. Is still not much difference.",
    "3004236": "@sacuscreed \n\nThis is true.\nIn the beginning I did exactly the same and even trained the model with the flipped left and right, the cv did not change much.",
    "3009891": "At one point I thought to create single output for left and right sides for nfn and ss but didn't follow through.\nI think itwill be good regularization as we don't see many instances of single sided case.",
    "3002608": "what is flipped left/right foraminal predictions.",
    "3002761": ""
  }
}