{
  "id": 181363,
  "title": "LB vs Mean Confidence",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/181363",
  "author_name": "",
  "post_date": "2020-09-08T14:57:07.457422500Z",
  "votes": 7,
  "comment_count": 9,
  "views": 0,
  "content": "<p>It would be interesting to watch how LB varries with mean Confidence that we get for our observations. Although not good indicative but still some hint you get how is your confidence.</p>\n<p>Mine latest lb is 6.8018  Mean confidence around 215 to 270..<br>\nPlease post yours..</p>",
  "messages": [
    {
      "id": "1002958",
      "postDate": "09/08/2020 14:57:07",
      "content": "<p>It would be interesting to watch how LB varries with mean Confidence that we get for our observations. Although not good indicative but still some hint you get how is your confidence.</p>\n<p>Mine latest lb is 6.8018  Mean confidence around 215 to 270..<br>\nPlease post yours..</p>",
      "rawMarkdown": "It would be interesting to watch how LB varries with mean Confidence that we get for our observations. Although not good indicative but still some hint you get how is your confidence.\n\nMine latest lb is 6.8018  Mean confidence around 215 to 270..\nPlease post yours..",
      "votes": null
    },
    {
      "id": "1003787",
      "postDate": "09/09/2020 08:45:56",
      "content": "<p>That's a good point. I agree that Confidence values can give hints about the submission. This is my best submission's stats. I think the min Confidence might be too low compared to yours. Do you do any post processing?</p>\n<pre><code>               FVC  Confidence\ncount   730.000000  730.000000\nmean   2617.753350  262.778623\nstd     394.719185   89.683688\nmin    1856.668213   85.902252\n25%    2528.000580  188.157211\n50%    2683.444824  259.513535\n75%    2867.255890  336.734806\nmax    3281.783691  430.558899\n</code></pre>",
      "rawMarkdown": "That's a good point. I agree that Confidence values can give hints about the submission. This is my best submission's stats. I think the min Confidence might be too low compared to yours. Do you do any post processing?\n\n```\n               FVC  Confidence\ncount   730.000000  730.000000\nmean   2617.753350  262.778623\nstd     394.719185   89.683688\nmin    1856.668213   85.902252\n25%    2528.000580  188.157211\n50%    2683.444824  259.513535\n75%    2867.255890  336.734806\nmax    3281.783691  430.558899\n```",
      "votes": null
    },
    {
      "id": "1003811",
      "postDate": "09/09/2020 09:12:35",
      "content": "<p>thnx dude thats what i was looking for can u tell me your LB score  </p>",
      "rawMarkdown": "thnx dude thats what i was looking for can u tell me your LB score",
      "votes": null
    },
    {
      "id": "1003815",
      "postDate": "09/09/2020 09:18:06",
      "content": "<p>Sorry I forgot to add that. This model scored -6.8255 on public leaderboard.</p>",
      "rawMarkdown": "Sorry I forgot to add that. This model scored -6.8255 on public leaderboard.",
      "votes": null
    },
    {
      "id": "1003854",
      "postDate": "09/09/2020 09:57:56",
      "content": "<p>No not yet.. as there is no generic applicable post processing to adjust confidence i suppose.. it might fail for larger patients group..<br>\nI tried but no results…</p>",
      "rawMarkdown": "No not yet.. as there is no generic applicable post processing to adjust confidence i suppose.. it might fail for larger patients group..\nI tried but no results...",
      "votes": null
    },
    {
      "id": "1004543",
      "postDate": "09/09/2020 19:30:18",
      "content": "<p>That's an interesting idea, thanks. For not being too overrated: if someone has relatively low Confidence it doesn't mean that you will get a good result with this confidence too. Confidence is a type of estimation samples accuracy and decreasing confidence with the same FVC prediction precision may give a negative effect.</p>",
      "rawMarkdown": "That's an interesting idea, thanks. For not being too overrated: if someone has relatively low Confidence it doesn't mean that you will get a good result with this confidence too. Confidence is a type of estimation samples accuracy and decreasing confidence with the same FVC prediction precision may give a negative effect.",
      "votes": null
    },
    {
      "id": "1005209",
      "postDate": "09/10/2020 10:29:04",
      "content": "<p>Yup dont fall prey  to getting some how low confidence in return of high score. </p>",
      "rawMarkdown": "Yup dont fall prey  to getting some how low confidence in return of high score.",
      "votes": null
    },
    {
      "id": "1008239",
      "postDate": "09/12/2020 21:39:24",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4246137%2F7e1f6f6b2d80440d5d445a26e0defc2a%2Fssq.png?generation=1599946759542735&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4246137%2F7e1f6f6b2d80440d5d445a26e0defc2a%2Fssq.png?generation=1599946759542735&alt=media)",
      "votes": null
    },
    {
      "id": "1008413",
      "postDate": "09/13/2020 05:26:59",
      "content": "<p>Hi ,I m expecting one person to join post Cornell competition ,he is my previous mate . I will reconsider if he is not joining <br>\nThanks for your invite </p>",
      "rawMarkdown": "Hi ,I m expecting one person to join post Cornell competition ,he is my previous mate . I will reconsider if he is not joining \nThanks for your invite",
      "votes": null
    },
    {
      "id": "1013750",
      "postDate": "09/17/2020 00:16:49",
      "content": "<p>Hey, <a href=\"https://www.kaggle.com/jaideepvalani\" target=\"_blank\">@jaideepvalani</a>.<br>\nThank you for your reply. If your friend is not going to join you. I would like to merge with you if this is appropriate for you and your team mates</p>",
      "rawMarkdown": "Hey, @jaideepvalani.\nThank you for your reply. If your friend is not going to join you. I would like to merge with you if this is appropriate for you and your team mates",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1003787,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "09/09/2020 08:45:56",
      "content": "<p>That's a good point. I agree that Confidence values can give hints about the submission. This is my best submission's stats. I think the min Confidence might be too low compared to yours. Do you do any post processing?</p>\n<pre><code>               FVC  Confidence\ncount   730.000000  730.000000\nmean   2617.753350  262.778623\nstd     394.719185   89.683688\nmin    1856.668213   85.902252\n25%    2528.000580  188.157211\n50%    2683.444824  259.513535\n75%    2867.255890  336.734806\nmax    3281.783691  430.558899\n</code></pre>",
      "votes": null,
      "replies": [
        {
          "id": 1003811,
          "author_name": "shubhamthapa8",
          "author_url": "",
          "post_date": "09/09/2020 09:12:35",
          "content": "<p>thnx dude thats what i was looking for can u tell me your LB score  </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1003815,
          "author_name": "gunesevitan",
          "author_url": "",
          "post_date": "09/09/2020 09:18:06",
          "content": "<p>Sorry I forgot to add that. This model scored -6.8255 on public leaderboard.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1003854,
          "author_name": "jaideepvalani",
          "author_url": "",
          "post_date": "09/09/2020 09:57:56",
          "content": "<p>No not yet.. as there is no generic applicable post processing to adjust confidence i suppose.. it might fail for larger patients group..<br>\nI tried but no results…</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1004543,
      "author_name": "koza4ukdmitrij",
      "author_url": "",
      "post_date": "09/09/2020 19:30:18",
      "content": "<p>That's an interesting idea, thanks. For not being too overrated: if someone has relatively low Confidence it doesn't mean that you will get a good result with this confidence too. Confidence is a type of estimation samples accuracy and decreasing confidence with the same FVC prediction precision may give a negative effect.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1005209,
          "author_name": "jaideepvalani",
          "author_url": "",
          "post_date": "09/10/2020 10:29:04",
          "content": "<p>Yup dont fall prey  to getting some how low confidence in return of high score. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1008239,
      "author_name": "tauilabdelilah",
      "author_url": "",
      "post_date": "09/12/2020 21:39:24",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4246137%2F7e1f6f6b2d80440d5d445a26e0defc2a%2Fssq.png?generation=1599946759542735&amp;alt=media\" alt=\"\"></p>",
      "votes": null,
      "replies": [
        {
          "id": 1008413,
          "author_name": "jaideepvalani",
          "author_url": "",
          "post_date": "09/13/2020 05:26:59",
          "content": "<p>Hi ,I m expecting one person to join post Cornell competition ,he is my previous mate . I will reconsider if he is not joining <br>\nThanks for your invite </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1013750,
          "author_name": "tauilabdelilah",
          "author_url": "",
          "post_date": "09/17/2020 00:16:49",
          "content": "<p>Hey, <a href=\"https://www.kaggle.com/jaideepvalani\" target=\"_blank\">@jaideepvalani</a>.<br>\nThank you for your reply. If your friend is not going to join you. I would like to merge with you if this is appropriate for you and your team mates</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1002958": "It would be interesting to watch how LB varries with mean Confidence that we get for our observations. Although not good indicative but still some hint you get how is your confidence.\n\nMine latest lb is 6.8018  Mean confidence around 215 to 270..\nPlease post yours..",
    "1003787": "That's a good point. I agree that Confidence values can give hints about the submission. This is my best submission's stats. I think the min Confidence might be too low compared to yours. Do you do any post processing?\n\n```\n               FVC  Confidence\ncount   730.000000  730.000000\nmean   2617.753350  262.778623\nstd     394.719185   89.683688\nmin    1856.668213   85.902252\n25%    2528.000580  188.157211\n50%    2683.444824  259.513535\n75%    2867.255890  336.734806\nmax    3281.783691  430.558899\n```",
    "1003811": "thnx dude thats what i was looking for can u tell me your LB score",
    "1003815": "Sorry I forgot to add that. This model scored -6.8255 on public leaderboard.",
    "1003854": "No not yet.. as there is no generic applicable post processing to adjust confidence i suppose.. it might fail for larger patients group..\nI tried but no results...",
    "1004543": "That's an interesting idea, thanks. For not being too overrated: if someone has relatively low Confidence it doesn't mean that you will get a good result with this confidence too. Confidence is a type of estimation samples accuracy and decreasing confidence with the same FVC prediction precision may give a negative effect.",
    "1005209": "Yup dont fall prey  to getting some how low confidence in return of high score.",
    "1008239": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4246137%2F7e1f6f6b2d80440d5d445a26e0defc2a%2Fssq.png?generation=1599946759542735&alt=media)",
    "1008413": "Hi ,I m expecting one person to join post Cornell competition ,he is my previous mate . I will reconsider if he is not joining \nThanks for your invite",
    "1013750": "Hey, @jaideepvalani.\nThank you for your reply. If your friend is not going to join you. I would like to merge with you if this is appropriate for you and your team mates"
  },
  "source": "meta"
}