{
  "id": 64894,
  "title": "0.353 Simple baseline model = Radiologist ?",
  "url": "/competitions/rsna-pneumonia-detection-challenge/discussion/64894",
  "author_name": "",
  "post_date": "2018-09-04T01:24:35.785763200Z",
  "votes": 12,
  "comment_count": 8,
  "views": 0,
  "content": "<p>After one week with 458 teams, LB is still far away from the 0.353 \"Simple\" baseline model.</p>\n\n<p>Is there a possibility that this \"Simple\" model is actually a radiologist like the leading score in Stanford MURA competition (<a href=\"https://stanfordmlgroup.github.io/competitions/mura/\">https://stanfordmlgroup.github.io/competitions/mura/</a>) ?</p>\n\n<p>This could be a cool challenge narrative by RSNA.</p>",
  "messages": [
    {
      "id": "381041",
      "postDate": "09/04/2018 01:24:35",
      "content": "<p>After one week with 458 teams, LB is still far away from the 0.353 \"Simple\" baseline model.</p>\n\n<p>Is there a possibility that this \"Simple\" model is actually a radiologist like the leading score in Stanford MURA competition (<a href=\"https://stanfordmlgroup.github.io/competitions/mura/\">https://stanfordmlgroup.github.io/competitions/mura/</a>) ?</p>\n\n<p>This could be a cool challenge narrative by RSNA.</p>",
      "rawMarkdown": "After one week with 458 teams, LB is still far away from the 0.353 \"Simple\" baseline model.\n\nIs there a possibility that this \"Simple\" model is actually a radiologist like the leading score in Stanford MURA competition (https://stanfordmlgroup.github.io/competitions/mura/) ?\n\nThis could be a cool challenge narrative by RSNA.",
      "votes": null
    },
    {
      "id": "381116",
      "postDate": "09/04/2018 04:37:38",
      "content": "<p>@Alexandre, thanks for this link, it looks interesting. I will definitely take a closer look at it when I have time to work on the RSNA data properly in a couple of weeks or so.</p>",
      "rawMarkdown": "Alexandre, thanks for this link, it looks interesting. I will definitely take a closer look at it when I have time to work on the RSNA data properly in a couple of weeks or so.",
      "votes": null
    },
    {
      "id": "381306",
      "postDate": "09/04/2018 13:00:08",
      "content": "<p>Is this the correct link?: <a href=\"https://stanfordmlgroup.github.io/projects/chexnet/\">https://stanfordmlgroup.github.io/projects/chexnet/</a> </p>",
      "rawMarkdown": "Is this the correct link?: https://stanfordmlgroup.github.io/projects/chexnet/",
      "votes": null
    },
    {
      "id": "381319",
      "postDate": "09/04/2018 13:20:34",
      "content": "<p>MURA (upper extremity dataset, model and competition) and ChexNet (chest xrays model trained on NIH dataset) are two different Stanford projects from the Stanford ML group. As specified in original post, I was refering to the MURA project because they are reporting a radiologist score as the leader in their ML competition.</p>\n\n<p>ChexNet is interesting of course, but it was not the link that I was refering.</p>",
      "rawMarkdown": "MURA (upper extremity dataset, model and competition) and ChexNet (chest xrays model trained on NIH dataset) are two different Stanford projects from the Stanford ML group. As specified in original post, I was refering to the MURA project because they are reporting a radiologist score as the leader in their ML competition.\n\nChexNet is interesting of course, but it was not the link that I was refering.",
      "votes": null
    },
    {
      "id": "381331",
      "postDate": "09/04/2018 13:35:17",
      "content": "<p>That would be very cool Alex.  Nice observaton.</p>",
      "rawMarkdown": "That would be very cool Alex.  Nice observaton.",
      "votes": null
    },
    {
      "id": "381428",
      "postDate": "09/04/2018 16:22:06",
      "content": "<p>I don't think that Model 0.353 is simple. :)</p>",
      "rawMarkdown": "I don't think that Model 0.353 is simple. :)",
      "votes": null
    },
    {
      "id": "381430",
      "postDate": "09/04/2018 16:23:50",
      "content": "<p>ChexNet  is the 121-layer convolutional neural network... Cool!</p>",
      "rawMarkdown": "ChexNet  is the 121-layer convolutional neural network... Cool!",
      "votes": null
    },
    {
      "id": "383136",
      "postDate": "09/07/2018 18:59:13",
      "content": "<p>This line 0.353 is disappeared. :)</p>",
      "rawMarkdown": "This line 0.353 is disappeared. :)",
      "votes": null
    },
    {
      "id": "383140",
      "postDate": "09/07/2018 19:17:46",
      "content": "<p>Thanks for the info ! Fortunately, I should be able to sleep again without being haunted by the simple model benchmark!  ... or maybe not for too long when the simplicity doubt will come back in the next weeks.</p>",
      "rawMarkdown": "Thanks for the info ! Fortunately, I should be able to sleep again without being haunted by the simple model benchmark!  ... or maybe not for too long when the simplicity doubt will come back in the next weeks.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 381116,
      "author_name": "sheriytm",
      "author_url": "",
      "post_date": "09/04/2018 04:37:38",
      "content": "<p>@Alexandre, thanks for this link, it looks interesting. I will definitely take a closer look at it when I have time to work on the RSNA data properly in a couple of weeks or so.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 381306,
      "author_name": "paulorzp",
      "author_url": "",
      "post_date": "09/04/2018 13:00:08",
      "content": "<p>Is this the correct link?: <a href=\"https://stanfordmlgroup.github.io/projects/chexnet/\">https://stanfordmlgroup.github.io/projects/chexnet/</a> </p>",
      "votes": null,
      "replies": [
        {
          "id": 381319,
          "author_name": "alexandrecc",
          "author_url": "",
          "post_date": "09/04/2018 13:20:34",
          "content": "<p>MURA (upper extremity dataset, model and competition) and ChexNet (chest xrays model trained on NIH dataset) are two different Stanford projects from the Stanford ML group. As specified in original post, I was refering to the MURA project because they are reporting a radiologist score as the leader in their ML competition.</p>\n\n<p>ChexNet is interesting of course, but it was not the link that I was refering.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 381430,
          "author_name": "sergeyzlobin",
          "author_url": "",
          "post_date": "09/04/2018 16:23:50",
          "content": "<p>ChexNet  is the 121-layer convolutional neural network... Cool!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 381331,
      "author_name": "drsxr1",
      "author_url": "",
      "post_date": "09/04/2018 13:35:17",
      "content": "<p>That would be very cool Alex.  Nice observaton.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 381428,
      "author_name": "sergeyzlobin",
      "author_url": "",
      "post_date": "09/04/2018 16:22:06",
      "content": "<p>I don't think that Model 0.353 is simple. :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 383136,
      "author_name": "sergeyzlobin",
      "author_url": "",
      "post_date": "09/07/2018 18:59:13",
      "content": "<p>This line 0.353 is disappeared. :)</p>",
      "votes": null,
      "replies": [
        {
          "id": 383140,
          "author_name": "alexandrecc",
          "author_url": "",
          "post_date": "09/07/2018 19:17:46",
          "content": "<p>Thanks for the info ! Fortunately, I should be able to sleep again without being haunted by the simple model benchmark!  ... or maybe not for too long when the simplicity doubt will come back in the next weeks.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "381041": "After one week with 458 teams, LB is still far away from the 0.353 \"Simple\" baseline model.\n\nIs there a possibility that this \"Simple\" model is actually a radiologist like the leading score in Stanford MURA competition (https://stanfordmlgroup.github.io/competitions/mura/) ?\n\nThis could be a cool challenge narrative by RSNA.",
    "381116": "Alexandre, thanks for this link, it looks interesting. I will definitely take a closer look at it when I have time to work on the RSNA data properly in a couple of weeks or so.",
    "381306": "Is this the correct link?: https://stanfordmlgroup.github.io/projects/chexnet/",
    "381319": "MURA (upper extremity dataset, model and competition) and ChexNet (chest xrays model trained on NIH dataset) are two different Stanford projects from the Stanford ML group. As specified in original post, I was refering to the MURA project because they are reporting a radiologist score as the leader in their ML competition.\n\nChexNet is interesting of course, but it was not the link that I was refering.",
    "381331": "That would be very cool Alex.  Nice observaton.",
    "381428": "I don't think that Model 0.353 is simple. :)",
    "381430": "ChexNet  is the 121-layer convolutional neural network... Cool!",
    "383136": "This line 0.353 is disappeared. :)",
    "383140": "Thanks for the info ! Fortunately, I should be able to sleep again without being haunted by the simple model benchmark!  ... or maybe not for too long when the simplicity doubt will come back in the next weeks."
  },
  "source": "meta"
}