{
  "id": 175451,
  "title": "Benchmark for Pulmonary Fibrosis Progression",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/175451",
  "author_name": "Andreas Quauke",
  "post_date": "2020-08-18T08:30:46.099000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I wonder if there is some kind of benchmark for healthcare challenges like this we have to beat. </p>\n<p>For instance, how accurate should our pulmonary fibrosis image classifier be or what is a good value for our modified version of Laplace Log Likelihood?</p>\n<p>In his <a href=\"https://www.kaggle.com/rohanrao/osic-understanding-laplace-log-likelihood\" target=\"_blank\">notebook</a> <a href=\"https://www.kaggle.com/rohanrao\" target=\"_blank\">@rohanrao</a> calculates a minimum result for the target metric:</p>\n<blockquote>\n  <p>-8.023 is the default score to beat while cross-validating models on train data. Any model scoring worse than this is not useful. You can get this default score to beat for each (fold of) validation data as well.</p>\n</blockquote>\n<p>My question is if we have other indications (maybe learned from similar problems in healthcare) for problems like this?</p>",
  "messages": [
    {
      "id": 975299,
      "postDate": "2020-08-18T08:30:46.100Z",
      "content": "<p>I wonder if there is some kind of benchmark for healthcare challenges like this we have to beat. </p>\n<p>For instance, how accurate should our pulmonary fibrosis image classifier be or what is a good value for our modified version of Laplace Log Likelihood?</p>\n<p>In his <a href=\"https://www.kaggle.com/rohanrao/osic-understanding-laplace-log-likelihood\" target=\"_blank\">notebook</a> <a href=\"https://www.kaggle.com/rohanrao\" target=\"_blank\">@rohanrao</a> calculates a minimum result for the target metric:</p>\n<blockquote>\n  <p>-8.023 is the default score to beat while cross-validating models on train data. Any model scoring worse than this is not useful. You can get this default score to beat for each (fold of) validation data as well.</p>\n</blockquote>\n<p>My question is if we have other indications (maybe learned from similar problems in healthcare) for problems like this?</p>",
      "rawMarkdown": "I wonder if there is some kind of benchmark for healthcare challenges like this we have to beat. \n\nFor instance, how accurate should our pulmonary fibrosis image classifier be or what is a good value for our modified version of Laplace Log Likelihood?\n\nIn his [notebook](https://www.kaggle.com/rohanrao/osic-understanding-laplace-log-likelihood) @rohanrao calculates a minimum result for the target metric:\n\n> -8.023 is the default score to beat while cross-validating models on train data. Any model scoring worse than this is not useful. You can get this default score to beat for each (fold of) validation data as well.\n\nMy question is if we have other indications (maybe learned from similar problems in healthcare) for problems like this?",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "975299": "I wonder if there is some kind of benchmark for healthcare challenges like this we have to beat. \n\nFor instance, how accurate should our pulmonary fibrosis image classifier be or what is a good value for our modified version of Laplace Log Likelihood?\n\nIn his [notebook](https://www.kaggle.com/rohanrao/osic-understanding-laplace-log-likelihood) @rohanrao calculates a minimum result for the target metric:\n\n> -8.023 is the default score to beat while cross-validating models on train data. Any model scoring worse than this is not useful. You can get this default score to beat for each (fold of) validation data as well.\n\nMy question is if we have other indications (maybe learned from similar problems in healthcare) for problems like this?"
  }
}