{
  "id": 89502,
  "title": "F2 weights recall higher than precision",
  "url": "/competitions/imet-2019-fgvc6/discussion/89502",
  "author_name": "",
  "post_date": "2019-04-15T04:40:24.034229400Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Where did you come up with 2 for Beta, valuing the recall higher than the precision here?  To put it in an analogy for this classification, is it better having increased false alarms to catch all the shoplifters?</p>",
  "messages": [
    {
      "id": "516847",
      "postDate": "04/15/2019 04:40:24",
      "content": "<p>Where did you come up with 2 for Beta, valuing the recall higher than the precision here?  To put it in an analogy for this classification, is it better having increased false alarms to catch all the shoplifters?</p>",
      "rawMarkdown": "Where did you come up with 2 for Beta, valuing the recall higher than the precision here?  To put it in an analogy for this classification, is it better having increased false alarms to catch all the shoplifters?",
      "votes": null
    },
    {
      "id": "516871",
      "postDate": "04/15/2019 05:53:53",
      "content": "<p>In fbeta-score, if beta=2 , it becomes F2 score.\nIf beta=2,  the importance of recall is higher than precision. For example, I would rather incorrectly add one more label than leave out the possible label. So we can see many kernel set  0.28 as the threshold rather than 0.50.</p>",
      "rawMarkdown": "In fbeta-score, if beta=2 , it becomes F2 score.\nIf beta=2,  the importance of recall is higher than precision. For example, I would rather incorrectly add one more label than leave out the possible label. So we can see many kernel set  0.28 as the threshold rather than 0.50.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 516871,
      "author_name": "seefun",
      "author_url": "",
      "post_date": "04/15/2019 05:53:53",
      "content": "<p>In fbeta-score, if beta=2 , it becomes F2 score.\nIf beta=2,  the importance of recall is higher than precision. For example, I would rather incorrectly add one more label than leave out the possible label. So we can see many kernel set  0.28 as the threshold rather than 0.50.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "516847": "Where did you come up with 2 for Beta, valuing the recall higher than the precision here?  To put it in an analogy for this classification, is it better having increased false alarms to catch all the shoplifters?",
    "516871": "In fbeta-score, if beta=2 , it becomes F2 score.\nIf beta=2,  the importance of recall is higher than precision. For example, I would rather incorrectly add one more label than leave out the possible label. So we can see many kernel set  0.28 as the threshold rather than 0.50."
  },
  "source": "meta"
}