{
  "id": 584540,
  "title": "The non-target is dispensable for evaluation?",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/584540",
  "author_name": "",
  "post_date": "2025-06-14T06:54:26.895387300Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>The evaluation metric measures</p>\n<ol>\n<li>binary-F1, which focus on only target of non-target.</li>\n<li>macro-F1, all non-target actions are collapsed into 1 single label.</li>\n</ol>\n<p>Given the description, I believe without predicting the non-target should give the same score, right? I am testing my hypothesis.</p>",
  "messages": [
    {
      "id": "3223955",
      "postDate": "06/14/2025 06:54:26",
      "content": "<p>The evaluation metric measures</p>\n<ol>\n<li>binary-F1, which focus on only target of non-target.</li>\n<li>macro-F1, all non-target actions are collapsed into 1 single label.</li>\n</ol>\n<p>Given the description, I believe without predicting the non-target should give the same score, right? I am testing my hypothesis.</p>",
      "rawMarkdown": "The evaluation metric measures\n1. binary-F1, which focus on only target of non-target.\n2. macro-F1, all non-target actions are collapsed into 1 single label.\n\nGiven the description, I believe without predicting the non-target should give the same score, right? I am testing my hypothesis.",
      "votes": null
    },
    {
      "id": "3223956",
      "postDate": "06/14/2025 06:54:59",
      "content": "<p>If my hypo is right, there should be a different way to do this comp</p>",
      "rawMarkdown": "If my hypo is right, there should be a different way to do this comp",
      "votes": null
    },
    {
      "id": "3224077",
      "postDate": "06/14/2025 10:50:45",
      "content": "<p>I thought about it too, i was thinking on firstly classifying target/non-target, and after predicting target gestures.</p>",
      "rawMarkdown": "I thought about it too, i was thinking on firstly classifying target/non-target, and after predicting target gestures.",
      "votes": null
    },
    {
      "id": "3224130",
      "postDate": "06/14/2025 12:54:36",
      "content": "<p>tested, the hypothesis works</p>",
      "rawMarkdown": "tested, the hypothesis works",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3223956,
      "author_name": "zhangyue199",
      "author_url": "",
      "post_date": "06/14/2025 06:54:59",
      "content": "<p>If my hypo is right, there should be a different way to do this comp</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3224077,
      "author_name": "eikyou",
      "author_url": "",
      "post_date": "06/14/2025 10:50:45",
      "content": "<p>I thought about it too, i was thinking on firstly classifying target/non-target, and after predicting target gestures.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3224130,
      "author_name": "zhangyue199",
      "author_url": "",
      "post_date": "06/14/2025 12:54:36",
      "content": "<p>tested, the hypothesis works</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3223955": "The evaluation metric measures\n1. binary-F1, which focus on only target of non-target.\n2. macro-F1, all non-target actions are collapsed into 1 single label.\n\nGiven the description, I believe without predicting the non-target should give the same score, right? I am testing my hypothesis.",
    "3223956": "If my hypo is right, there should be a different way to do this comp",
    "3224077": "I thought about it too, i was thinking on firstly classifying target/non-target, and after predicting target gestures.",
    "3224130": "tested, the hypothesis works"
  },
  "source": "meta"
}