{
  "id": 233414,
  "title": "An appropriate threshold seems important in this competition",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/233414",
  "author_name": "",
  "post_date": "2021-04-19T06:29:14.993165900Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I tried threshold 0.28 - 0.6, the LB score varies from 0.902 - 0.912, can anyone share the experience about threshold tunning?</p>",
  "messages": [
    {
      "id": "1277737",
      "postDate": "04/19/2021 06:29:14",
      "content": "<p>I tried threshold 0.28 - 0.6, the LB score varies from 0.902 - 0.912, can anyone share the experience about threshold tunning?</p>",
      "rawMarkdown": "I tried threshold 0.28 - 0.6, the LB score varies from 0.902 - 0.912, can anyone share the experience about threshold tunning?",
      "votes": null
    },
    {
      "id": "1277943",
      "postDate": "04/19/2021 11:50:59",
      "content": "<p>Hello,is 0.6 better than 0.28?</p>",
      "rawMarkdown": "Hello,is 0.6 better than 0.28?",
      "votes": null
    },
    {
      "id": "1278502",
      "postDate": "04/20/2021 01:17:17",
      "content": "<p>There is no silver bullet for the threshold. Results may vary, based on the trained model, etc. If you really want to adjust the threshold, you may consider doing it dynamically in the post processing phase…</p>",
      "rawMarkdown": "There is no silver bullet for the threshold. Results may vary, based on the trained model, etc. If you really want to adjust the threshold, you may consider doing it dynamically in the post processing phase...",
      "votes": null
    },
    {
      "id": "1279261",
      "postDate": "04/20/2021 18:40:25",
      "content": "<p>How would you do that? run the inference multiple times and pick the best one?</p>",
      "rawMarkdown": "How would you do that? run the inference multiple times and pick the best one?",
      "votes": null
    },
    {
      "id": "1279264",
      "postDate": "04/20/2021 18:41:42",
      "content": "<p>Did you implemented some sort of grid search for the threshold? Or youa are hardcoding it?</p>",
      "rawMarkdown": "Did you implemented some sort of grid search for the threshold? Or youa are hardcoding it?",
      "votes": null
    },
    {
      "id": "1279540",
      "postDate": "04/21/2021 02:52:42",
      "content": "<p>I hardcode it</p>",
      "rawMarkdown": "I hardcode it",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1277943,
      "author_name": "zekunn",
      "author_url": "",
      "post_date": "04/19/2021 11:50:59",
      "content": "<p>Hello,is 0.6 better than 0.28?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1278502,
      "author_name": "andrasferenczi",
      "author_url": "",
      "post_date": "04/20/2021 01:17:17",
      "content": "<p>There is no silver bullet for the threshold. Results may vary, based on the trained model, etc. If you really want to adjust the threshold, you may consider doing it dynamically in the post processing phase…</p>",
      "votes": null,
      "replies": [
        {
          "id": 1279261,
          "author_name": "victorasso",
          "author_url": "",
          "post_date": "04/20/2021 18:40:25",
          "content": "<p>How would you do that? run the inference multiple times and pick the best one?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1279264,
      "author_name": "victorasso",
      "author_url": "",
      "post_date": "04/20/2021 18:41:42",
      "content": "<p>Did you implemented some sort of grid search for the threshold? Or youa are hardcoding it?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1279540,
          "author_name": "plugin1689",
          "author_url": "",
          "post_date": "04/21/2021 02:52:42",
          "content": "<p>I hardcode it</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1277737": "I tried threshold 0.28 - 0.6, the LB score varies from 0.902 - 0.912, can anyone share the experience about threshold tunning?",
    "1277943": "Hello,is 0.6 better than 0.28?",
    "1278502": "There is no silver bullet for the threshold. Results may vary, based on the trained model, etc. If you really want to adjust the threshold, you may consider doing it dynamically in the post processing phase...",
    "1279261": "How would you do that? run the inference multiple times and pick the best one?",
    "1279264": "Did you implemented some sort of grid search for the threshold? Or youa are hardcoding it?",
    "1279540": "I hardcode it"
  },
  "source": "meta"
}