{
  "id": 347869,
  "title": "Question: Dice loss",
  "url": "/competitions/hubmap-organ-segmentation/discussion/347869",
  "author_name": "",
  "post_date": "2022-08-25T18:10:22.851984900Z",
  "votes": -1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Dear all, </p>\n<p>I have one question regarding the Dice loss.</p>\n<p>I understand that the Dice coefficient should be close to 1, the larger will be better.</p>\n<p>However, when trainning, my Dice loss get smaller and smaller. May I ask that is this normal, or how to understand such phenomenon?</p>\n<p>Thank you so much!! </p>",
  "messages": [
    {
      "id": "1914094",
      "postDate": "08/25/2022 18:10:22",
      "content": "<p>Dear all, </p>\n<p>I have one question regarding the Dice loss.</p>\n<p>I understand that the Dice coefficient should be close to 1, the larger will be better.</p>\n<p>However, when trainning, my Dice loss get smaller and smaller. May I ask that is this normal, or how to understand such phenomenon?</p>\n<p>Thank you so much!! </p>",
      "rawMarkdown": "Dear all, \n\nI have one question regarding the Dice loss.\n\nI understand that the Dice coefficient should be close to 1, the larger will be better.\n\nHowever, when trainning, my Dice loss get smaller and smaller. May I ask that is this normal, or how to understand such phenomenon?\n\nThank you so much!!",
      "votes": null
    },
    {
      "id": "1914709",
      "postDate": "08/26/2022 10:30:29",
      "content": "<p>in fact DiceLoss = 1 - Dice coefficient</p>",
      "rawMarkdown": "in fact DiceLoss = 1 - Dice coefficient",
      "votes": null
    },
    {
      "id": "1934073",
      "postDate": "09/11/2022 06:11:42",
      "content": "<p><a href=\"https://www.kaggle.com/hubingqian2006\" target=\"_blank\">@hubingqian2006</a> , it is normal. Optimizers are supposed to minimize the loss. And dice loss is constructed as (1 - dice coefficient).</p>",
      "rawMarkdown": "hubingqian2006 , it is normal. Optimizers are supposed to minimize the loss. And dice loss is constructed as (1 - dice coefficient).",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1914709,
      "author_name": "xiafire",
      "author_url": "",
      "post_date": "08/26/2022 10:30:29",
      "content": "<p>in fact DiceLoss = 1 - Dice coefficient</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1934073,
      "author_name": "electro",
      "author_url": "",
      "post_date": "09/11/2022 06:11:42",
      "content": "<p><a href=\"https://www.kaggle.com/hubingqian2006\" target=\"_blank\">@hubingqian2006</a> , it is normal. Optimizers are supposed to minimize the loss. And dice loss is constructed as (1 - dice coefficient).</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1914094": "Dear all, \n\nI have one question regarding the Dice loss.\n\nI understand that the Dice coefficient should be close to 1, the larger will be better.\n\nHowever, when trainning, my Dice loss get smaller and smaller. May I ask that is this normal, or how to understand such phenomenon?\n\nThank you so much!!",
    "1914709": "in fact DiceLoss = 1 - Dice coefficient",
    "1934073": "hubingqian2006 , it is normal. Optimizers are supposed to minimize the loss. And dice loss is constructed as (1 - dice coefficient)."
  },
  "source": "meta"
}