{
  "id": 163479,
  "title": "How to implement focal loss? and why focal loss performs better on this task then binary_crossentropy?",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/163479",
  "author_name": "",
  "post_date": "2020-07-02T08:08:30.508998500Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "",
  "messages": [
    {
      "id": "912093",
      "postDate": "07/02/2020 08:08:30",
      "content": "",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "912119",
      "postDate": "07/02/2020 08:38:06",
      "content": "<p>Focal loss is the reshaping of cross entropy loss such that it down-weights the loss assigned to well-classified examples. Specificaly, adds very less weight to well classified examples and large weight to miss-classified or hard classified examples. Therefore, it provides better results when dealing with imabalanced datasets such as this one.</p>\n\n<p>The implementation depends on the framework you are using. </p>\n\n<p>Focal Loss for Dense Object Detection <a href=\"https://arxiv.org/abs/1708.02002\">Arxiv link</a></p>",
      "rawMarkdown": "Focal loss is the reshaping of cross entropy loss such that it down-weights the loss assigned to well-classified examples. Specificaly, adds very less weight to well classified examples and large weight to miss-classified or hard classified examples. Therefore, it provides better results when dealing with imabalanced datasets such as this one.\n\nThe implementation depends on the framework you are using. \n\nFocal Loss for Dense Object Detection [Arxiv link](https://arxiv.org/abs/1708.02002)",
      "votes": null
    },
    {
      "id": "912216",
      "postDate": "07/02/2020 10:13:20",
      "content": "<p>Thank You . Will surely give it a try. 😃 </p>",
      "rawMarkdown": "Thank You . Will surely give it a try. 😃",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 912119,
      "author_name": "cayala",
      "author_url": "",
      "post_date": "07/02/2020 08:38:06",
      "content": "<p>Focal loss is the reshaping of cross entropy loss such that it down-weights the loss assigned to well-classified examples. Specificaly, adds very less weight to well classified examples and large weight to miss-classified or hard classified examples. Therefore, it provides better results when dealing with imabalanced datasets such as this one.</p>\n\n<p>The implementation depends on the framework you are using. </p>\n\n<p>Focal Loss for Dense Object Detection <a href=\"https://arxiv.org/abs/1708.02002\">Arxiv link</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 912216,
          "author_name": "prashantarorat",
          "author_url": "",
          "post_date": "07/02/2020 10:13:20",
          "content": "<p>Thank You . Will surely give it a try. 😃 </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "912093": "",
    "912119": "Focal loss is the reshaping of cross entropy loss such that it down-weights the loss assigned to well-classified examples. Specificaly, adds very less weight to well classified examples and large weight to miss-classified or hard classified examples. Therefore, it provides better results when dealing with imabalanced datasets such as this one.\n\nThe implementation depends on the framework you are using. \n\nFocal Loss for Dense Object Detection [Arxiv link](https://arxiv.org/abs/1708.02002)",
    "912216": "Thank You . Will surely give it a try. 😃"
  },
  "source": "meta"
}