{
  "id": 143658,
  "title": "Loss function to use?",
  "url": "/competitions/imet-2020-fgvc7/discussion/143658",
  "author_name": "",
  "post_date": "2020-04-15T18:44:20.748132400Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I have seen some kernels use categorical cross entropy and others use binary cross entropy.</p>\n\n<p>Which is more suited for this competition?</p>",
  "messages": [
    {
      "id": "808968",
      "postDate": "04/15/2020 18:44:20",
      "content": "<p>I have seen some kernels use categorical cross entropy and others use binary cross entropy.</p>\n\n<p>Which is more suited for this competition?</p>",
      "rawMarkdown": "I have seen some kernels use categorical cross entropy and others use binary cross entropy.\n\nWhich is more suited for this competition?",
      "votes": null
    },
    {
      "id": "809017",
      "postDate": "04/15/2020 19:13:47",
      "content": "<p>also you should check:\n- <strong><a href=\"http://openaccess.thecvf.com/content_CVPR_2019/papers/Cui_Class-Balanced_Loss_Based_on_Effective_Number_of_Samples_CVPR_2019_paper.pdf\">Class-Balanced Loss Based on Effective Number of Samples</a></strong> (recommended)\n- <strong><a href=\"https://arxiv.org/abs/1708.02002\">Focal Loss</a></strong></p>",
      "rawMarkdown": "also you should check:\n- **[Class-Balanced Loss Based on Effective Number of Samples](http://openaccess.thecvf.com/content_CVPR_2019/papers/Cui_Class-Balanced_Loss_Based_on_Effective_Number_of_Samples_CVPR_2019_paper.pdf)** (recommended)\n- **[Focal Loss](https://arxiv.org/abs/1708.02002)**",
      "votes": null
    },
    {
      "id": "809343",
      "postDate": "04/16/2020 04:59:10",
      "content": "<p>I use BCEWithLogitsLoss</p>",
      "rawMarkdown": "I use BCEWithLogitsLoss",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 809017,
      "author_name": "jesucristo",
      "author_url": "",
      "post_date": "04/15/2020 19:13:47",
      "content": "<p>also you should check:\n- <strong><a href=\"http://openaccess.thecvf.com/content_CVPR_2019/papers/Cui_Class-Balanced_Loss_Based_on_Effective_Number_of_Samples_CVPR_2019_paper.pdf\">Class-Balanced Loss Based on Effective Number of Samples</a></strong> (recommended)\n- <strong><a href=\"https://arxiv.org/abs/1708.02002\">Focal Loss</a></strong></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 809343,
      "author_name": "alimbekovkz",
      "author_url": "",
      "post_date": "04/16/2020 04:59:10",
      "content": "<p>I use BCEWithLogitsLoss</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "808968": "I have seen some kernels use categorical cross entropy and others use binary cross entropy.\n\nWhich is more suited for this competition?",
    "809017": "also you should check:\n- **[Class-Balanced Loss Based on Effective Number of Samples](http://openaccess.thecvf.com/content_CVPR_2019/papers/Cui_Class-Balanced_Loss_Based_on_Effective_Number_of_Samples_CVPR_2019_paper.pdf)** (recommended)\n- **[Focal Loss](https://arxiv.org/abs/1708.02002)**",
    "809343": "I use BCEWithLogitsLoss"
  },
  "source": "meta"
}