{
  "id": 74583,
  "title": "Loss functions for optimizing top-K accuracy",
  "url": "/competitions/humpback-whale-identification/discussion/74583",
  "author_name": "",
  "post_date": "2018-12-13T19:08:30.868205800Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi,\nAfter going through the problem statement, my first try was to optimize over categorical cross entropy. But since this is a top-K identification problem, I think a loss function designed for such a goal will be better. \nCan anybody provide some good references?</p>",
  "messages": [
    {
      "id": "438478",
      "postDate": "12/13/2018 19:08:30",
      "content": "<p>Hi,\nAfter going through the problem statement, my first try was to optimize over categorical cross entropy. But since this is a top-K identification problem, I think a loss function designed for such a goal will be better. \nCan anybody provide some good references?</p>",
      "rawMarkdown": "Hi,\nAfter going through the problem statement, my first try was to optimize over categorical cross entropy. But since this is a top-K identification problem, I think a loss function designed for such a goal will be better. \nCan anybody provide some good references?",
      "votes": null
    },
    {
      "id": "438673",
      "postDate": "12/14/2018 02:17:55",
      "content": "<p>Try implementing it. ; )</p>",
      "rawMarkdown": "Try implementing it. ; )",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 438673,
      "author_name": "gxkok21",
      "author_url": "",
      "post_date": "12/14/2018 02:17:55",
      "content": "<p>Try implementing it. ; )</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "438478": "Hi,\nAfter going through the problem statement, my first try was to optimize over categorical cross entropy. But since this is a top-K identification problem, I think a loss function designed for such a goal will be better. \nCan anybody provide some good references?",
    "438673": "Try implementing it. ; )"
  },
  "source": "meta"
}