{
  "id": 182191,
  "title": "Label noise in CV",
  "url": "/competitions/birdsong-recognition/discussion/182191",
  "author_name": "Matthias",
  "post_date": "2020-09-11T15:48:34.265000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>When using a split of the training data or comparable external sources for your CV obviously there will be a lot of label noise in the validation splits.</p>\n<p>How do you handle this - so far I do not feel confident to just exclude samples based on a \"high confusion threshold\" …</p>\n<p>I have the feeling that this might influence any kind of local CV attempts a lot….</p>",
  "messages": [
    {
      "id": 1006871,
      "postDate": "2020-09-11T15:48:34.267Z",
      "content": "<p>When using a split of the training data or comparable external sources for your CV obviously there will be a lot of label noise in the validation splits.</p>\n<p>How do you handle this - so far I do not feel confident to just exclude samples based on a \"high confusion threshold\" …</p>\n<p>I have the feeling that this might influence any kind of local CV attempts a lot….</p>",
      "rawMarkdown": "When using a split of the training data or comparable external sources for your CV obviously there will be a lot of label noise in the validation splits.\n\nHow do you handle this - so far I do not feel confident to just exclude samples based on a \"high confusion threshold\" ...\n\nI have the feeling that this might influence any kind of local CV attempts a lot...."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1006871": "When using a split of the training data or comparable external sources for your CV obviously there will be a lot of label noise in the validation splits.\n\nHow do you handle this - so far I do not feel confident to just exclude samples based on a \"high confusion threshold\" ...\n\nI have the feeling that this might influence any kind of local CV attempts a lot...."
  }
}