{
  "id": 211173,
  "title": "Is it possible for us to implement distillation learning by ourselves? ",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/211173",
  "author_name": "Junyi Ng",
  "post_date": "2021-01-14T03:33:00.513000",
  "votes": 7,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Ever since I came across the concept of \"noisy student\", I've been wondering if we could implement something like that for this competition (on a smaller scale, of course).</p>\n<p>After reading about \"meta pseudo labels\" yesterday, which broke past 90% on image net, I can't help but ask, is it feasible?</p>\n<p>Link: <a href=\"https://paperswithcode.com/paper/meta-pseudo-labels\" target=\"_blank\">https://paperswithcode.com/paper/meta-pseudo-labels</a></p>",
  "messages": [
    {
      "id": 1152335,
      "postDate": "2021-01-14T03:33:00.513Z",
      "content": "<p>Ever since I came across the concept of \"noisy student\", I've been wondering if we could implement something like that for this competition (on a smaller scale, of course).</p>\n<p>After reading about \"meta pseudo labels\" yesterday, which broke past 90% on image net, I can't help but ask, is it feasible?</p>\n<p>Link: <a href=\"https://paperswithcode.com/paper/meta-pseudo-labels\" target=\"_blank\">https://paperswithcode.com/paper/meta-pseudo-labels</a></p>",
      "rawMarkdown": "Ever since I came across the concept of \"noisy student\", I've been wondering if we could implement something like that for this competition (on a smaller scale, of course).\n\nAfter reading about \"meta pseudo labels\" yesterday, which broke past 90% on image net, I can't help but ask, is it feasible?\n\nLink: https://paperswithcode.com/paper/meta-pseudo-labels",
      "votes": 7
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1152335": "Ever since I came across the concept of \"noisy student\", I've been wondering if we could implement something like that for this competition (on a smaller scale, of course).\n\nAfter reading about \"meta pseudo labels\" yesterday, which broke past 90% on image net, I can't help but ask, is it feasible?\n\nLink: https://paperswithcode.com/paper/meta-pseudo-labels"
  }
}