{
  "id": 321311,
  "title": "Semi-supervised Learning Techniques",
  "url": "/competitions/sorghum-id-fgvc-9/discussion/321311",
  "author_name": "",
  "post_date": "2022-04-26T07:40:36.081508400Z",
  "votes": 4,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Would it be against the competition rules if I used semi-supervised learning techniques like Noisy Student on the test set? I'm not sure if it is considered as \"collecting annotations on the test set\" since the pseudo-labels are generated by a teacher model trained on the training set and the test set is not hand labelled. The overall method can also just be viewed as a computationally intensive and slow way of getting more accurate labels. But training on the test set does also sound quite suspicious so I don't really know.</p>",
  "messages": [
    {
      "id": "1768353",
      "postDate": "04/26/2022 07:40:36",
      "content": "<p>Would it be against the competition rules if I used semi-supervised learning techniques like Noisy Student on the test set? I'm not sure if it is considered as \"collecting annotations on the test set\" since the pseudo-labels are generated by a teacher model trained on the training set and the test set is not hand labelled. The overall method can also just be viewed as a computationally intensive and slow way of getting more accurate labels. But training on the test set does also sound quite suspicious so I don't really know.</p>",
      "rawMarkdown": "Would it be against the competition rules if I used semi-supervised learning techniques like Noisy Student on the test set? I'm not sure if it is considered as \"collecting annotations on the test set\" since the pseudo-labels are generated by a teacher model trained on the training set and the test set is not hand labelled. The overall method can also just be viewed as a computationally intensive and slow way of getting more accurate labels. But training on the test set does also sound quite suspicious so I don't really know.",
      "votes": null
    },
    {
      "id": "1772828",
      "postDate": "04/30/2022 15:50:28",
      "content": "<p>In my interpretation it would be allowed, since pseudo-labels are not real annotations. I'd be interested in an official statement on this.</p>",
      "rawMarkdown": "In my interpretation it would be allowed, since pseudo-labels are not real annotations. I'd be interested in an official statement on this.",
      "votes": null
    },
    {
      "id": "1775026",
      "postDate": "05/02/2022 16:50:55",
      "content": "<p>Semi-supervised learning is fine.</p>",
      "rawMarkdown": "Semi-supervised learning is fine.",
      "votes": null
    },
    {
      "id": "3523111",
      "postDate": "09/10/2026 15:55:33",
      "content": "<p>Cultivar identification needs labels—<strong>but how much could ML learn from unlabeled images?</strong> 🔍</p>\n<p><a href=\"https://mlguidance.blogspot.com/2026/09/what-are-models-of-semi-supervised.html\" target=\"_blank\">https://mlguidance.blogspot.com/2026/09/what-are-models-of-semi-supervised.html</a></p>",
      "rawMarkdown": "Cultivar identification needs labels—**but how much could ML learn from unlabeled images?** 🔍\n\nhttps://mlguidance.blogspot.com/2026/09/what-are-models-of-semi-supervised.html",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1772828,
      "author_name": "tfriedel",
      "author_url": "",
      "post_date": "04/30/2022 15:50:28",
      "content": "<p>In my interpretation it would be allowed, since pseudo-labels are not real annotations. I'd be interested in an official statement on this.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1775026,
      "author_name": "sohier",
      "author_url": "",
      "post_date": "05/02/2022 16:50:55",
      "content": "<p>Semi-supervised learning is fine.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3523111,
      "author_name": "kaustubh994",
      "author_url": "",
      "post_date": "09/10/2026 15:55:33",
      "content": "<p>Cultivar identification needs labels—<strong>but how much could ML learn from unlabeled images?</strong> 🔍</p>\n<p><a href=\"https://mlguidance.blogspot.com/2026/09/what-are-models-of-semi-supervised.html\" target=\"_blank\">https://mlguidance.blogspot.com/2026/09/what-are-models-of-semi-supervised.html</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1768353": "Would it be against the competition rules if I used semi-supervised learning techniques like Noisy Student on the test set? I'm not sure if it is considered as \"collecting annotations on the test set\" since the pseudo-labels are generated by a teacher model trained on the training set and the test set is not hand labelled. The overall method can also just be viewed as a computationally intensive and slow way of getting more accurate labels. But training on the test set does also sound quite suspicious so I don't really know.",
    "1772828": "In my interpretation it would be allowed, since pseudo-labels are not real annotations. I'd be interested in an official statement on this.",
    "1775026": "Semi-supervised learning is fine.",
    "3523111": "Cultivar identification needs labels—**but how much could ML learn from unlabeled images?** 🔍\n\nhttps://mlguidance.blogspot.com/2026/09/what-are-models-of-semi-supervised.html"
  },
  "source": "meta"
}