{
  "id": 421460,
  "title": "Why isn't any one doing semi supervised learning??",
  "url": "/competitions/hubmap-hacking-the-human-vasculature/discussion/421460",
  "author_name": "Vishak K Bhat",
  "post_date": "2023-07-05T12:50:01.170000",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Whatever notebooks i went through in the public code,<br>\nnone included semi supervised learning technique…</p>\n<p>I mean to say why havent any one used the dataset3 ??</p>\n<p>Is there any problem??</p>",
  "messages": [
    {
      "id": 2332099,
      "postDate": "2023-07-06T01:38:21.520Z",
      "content": "<p>I'm reading a paper to apply semi supervised learning and I think its performance increment would be not that big.. <br>\nA standard benchmark tests with 1%, 2%, 5%, 10% of labeled data and the effect of semi-supervised learning gets smaller as the labeled data increases. In this competition, we have almost 20% of labeled data and it's quite more than the standard benchmark.<br>\nI'm not tested yet but I think it may show similar effect with pseudo labelling.</p>",
      "rawMarkdown": "I'm reading a paper to apply semi supervised learning and I think its performance increment would be not that big.. \nA standard benchmark tests with 1%, 2%, 5%, 10% of labeled data and the effect of semi-supervised learning gets smaller as the labeled data increases. In this competition, we have almost 20% of labeled data and it's quite more than the standard benchmark.\nI'm not tested yet but I think it may show similar effect with pseudo labelling.",
      "votes": 1,
      "replies": [
        {
          "id": 2332192,
          "postDate": "2023-07-06T04:04:20.037Z",
          "content": "<p>may I know which paper are you reading ?</p>",
          "rawMarkdown": "may I know which paper are you reading ?",
          "replies": [
            {
              "id": 2332253,
              "postDate": "2023-07-06T05:01:09.357Z",
              "content": "<p><a href=\"https://github.com/AlibabaResearch/efficientteacher\" target=\"_blank\">https://github.com/AlibabaResearch/efficientteacher</a></p>\n<p>Paper is also included in that link</p>",
              "rawMarkdown": "https://github.com/AlibabaResearch/efficientteacher\n\nPaper is also included in that link"
            }
          ]
        }
      ]
    },
    {
      "id": 2331240,
      "postDate": "2023-07-05T12:50:01.170Z",
      "content": "<p>Whatever notebooks i went through in the public code,<br>\nnone included semi supervised learning technique…</p>\n<p>I mean to say why havent any one used the dataset3 ??</p>\n<p>Is there any problem??</p>",
      "rawMarkdown": "Whatever notebooks i went through in the public code,\nnone included semi supervised learning technique...\n\nI mean to say why havent any one used the dataset3 ??\n\nIs there any problem??\n",
      "votes": 2
    },
    {
      "id": 2331291,
      "postDate": "2023-07-05T13:23:36.597Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2332099,
      "author_name": "dodoit",
      "author_url": "",
      "post_date": "2023-07-06T01:38:21.520000",
      "content": "<p>I'm reading a paper to apply semi supervised learning and I think its performance increment would be not that big.. <br>\nA standard benchmark tests with 1%, 2%, 5%, 10% of labeled data and the effect of semi-supervised learning gets smaller as the labeled data increases. In this competition, we have almost 20% of labeled data and it's quite more than the standard benchmark.<br>\nI'm not tested yet but I think it may show similar effect with pseudo labelling.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 2332192,
          "author_name": "Vishak K Bhat",
          "author_url": "",
          "post_date": "2023-07-06T04:04:20.037000",
          "content": "<p>may I know which paper are you reading ?</p>",
          "votes": 0,
          "replies": [
            {
              "id": 2332253,
              "author_name": "dodoit",
              "author_url": "",
              "post_date": "2023-07-06T05:01:09.357000",
              "content": "<p><a href=\"https://github.com/AlibabaResearch/efficientteacher\" target=\"_blank\">https://github.com/AlibabaResearch/efficientteacher</a></p>\n<p>Paper is also included in that link</p>",
              "votes": 0,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 2331291,
      "author_name": "",
      "author_url": "",
      "post_date": "2023-07-05T13:23:36.597000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2332099": "I'm reading a paper to apply semi supervised learning and I think its performance increment would be not that big.. \nA standard benchmark tests with 1%, 2%, 5%, 10% of labeled data and the effect of semi-supervised learning gets smaller as the labeled data increases. In this competition, we have almost 20% of labeled data and it's quite more than the standard benchmark.\nI'm not tested yet but I think it may show similar effect with pseudo labelling.",
    "2331240": "Whatever notebooks i went through in the public code,\nnone included semi supervised learning technique...\n\nI mean to say why havent any one used the dataset3 ??\n\nIs there any problem??\n",
    "2331291": ""
  }
}