{
  "id": 319988,
  "title": "Pseudolabelling notebook",
  "url": "/competitions/digit-recognizer/discussion/319988",
  "author_name": "Pavel Motloch",
  "post_date": "2022-04-19T16:47:32.270000",
  "votes": 2,
  "comment_count": 2,
  "views": null,
  "content": "<p>See <a href=\"https://www.kaggle.com/code/motloch/demonstrating-pseudolabelling-on-mnist\" target=\"_blank\">this notebook</a> for an example of pseudolabelling, a semi-supervised technique useful when we have access to only a small number of labelled data and a significantly larger set of unlabelled data. </p>\n<p>Using a simple CNN model, the performance improves from 90% to over 91%.</p>",
  "messages": [
    {
      "id": 1761044,
      "postDate": "2022-04-19T16:47:32.270Z",
      "content": "<p>See <a href=\"https://www.kaggle.com/code/motloch/demonstrating-pseudolabelling-on-mnist\" target=\"_blank\">this notebook</a> for an example of pseudolabelling, a semi-supervised technique useful when we have access to only a small number of labelled data and a significantly larger set of unlabelled data. </p>\n<p>Using a simple CNN model, the performance improves from 90% to over 91%.</p>",
      "rawMarkdown": "See [this notebook](https://www.kaggle.com/code/motloch/demonstrating-pseudolabelling-on-mnist) for an example of pseudolabelling, a semi-supervised technique useful when we have access to only a small number of labelled data and a significantly larger set of unlabelled data. \n\nUsing a simple CNN model, the performance improves from 90% to over 91%.",
      "votes": 2
    },
    {
      "id": 1762646,
      "postDate": "2022-04-20T21:36:59.857Z",
      "content": "<p>ty for it!</p>",
      "rawMarkdown": "ty for it!"
    },
    {
      "id": 1761914,
      "postDate": "2022-04-20T09:49:53.217Z",
      "content": "<p>its helpful information one can apply thanks for sharing .</p>",
      "rawMarkdown": "its helpful information one can apply thanks for sharing ."
    }
  ],
  "comments": [
    {
      "id": 1762646,
      "author_name": "RebelRaider",
      "author_url": "",
      "post_date": "2022-04-20T21:36:59.857000",
      "content": "<p>ty for it!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1761914,
      "author_name": "Iqbal Ali",
      "author_url": "",
      "post_date": "2022-04-20T09:49:53.217000",
      "content": "<p>its helpful information one can apply thanks for sharing .</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1761044": "See [this notebook](https://www.kaggle.com/code/motloch/demonstrating-pseudolabelling-on-mnist) for an example of pseudolabelling, a semi-supervised technique useful when we have access to only a small number of labelled data and a significantly larger set of unlabelled data. \n\nUsing a simple CNN model, the performance improves from 90% to over 91%.",
    "1762646": "ty for it!",
    "1761914": "its helpful information one can apply thanks for sharing ."
  }
}