{
  "id": 76154,
  "title": "Correct way to implement pseudo labelling",
  "url": "/competitions/quora-insincere-questions-classification/discussion/76154",
  "author_name": "bilal2vec",
  "post_date": "2018-12-29T20:56:40.233000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi.</p>\n\n<p>I tried to implement pseudo labeling and got a kernel with a public LB score of 0.688 but I'm not sure how to make sure that I don't overfit to the public test set.</p>\n\n<p>Using the validation set as part of the training set means that relying on the validation loss to determine when to stop training isn't possible</p>\n\n<p>My pseudo labeling kernel is: <a href=\"https://www.kaggle.com/bkkaggle/skip-rnn-meta-features-pseudo-labeling\">https://www.kaggle.com/bkkaggle/skip-rnn-meta-features-pseudo-labeling</a></p>",
  "messages": [
    {
      "id": 447454,
      "postDate": "2018-12-29T20:56:40.233Z",
      "content": "<p>Hi.</p>\n\n<p>I tried to implement pseudo labeling and got a kernel with a public LB score of 0.688 but I'm not sure how to make sure that I don't overfit to the public test set.</p>\n\n<p>Using the validation set as part of the training set means that relying on the validation loss to determine when to stop training isn't possible</p>\n\n<p>My pseudo labeling kernel is: <a href=\"https://www.kaggle.com/bkkaggle/skip-rnn-meta-features-pseudo-labeling\">https://www.kaggle.com/bkkaggle/skip-rnn-meta-features-pseudo-labeling</a></p>",
      "rawMarkdown": "Hi.\n\nI tried to implement pseudo labeling and got a kernel with a public LB score of 0.688 but I'm not sure how to make sure that I don't overfit to the public test set.\n\nUsing the validation set as part of the training set means that relying on the validation loss to determine when to stop training isn't possible\n\nMy pseudo labeling kernel is: https://www.kaggle.com/bkkaggle/skip-rnn-meta-features-pseudo-labeling",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "447454": "Hi.\n\nI tried to implement pseudo labeling and got a kernel with a public LB score of 0.688 but I'm not sure how to make sure that I don't overfit to the public test set.\n\nUsing the validation set as part of the training set means that relying on the validation loss to determine when to stop training isn't possible\n\nMy pseudo labeling kernel is: https://www.kaggle.com/bkkaggle/skip-rnn-meta-features-pseudo-labeling"
  }
}