{
  "id": 53893,
  "title": "Stage2 questions: Re-train with stage1 test release after 04/11",
  "url": "/competitions/data-science-bowl-2018/discussion/53893",
  "author_name": "",
  "post_date": "2018-04-06T14:17:13.821589600Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>According to the rules we need to upload our model by 04/11. \nHowever, we're allowed to re-train after the 04/11 by including stage1 test that will be released (masks) at the same date.</p>\n\n<p>2 questions:</p>\n\n<ul>\n<li>Do we need to document how to include this dataset for re-train prior\nto upload? </li>\n<li>Are we allowed to update parameters passed to our train script after 04/11 (such as maximum epochs)? or is it considered as hyperparameter and then forbidden?</li>\n</ul>\n\n<p>Thanks.</p>",
  "messages": [
    {
      "id": "310061",
      "postDate": "04/06/2018 14:17:13",
      "content": "<p>According to the rules we need to upload our model by 04/11. \nHowever, we're allowed to re-train after the 04/11 by including stage1 test that will be released (masks) at the same date.</p>\n\n<p>2 questions:</p>\n\n<ul>\n<li>Do we need to document how to include this dataset for re-train prior\nto upload? </li>\n<li>Are we allowed to update parameters passed to our train script after 04/11 (such as maximum epochs)? or is it considered as hyperparameter and then forbidden?</li>\n</ul>\n\n<p>Thanks.</p>",
      "rawMarkdown": "According to the rules we need to upload our model by 04/11. \nHowever, we're allowed to re-train after the 04/11 by including stage1 test that will be released (masks) at the same date.\n\n2 questions:\n\n - Do we need to document how to include this dataset for re-train prior\n   to upload? \n - Are we allowed to update parameters passed to our train script after 04/11 (such as maximum epochs)? or is it considered as hyperparameter and then forbidden?\n\nThanks.",
      "votes": null
    },
    {
      "id": "310156",
      "postDate": "04/06/2018 17:35:39",
      "content": "<p>According to the <a href=\"https://www.kaggle.com/two-stage-faq\">2-stage faq</a>, you may change your code as necessary to deal with the new dataset's structure, and you may do automated tuning of parameters (which sounds like another term for retraining to me), but you may not tune hyperparameters, manually at least. I'm not sure if automated hyperparameter tuning is okay. Maximum epochs sounds like a hyperparameter to me. But I'm not an expert.</p>",
      "rawMarkdown": "According to the [2-stage faq](https://www.kaggle.com/two-stage-faq), you may change your code as necessary to deal with the new dataset's structure, and you may do automated tuning of parameters (which sounds like another term for retraining to me), but you may not tune hyperparameters, manually at least. I'm not sure if automated hyperparameter tuning is okay. Maximum epochs sounds like a hyperparameter to me. But I'm not an expert.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 310156,
      "author_name": "pointyointment",
      "author_url": "",
      "post_date": "04/06/2018 17:35:39",
      "content": "<p>According to the <a href=\"https://www.kaggle.com/two-stage-faq\">2-stage faq</a>, you may change your code as necessary to deal with the new dataset's structure, and you may do automated tuning of parameters (which sounds like another term for retraining to me), but you may not tune hyperparameters, manually at least. I'm not sure if automated hyperparameter tuning is okay. Maximum epochs sounds like a hyperparameter to me. But I'm not an expert.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "310061": "According to the rules we need to upload our model by 04/11. \nHowever, we're allowed to re-train after the 04/11 by including stage1 test that will be released (masks) at the same date.\n\n2 questions:\n\n - Do we need to document how to include this dataset for re-train prior\n   to upload? \n - Are we allowed to update parameters passed to our train script after 04/11 (such as maximum epochs)? or is it considered as hyperparameter and then forbidden?\n\nThanks.",
    "310156": "According to the [2-stage faq](https://www.kaggle.com/two-stage-faq), you may change your code as necessary to deal with the new dataset's structure, and you may do automated tuning of parameters (which sounds like another term for retraining to me), but you may not tune hyperparameters, manually at least. I'm not sure if automated hyperparameter tuning is okay. Maximum epochs sounds like a hyperparameter to me. But I'm not an expert."
  },
  "source": "meta"
}