{
  "id": 542628,
  "title": "Training dataset at evaluation stage",
  "url": "/competitions/child-mind-institute-problematic-internet-use/discussion/542628",
  "author_name": "",
  "post_date": "2024-10-25T22:03:39.370095700Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I understand that during the evaluation stage, both the test actigraph and test metadata are replaced - Are you however replacing also the training dataset during evaluation? I am asking because I do not know if I also need to train my models in notebooks or can I just upload the trained models for evaluation.</p>",
  "messages": [
    {
      "id": "3028333",
      "postDate": "10/25/2024 22:03:39",
      "content": "<p>I understand that during the evaluation stage, both the test actigraph and test metadata are replaced - Are you however replacing also the training dataset during evaluation? I am asking because I do not know if I also need to train my models in notebooks or can I just upload the trained models for evaluation.</p>",
      "rawMarkdown": "I understand that during the evaluation stage, both the test actigraph and test metadata are replaced - Are you however replacing also the training dataset during evaluation? I am asking because I do not know if I also need to train my models in notebooks or can I just upload the trained models for evaluation.",
      "votes": null
    },
    {
      "id": "3029383",
      "postDate": "10/27/2024 08:28:59",
      "content": "<p>The training set is just one and you can use a pre-trained model that predicts the test set. To make it simpler - if you submission is succesful and you have a LB score than this notebook is ok because behind the scene you are predicting the full test set but just the partial one is showed as public.</p>",
      "rawMarkdown": "The training set is just one and you can use a pre-trained model that predicts the test set. To make it simpler - if you submission is succesful and you have a LB score than this notebook is ok because behind the scene you are predicting the full test set but just the partial one is showed as public.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3029383,
      "author_name": "eu1234",
      "author_url": "",
      "post_date": "10/27/2024 08:28:59",
      "content": "<p>The training set is just one and you can use a pre-trained model that predicts the test set. To make it simpler - if you submission is succesful and you have a LB score than this notebook is ok because behind the scene you are predicting the full test set but just the partial one is showed as public.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3028333": "I understand that during the evaluation stage, both the test actigraph and test metadata are replaced - Are you however replacing also the training dataset during evaluation? I am asking because I do not know if I also need to train my models in notebooks or can I just upload the trained models for evaluation.",
    "3029383": "The training set is just one and you can use a pre-trained model that predicts the test set. To make it simpler - if you submission is succesful and you have a LB score than this notebook is ok because behind the scene you are predicting the full test set but just the partial one is showed as public."
  },
  "source": "meta"
}