{
  "id": 295253,
  "title": "Expectation for using the supplemental train data after the submission period?",
  "url": "/competitions/g-research-crypto-forecasting/discussion/295253",
  "author_name": "",
  "post_date": "2021-12-15T05:49:59.532807Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>The description for the supplemental train data is given as:</p>\n<blockquote>\n  <p>After the submission period is over this file's data will be replaced with cryptoasset prices from the submission period. In the Evaluation phase, the train, train supplement, and test set will be contiguous in time, apart from any missing data. The current copy, which is just filled approximately the right amount of data from train.csv is provided as a placeholder.</p>\n</blockquote>\n<p>This is an intuitive way to use this information for the competition, as more recent observations are generally more important than older ones. Therefore these observations should not be discarded, but obviously they aren't available during the submission period for use. </p>\n<p>However, when planning the submission notebook for the Evaluation phase, this also made me think about how this data would be incorporated into a solution. </p>\n<p>For example, to use it for model training as the name implies, wouldn't the submission notebook now have to both train model(s) with the supplemental data AND make the test evaluations after training? I assume that would be an important part of how we weigh notebook runtime?</p>\n<p>If I am misunderstanding how this relatively important data is intended to be used, I would appreciate any insight or feedback. Thanks in advance!</p>",
  "messages": [
    {
      "id": "1618583",
      "postDate": "12/15/2021 05:49:59",
      "content": "<p>Hi,</p>\n<p>The description for the supplemental train data is given as:</p>\n<blockquote>\n  <p>After the submission period is over this file's data will be replaced with cryptoasset prices from the submission period. In the Evaluation phase, the train, train supplement, and test set will be contiguous in time, apart from any missing data. The current copy, which is just filled approximately the right amount of data from train.csv is provided as a placeholder.</p>\n</blockquote>\n<p>This is an intuitive way to use this information for the competition, as more recent observations are generally more important than older ones. Therefore these observations should not be discarded, but obviously they aren't available during the submission period for use. </p>\n<p>However, when planning the submission notebook for the Evaluation phase, this also made me think about how this data would be incorporated into a solution. </p>\n<p>For example, to use it for model training as the name implies, wouldn't the submission notebook now have to both train model(s) with the supplemental data AND make the test evaluations after training? I assume that would be an important part of how we weigh notebook runtime?</p>\n<p>If I am misunderstanding how this relatively important data is intended to be used, I would appreciate any insight or feedback. Thanks in advance!</p>",
      "rawMarkdown": "Hi,\n\nThe description for the supplemental train data is given as:\n> After the submission period is over this file's data will be replaced with cryptoasset prices from the submission period. In the Evaluation phase, the train, train supplement, and test set will be contiguous in time, apart from any missing data. The current copy, which is just filled approximately the right amount of data from train.csv is provided as a placeholder.\n\nThis is an intuitive way to use this information for the competition, as more recent observations are generally more important than older ones. Therefore these observations should not be discarded, but obviously they aren't available during the submission period for use. \n\nHowever, when planning the submission notebook for the Evaluation phase, this also made me think about how this data would be incorporated into a solution. \n\nFor example, to use it for model training as the name implies, wouldn't the submission notebook now have to both train model(s) with the supplemental data AND make the test evaluations after training? I assume that would be an important part of how we weigh notebook runtime?\n\nIf I am misunderstanding how this relatively important data is intended to be used, I would appreciate any insight or feedback. Thanks in advance!",
      "votes": null
    },
    {
      "id": "1618589",
      "postDate": "12/15/2021 05:55:14",
      "content": "<p>To clarify in advance I also see how the supplemental data could be used for other purposes, such as inputs to a time series model in order to generate a prediction for the first test point. But for actual training, as it seems was the intention behind the supplemental dataset, it seems like there are serious considerations if using it?</p>",
      "rawMarkdown": "To clarify in advance I also see how the supplemental data could be used for other purposes, such as inputs to a time series model in order to generate a prediction for the first test point. But for actual training, as it seems was the intention behind the supplemental dataset, it seems like there are serious considerations if using it?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1618589,
      "author_name": "ajaxdata",
      "author_url": "",
      "post_date": "12/15/2021 05:55:14",
      "content": "<p>To clarify in advance I also see how the supplemental data could be used for other purposes, such as inputs to a time series model in order to generate a prediction for the first test point. But for actual training, as it seems was the intention behind the supplemental dataset, it seems like there are serious considerations if using it?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1618583": "Hi,\n\nThe description for the supplemental train data is given as:\n> After the submission period is over this file's data will be replaced with cryptoasset prices from the submission period. In the Evaluation phase, the train, train supplement, and test set will be contiguous in time, apart from any missing data. The current copy, which is just filled approximately the right amount of data from train.csv is provided as a placeholder.\n\nThis is an intuitive way to use this information for the competition, as more recent observations are generally more important than older ones. Therefore these observations should not be discarded, but obviously they aren't available during the submission period for use. \n\nHowever, when planning the submission notebook for the Evaluation phase, this also made me think about how this data would be incorporated into a solution. \n\nFor example, to use it for model training as the name implies, wouldn't the submission notebook now have to both train model(s) with the supplemental data AND make the test evaluations after training? I assume that would be an important part of how we weigh notebook runtime?\n\nIf I am misunderstanding how this relatively important data is intended to be used, I would appreciate any insight or feedback. Thanks in advance!",
    "1618589": "To clarify in advance I also see how the supplemental data could be used for other purposes, such as inputs to a time series model in order to generate a prediction for the first test point. But for actual training, as it seems was the intention behind the supplemental dataset, it seems like there are serious considerations if using it?"
  },
  "source": "meta"
}