{
  "id": 419512,
  "title": "Is test data truly private?",
  "url": "/competitions/google-research-identify-contrails-reduce-global-warming/discussion/419512",
  "author_name": "Bilzard",
  "post_date": "2023-06-26T09:23:59.183000",
  "votes": 10,
  "comment_count": 2,
  "views": 0,
  "content": "<p>According to the paper by host[1], they published dataset[2].</p>\n<p>In my understanding, this competition's dataset is exactly same as this dataset, or, subset of this dataset.<br>\n<br>\n-&gt; Considering the original dataset's resolution is float64 and this competition's dataset is float32, the dataset size must be nearly the same.</p>\n<p>Then I wonder if test dataset is truly private, i.e. they are not taken from this dataset.</p>\n<p><a href=\"https://www.kaggle.com/inversion\" target=\"_blank\">@inversion</a> Could you confirm this?</p>\n<p>Besides, is it O.K. to use this dataset to train model?</p>\n<p>[1] <a href=\"https://arxiv.org/abs/2304.02122\" target=\"_blank\">https://arxiv.org/abs/2304.02122</a><br>\n[2] gs://goes_contrails_dataset</p>",
  "messages": [
    {
      "id": 2318287,
      "postDate": "2023-06-26T09:23:59.183Z",
      "content": "<p>According to the paper by host[1], they published dataset[2].</p>\n<p>In my understanding, this competition's dataset is exactly same as this dataset, or, subset of this dataset.<br>\n<br>\n-&gt; Considering the original dataset's resolution is float64 and this competition's dataset is float32, the dataset size must be nearly the same.</p>\n<p>Then I wonder if test dataset is truly private, i.e. they are not taken from this dataset.</p>\n<p><a href=\"https://www.kaggle.com/inversion\" target=\"_blank\">@inversion</a> Could you confirm this?</p>\n<p>Besides, is it O.K. to use this dataset to train model?</p>\n<p>[1] <a href=\"https://arxiv.org/abs/2304.02122\" target=\"_blank\">https://arxiv.org/abs/2304.02122</a><br>\n[2] gs://goes_contrails_dataset</p>",
      "rawMarkdown": "According to the paper by host[1], they published dataset[2].\n\nIn my understanding, this competition's dataset is exactly same as this dataset, or, subset of this dataset.\n~~(Considering the bigger storage size (~900GB), I guess the latter is true.)~~\n-> Considering the original dataset's resolution is float64 and this competition's dataset is float32, the dataset size must be nearly the same.\n\nThen I wonder if test dataset is truly private, i.e. they are not taken from this dataset.\n\n@inversion Could you confirm this?\n\nBesides, is it O.K. to use this dataset to train model?\n\n[1] https://arxiv.org/abs/2304.02122\n[2] gs://goes_contrails_dataset",
      "votes": 9
    },
    {
      "id": 2318691,
      "postDate": "2023-06-26T14:13:38.077Z",
      "content": "<p>Thanks for asking. The test set for this competition was not released as part of the referenced dataset. Please see Rule 7.C. regarding the use of other datasets.</p>",
      "rawMarkdown": "Thanks for asking. The test set for this competition was not released as part of the referenced dataset. Please see Rule 7.C. regarding the use of other datasets.",
      "replies": [
        {
          "id": 2318834,
          "postDate": "2023-06-26T15:56:04.733Z",
          "content": "<p><a href=\"https://www.kaggle.com/inversion\" target=\"_blank\">@inversion</a> <br>\nThanks. Could you also confirm if they are exactly same except for precisions  (float64 &lt;-&gt; float32)?</p>",
          "rawMarkdown": "@inversion \nThanks. Could you also confirm if they are exactly same except for precisions  (float64 <-> float32)?",
          "votes": 1
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 2318691,
      "author_name": "inversion",
      "author_url": "",
      "post_date": "2023-06-26T14:13:38.077000",
      "content": "<p>Thanks for asking. The test set for this competition was not released as part of the referenced dataset. Please see Rule 7.C. regarding the use of other datasets.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 2318834,
          "author_name": "Bilzard",
          "author_url": "",
          "post_date": "2023-06-26T15:56:04.733000",
          "content": "<p><a href=\"https://www.kaggle.com/inversion\" target=\"_blank\">@inversion</a> <br>\nThanks. Could you also confirm if they are exactly same except for precisions  (float64 &lt;-&gt; float32)?</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2318287": "According to the paper by host[1], they published dataset[2].\n\nIn my understanding, this competition's dataset is exactly same as this dataset, or, subset of this dataset.\n~~(Considering the bigger storage size (~900GB), I guess the latter is true.)~~\n-> Considering the original dataset's resolution is float64 and this competition's dataset is float32, the dataset size must be nearly the same.\n\nThen I wonder if test dataset is truly private, i.e. they are not taken from this dataset.\n\n@inversion Could you confirm this?\n\nBesides, is it O.K. to use this dataset to train model?\n\n[1] https://arxiv.org/abs/2304.02122\n[2] gs://goes_contrails_dataset",
    "2318691": "Thanks for asking. The test set for this competition was not released as part of the referenced dataset. Please see Rule 7.C. regarding the use of other datasets."
  }
}