{
  "id": 417696,
  "title": "Training data augmentation",
  "url": "/competitions/cafa-5-protein-function-prediction/discussion/417696",
  "author_name": "",
  "post_date": "2023-06-16T19:23:15.490452100Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Are we allowed to train on a larger set of proteins other than the ones given in the training data? And if so how have others pursued such a strategy? </p>",
  "messages": [
    {
      "id": "2305578",
      "postDate": "06/16/2023 19:23:15",
      "content": "<p>Are we allowed to train on a larger set of proteins other than the ones given in the training data? And if so how have others pursued such a strategy? </p>",
      "rawMarkdown": "Are we allowed to train on a larger set of proteins other than the ones given in the training data? And if so how have others pursued such a strategy?",
      "votes": null
    },
    {
      "id": "2305603",
      "postDate": "06/16/2023 19:43:22",
      "content": "<blockquote>\n  <p>Are we allowed to train on a larger set of proteins other than the ones given in the training data?</p>\n</blockquote>\n<p>Yes, even encouraged.</p>\n<blockquote>\n  <p>And if so how have others pursued such a strategy? </p>\n</blockquote>\n<p>In previous CAFAs competitors built their own training sets that included other attributes such as protein structure, genomic localization, and co-experssion, and used them to build models. </p>",
      "rawMarkdown": "> Are we allowed to train on a larger set of proteins other than the ones given in the training data?\n\nYes, even encouraged.\n\n> And if so how have others pursued such a strategy? \n\nIn previous CAFAs competitors built their own training sets that included other attributes such as protein structure, genomic localization, and co-experssion, and used them to build models.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2305603,
      "author_name": "iddux1",
      "author_url": "",
      "post_date": "06/16/2023 19:43:22",
      "content": "<blockquote>\n  <p>Are we allowed to train on a larger set of proteins other than the ones given in the training data?</p>\n</blockquote>\n<p>Yes, even encouraged.</p>\n<blockquote>\n  <p>And if so how have others pursued such a strategy? </p>\n</blockquote>\n<p>In previous CAFAs competitors built their own training sets that included other attributes such as protein structure, genomic localization, and co-experssion, and used them to build models. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2305578": "Are we allowed to train on a larger set of proteins other than the ones given in the training data? And if so how have others pursued such a strategy?",
    "2305603": "> Are we allowed to train on a larger set of proteins other than the ones given in the training data?\n\nYes, even encouraged.\n\n> And if so how have others pursued such a strategy? \n\nIn previous CAFAs competitors built their own training sets that included other attributes such as protein structure, genomic localization, and co-experssion, and used them to build models."
  },
  "source": "meta"
}