{
  "id": 567388,
  "title": "Dumb Question: Do we split the train data?",
  "url": "/competitions/stanford-rna-3d-folding/discussion/567388",
  "author_name": "",
  "post_date": "2025-03-10T02:58:56.595536500Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi.<br>\nPerhaps this is a silly question, but as there is validation data, does this imply splitting the data in the conventional way via a test/train split is pointless as we can use the validation data to test on??<br>\nCheers<br>\nM</p>",
  "messages": [
    {
      "id": "3145598",
      "postDate": "03/10/2025 02:58:56",
      "content": "<p>Hi.<br>\nPerhaps this is a silly question, but as there is validation data, does this imply splitting the data in the conventional way via a test/train split is pointless as we can use the validation data to test on??<br>\nCheers<br>\nM</p>",
      "rawMarkdown": "Hi.\nPerhaps this is a silly question, but as there is validation data, does this imply splitting the data in the conventional way via a test/train split is pointless as we can use the validation data to test on??\nCheers\nM",
      "votes": null
    },
    {
      "id": "3146007",
      "postDate": "03/10/2025 13:13:21",
      "content": "<p>Think it like this:</p>\n<p>There is validation data already. =&gt; </p>\n<p>Effectively someone has done a train/val split already. =&gt; </p>\n<p>Does it matter who that someone is? No, for all we care, it could be Donald Trump. =&gt;</p>\n<p>So we can assume without any loss of generality that that someone was you yourself who did it yesterday. =&gt;</p>\n<p>Is it pointless that you wake up this morning and say to yourself: Gee, I really want to do my train/val split another way, different from how I did it yesterday.</p>",
      "rawMarkdown": "Think it like this:\n\nThere is validation data already. => \n\nEffectively someone has done a train/val split already. => \n\nDoes it matter who that someone is? No, for all we care, it could be Donald Trump. =>\n\nSo we can assume without any loss of generality that that someone was you yourself who did it yesterday. =>\n\nIs it pointless that you wake up this morning and say to yourself: Gee, I really want to do my train/val split another way, different from how I did it yesterday.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3146007,
      "author_name": "revealer",
      "author_url": "",
      "post_date": "03/10/2025 13:13:21",
      "content": "<p>Think it like this:</p>\n<p>There is validation data already. =&gt; </p>\n<p>Effectively someone has done a train/val split already. =&gt; </p>\n<p>Does it matter who that someone is? No, for all we care, it could be Donald Trump. =&gt;</p>\n<p>So we can assume without any loss of generality that that someone was you yourself who did it yesterday. =&gt;</p>\n<p>Is it pointless that you wake up this morning and say to yourself: Gee, I really want to do my train/val split another way, different from how I did it yesterday.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3145598": "Hi.\nPerhaps this is a silly question, but as there is validation data, does this imply splitting the data in the conventional way via a test/train split is pointless as we can use the validation data to test on??\nCheers\nM",
    "3146007": "Think it like this:\n\nThere is validation data already. => \n\nEffectively someone has done a train/val split already. => \n\nDoes it matter who that someone is? No, for all we care, it could be Donald Trump. =>\n\nSo we can assume without any loss of generality that that someone was you yourself who did it yesterday. =>\n\nIs it pointless that you wake up this morning and say to yourself: Gee, I really want to do my train/val split another way, different from how I did it yesterday."
  },
  "source": "meta"
}