{
  "id": 565746,
  "title": "`validation_labels.csv`: should -1e+18 be interpreted as NaN?",
  "url": "/competitions/stanford-rna-3d-folding/discussion/565746",
  "author_name": "",
  "post_date": "2025-03-02T00:41:14.040198300Z",
  "votes": 9,
  "comment_count": 3,
  "views": 0,
  "content": "<p>In <code>train_labels.csv</code>, NaN coordinates appear as blank cells. In contrast, in <code>validation_labels.csv</code>, there seem to be a large number of -1e+18 entries. </p>\n<p>Should these be interpreted as NaN/unknown, in the same way that blanks are treated as NaN for train_labels? </p>\n<p>If so, why not encode them as NaN/blank?</p>",
  "messages": [
    {
      "id": "3137801",
      "postDate": "03/02/2025 00:41:14",
      "content": "<p>In <code>train_labels.csv</code>, NaN coordinates appear as blank cells. In contrast, in <code>validation_labels.csv</code>, there seem to be a large number of -1e+18 entries. </p>\n<p>Should these be interpreted as NaN/unknown, in the same way that blanks are treated as NaN for train_labels? </p>\n<p>If so, why not encode them as NaN/blank?</p>",
      "rawMarkdown": "In `train_labels.csv`, NaN coordinates appear as blank cells. In contrast, in `validation_labels.csv`, there seem to be a large number of -1e+18 entries. \n\nShould these be interpreted as NaN/unknown, in the same way that blanks are treated as NaN for train_labels? \n\nIf so, why not encode them as NaN/blank?",
      "votes": null
    },
    {
      "id": "3137886",
      "postDate": "03/02/2025 04:03:10",
      "content": "<p>It looks like some atoms are somewhere in space near Earth. But still somehow connected to RNA? Wow, we didn't learn that in school!</p>",
      "rawMarkdown": "It looks like some atoms are somewhere in space near Earth. But still somehow connected to RNA? Wow, we didn't learn that in school!",
      "votes": null
    },
    {
      "id": "3137895",
      "postDate": "03/02/2025 04:14:32",
      "content": "<p>Yes -1e+18 should be interpreted as NaNs </p>",
      "rawMarkdown": "Yes -1e+18 should be interpreted as NaNs",
      "votes": null
    },
    {
      "id": "3138192",
      "postDate": "03/02/2025 10:32:03",
      "content": "<p>Thank you for confirming, <a href=\"https://www.kaggle.com/shujun717\" target=\"_blank\">@shujun717</a>.</p>",
      "rawMarkdown": "Thank you for confirming, @shujun717.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3137886,
      "author_name": "sapr3s",
      "author_url": "",
      "post_date": "03/02/2025 04:03:10",
      "content": "<p>It looks like some atoms are somewhere in space near Earth. But still somehow connected to RNA? Wow, we didn't learn that in school!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3137895,
      "author_name": "shujun717",
      "author_url": "",
      "post_date": "03/02/2025 04:14:32",
      "content": "<p>Yes -1e+18 should be interpreted as NaNs </p>",
      "votes": null,
      "replies": [
        {
          "id": 3138192,
          "author_name": "jagofc",
          "author_url": "",
          "post_date": "03/02/2025 10:32:03",
          "content": "<p>Thank you for confirming, <a href=\"https://www.kaggle.com/shujun717\" target=\"_blank\">@shujun717</a>.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3137801": "In `train_labels.csv`, NaN coordinates appear as blank cells. In contrast, in `validation_labels.csv`, there seem to be a large number of -1e+18 entries. \n\nShould these be interpreted as NaN/unknown, in the same way that blanks are treated as NaN for train_labels? \n\nIf so, why not encode them as NaN/blank?",
    "3137886": "It looks like some atoms are somewhere in space near Earth. But still somehow connected to RNA? Wow, we didn't learn that in school!",
    "3137895": "Yes -1e+18 should be interpreted as NaNs",
    "3138192": "Thank you for confirming, @shujun717."
  },
  "source": "meta"
}