{
  "id": 506187,
  "title": "Clarification on multiplying by sample_submission.csv?",
  "url": "/competitions/leap-atmospheric-physics-ai-climsim/discussion/506187",
  "author_name": "Jonathan Randich",
  "post_date": "2024-05-20T20:07:46.187000",
  "votes": 0,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Do we multiply our trained data just by the sample_submission file?</p>\n<p>Or do we first scale each column of our submission by its std dev? Then multiply by the sample_submission?</p>\n<p>And when finding the std devs, is that before or after we zero out the specified columns?</p>\n<p>Unsure the right order of the steps like:</p>\n<p>train your model, scale each variable by std dev, zero out specified columns, multiply by sample_submission</p>\n<p>^or something like that. what's the right order? or what am I misintepreting?</p>",
  "messages": [
    {
      "id": 2827468,
      "postDate": "2024-05-21T13:46:24.883Z",
      "content": "<p>Yes, the same problem here. It would be helpful to get one model line of submission. I also find the instruction unclear at this… </p>\n<p>My guess is that our submisson.csv should be in the same units as targets are in train.csv</p>",
      "rawMarkdown": "Yes, the same problem here. It would be helpful to get one model line of submission. I also find the instruction unclear at this... \n\nMy guess is that our submisson.csv should be in the same units as targets are in train.csv",
      "votes": 1
    },
    {
      "id": 2826278,
      "postDate": "2024-05-20T20:07:46.187Z",
      "content": "<p>Do we multiply our trained data just by the sample_submission file?</p>\n<p>Or do we first scale each column of our submission by its std dev? Then multiply by the sample_submission?</p>\n<p>And when finding the std devs, is that before or after we zero out the specified columns?</p>\n<p>Unsure the right order of the steps like:</p>\n<p>train your model, scale each variable by std dev, zero out specified columns, multiply by sample_submission</p>\n<p>^or something like that. what's the right order? or what am I misintepreting?</p>",
      "rawMarkdown": "Do we multiply our trained data just by the sample_submission file?\n\nOr do we first scale each column of our submission by its std dev? Then multiply by the sample_submission?\n\nAnd when finding the std devs, is that before or after we zero out the specified columns?\n\nUnsure the right order of the steps like:\n\ntrain your model, scale each variable by std dev, zero out specified columns, multiply by sample_submission\n\n^or something like that. what's the right order? or what am I misintepreting?"
    }
  ],
  "comments": [
    {
      "id": 2827468,
      "author_name": "Markus Kaukonen",
      "author_url": "",
      "post_date": "2024-05-21T13:46:24.883000",
      "content": "<p>Yes, the same problem here. It would be helpful to get one model line of submission. I also find the instruction unclear at this… </p>\n<p>My guess is that our submisson.csv should be in the same units as targets are in train.csv</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2827468": "Yes, the same problem here. It would be helpful to get one model line of submission. I also find the instruction unclear at this... \n\nMy guess is that our submisson.csv should be in the same units as targets are in train.csv",
    "2826278": "Do we multiply our trained data just by the sample_submission file?\n\nOr do we first scale each column of our submission by its std dev? Then multiply by the sample_submission?\n\nAnd when finding the std devs, is that before or after we zero out the specified columns?\n\nUnsure the right order of the steps like:\n\ntrain your model, scale each variable by std dev, zero out specified columns, multiply by sample_submission\n\n^or something like that. what's the right order? or what am I misintepreting?"
  }
}