{
  "id": 189655,
  "title": "Still one remaining question",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/189655",
  "author_name": "Alex",
  "post_date": "2020-10-08T07:53:22.392000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Congratulations to everyone ! This was my first challenge dealing with tabular data and learnt a lot, it was quite instructive. I've got some solutions reaching top 8/10% in private LB but did not select it as lot of people apparently but its nothing in comparison with the amount of things that I've learnt. <br>\nI've went through the top solution summaries and something remains unclear to me:<br>\n-&gt; In case of kaggle challenges should we normalize the training data including test set statistics ? In real world problems somehow it is a leak but would it help to get better LB here on kaggle ? </p>",
  "messages": [
    {
      "id": 1042394,
      "postDate": "2020-10-08T07:53:22.393Z",
      "content": "<p>Congratulations to everyone ! This was my first challenge dealing with tabular data and learnt a lot, it was quite instructive. I've got some solutions reaching top 8/10% in private LB but did not select it as lot of people apparently but its nothing in comparison with the amount of things that I've learnt. <br>\nI've went through the top solution summaries and something remains unclear to me:<br>\n-&gt; In case of kaggle challenges should we normalize the training data including test set statistics ? In real world problems somehow it is a leak but would it help to get better LB here on kaggle ? </p>",
      "rawMarkdown": "Congratulations to everyone ! This was my first challenge dealing with tabular data and learnt a lot, it was quite instructive. I've got some solutions reaching top 8/10% in private LB but did not select it as lot of people apparently but its nothing in comparison with the amount of things that I've learnt. \nI've went through the top solution summaries and something remains unclear to me:\n-> In case of kaggle challenges should we normalize the training data including test set statistics ? In real world problems somehow it is a leak but would it help to get better LB here on kaggle ? "
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1042394": "Congratulations to everyone ! This was my first challenge dealing with tabular data and learnt a lot, it was quite instructive. I've got some solutions reaching top 8/10% in private LB but did not select it as lot of people apparently but its nothing in comparison with the amount of things that I've learnt. \nI've went through the top solution summaries and something remains unclear to me:\n-> In case of kaggle challenges should we normalize the training data including test set statistics ? In real world problems somehow it is a leak but would it help to get better LB here on kaggle ? "
  }
}