{
  "id": 16220,
  "title": "How to interpret difference in dimensions in training.csv and test.csv",
  "url": "/competitions/flavours-of-physics/discussion/16220",
  "author_name": "",
  "post_date": "2015-08-30T11:39:10.180Z",
  "votes": null,
  "comment_count": 1,
  "views": 370,
  "content": "<p>The size of vectors in file training.csv is greater than test.csv. Mass, production, min_ANNmuon are absent in test.csv. What is the reason of dimension's reduction? Neural networks are usually trained and tested with the input vectors of the same dimension. </p>\n\n<p>Are absent values of vectors turn to 0 without reducing the dimension of vectors in test.csv?</p>",
  "messages": [
    {
      "id": "90940",
      "postDate": "08/30/2015 11:39:10",
      "content": "<p>The size of vectors in file training.csv is greater than test.csv. Mass, production, min_ANNmuon are absent in test.csv. What is the reason of dimension's reduction? Neural networks are usually trained and tested with the input vectors of the same dimension. </p>\n\n<p>Are absent values of vectors turn to 0 without reducing the dimension of vectors in test.csv?</p>",
      "rawMarkdown": "The size of vectors in file training.csv is greater than test.csv. Mass, production, min_ANNmuon are absent in test.csv. What is the reason of dimension's reduction? Neural networks are usually trained and tested with the input vectors of the same dimension. \r\n \r\nAre absent values of vectors turn to 0 without reducing the dimension of vectors in test.csv?",
      "votes": null
    },
    {
      "id": "91003",
      "postDate": "08/31/2015 07:49:53",
      "content": "<p>Short answer: don't use these features which don't exist in the test sample.</p>\n\n<p>A little bit longer one is that, some features only exist in the simulation which come from the ground truth (computer simulation), and in real life, there are no such features. For example, production feature is for the production channel of tau in the simulation, and in real life, tau production channel can not be observed.</p>",
      "rawMarkdown": "Short answer: don't use these features which don't exist in the test sample.\r\n\r\nA little bit longer one is that, some features only exist in the simulation which come from the ground truth (computer simulation), and in real life, there are no such features. For example, production feature is for the production channel of tau in the simulation, and in real life, tau production channel can not be observed.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 91003,
      "author_name": "phunter",
      "author_url": "",
      "post_date": "08/31/2015 07:49:53",
      "content": "<p>Short answer: don't use these features which don't exist in the test sample.</p>\n\n<p>A little bit longer one is that, some features only exist in the simulation which come from the ground truth (computer simulation), and in real life, there are no such features. For example, production feature is for the production channel of tau in the simulation, and in real life, tau production channel can not be observed.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "90940": "The size of vectors in file training.csv is greater than test.csv. Mass, production, min_ANNmuon are absent in test.csv. What is the reason of dimension's reduction? Neural networks are usually trained and tested with the input vectors of the same dimension. \r\n \r\nAre absent values of vectors turn to 0 without reducing the dimension of vectors in test.csv?",
    "91003": "Short answer: don't use these features which don't exist in the test sample.\r\n\r\nA little bit longer one is that, some features only exist in the simulation which come from the ground truth (computer simulation), and in real life, there are no such features. For example, production feature is for the production channel of tau in the simulation, and in real life, tau production channel can not be observed."
  },
  "source": "meta"
}