{
  "id": 15740,
  "title": "to admin: is it OK to use \"non-physics\" featres?",
  "url": "/competitions/flavours-of-physics/discussion/15740",
  "author_name": "",
  "post_date": "2015-08-03T21:48:12.527Z",
  "votes": 3,
  "comment_count": 1,
  "views": 747,
  "content": "<p>Is it OK to use a feature extract from the training set that have no relation to the physical world but have a relation to how programmers would have programmed a simulator?</p>\n\n<p>After all the training set is just a set of numbers and any features extracted from them should be OK?</p>",
  "messages": [
    {
      "id": "88164",
      "postDate": "08/03/2015 21:48:12",
      "content": "<p>Is it OK to use a feature extract from the training set that have no relation to the physical world but have a relation to how programmers would have programmed a simulator?</p>\n\n<p>After all the training set is just a set of numbers and any features extracted from them should be OK?</p>",
      "rawMarkdown": "Is it OK to use a feature extract from the training set that have no relation to the physical world but have a relation to how programmers would have programmed a simulator?\r\n\r\nAfter all the training set is just a set of numbers and any features extracted from them should be OK?",
      "votes": null
    },
    {
      "id": "88748",
      "postDate": "08/07/2015 00:29:54",
      "content": "<p>I may be compeletly wrong, and I may have bugs, but if you look at this plot <a href=\"https://www.kaggle.com/udibr1/flavours-of-physics/quantize\">https://www.kaggle.com/udibr1/flavours-of-physics/quantize</a> you will see that the feature 'IPSig' has different behaviour in signal and background on the training set which looks to me like a numerical error generated when running the simulation for the signal part of the set. The background contains real physical measurements and therefore does not suffer from this effect.</p>\n\n<p>This could also be just a random artificat and I'm sorry in advance for any mistake.</p>",
      "rawMarkdown": "I may be compeletly wrong, and I may have bugs, but if you look at this plot https://www.kaggle.com/udibr1/flavours-of-physics/quantize you will see that the feature 'IPSig' has different behaviour in signal and background on the training set which looks to me like a numerical error generated when running the simulation for the signal part of the set. The background contains real physical measurements and therefore does not suffer from this effect.\r\n\r\nThis could also be just a random artificat and I'm sorry in advance for any mistake.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 88748,
      "author_name": "udibr1",
      "author_url": "",
      "post_date": "08/07/2015 00:29:54",
      "content": "<p>I may be compeletly wrong, and I may have bugs, but if you look at this plot <a href=\"https://www.kaggle.com/udibr1/flavours-of-physics/quantize\">https://www.kaggle.com/udibr1/flavours-of-physics/quantize</a> you will see that the feature 'IPSig' has different behaviour in signal and background on the training set which looks to me like a numerical error generated when running the simulation for the signal part of the set. The background contains real physical measurements and therefore does not suffer from this effect.</p>\n\n<p>This could also be just a random artificat and I'm sorry in advance for any mistake.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "88164": "Is it OK to use a feature extract from the training set that have no relation to the physical world but have a relation to how programmers would have programmed a simulator?\r\n\r\nAfter all the training set is just a set of numbers and any features extracted from them should be OK?",
    "88748": "I may be compeletly wrong, and I may have bugs, but if you look at this plot https://www.kaggle.com/udibr1/flavours-of-physics/quantize you will see that the feature 'IPSig' has different behaviour in signal and background on the training set which looks to me like a numerical error generated when running the simulation for the signal part of the set. The background contains real physical measurements and therefore does not suffer from this effect.\r\n\r\nThis could also be just a random artificat and I'm sorry in advance for any mistake."
  },
  "source": "meta"
}