{
  "id": 10232,
  "title": "Training per patient",
  "url": "/competitions/seizure-prediction/discussion/10232",
  "author_name": "",
  "post_date": "2014-09-05T09:36:17.747Z",
  "votes": null,
  "comment_count": 5,
  "views": 1292,
  "content": "<p>Because I saw on the forum of the previous competition that some people train classifiers for each patient, does that mean that you can use the name(not segment) of the file/patient to use a different classifier?</p>",
  "messages": [
    {
      "id": "53163",
      "postDate": "09/05/2014 09:36:17",
      "content": "<p>Because I saw on the forum of the previous competition that some people train classifiers for each patient, does that mean that you can use the name(not segment) of the file/patient to use a different classifier?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "53164",
      "postDate": "09/05/2014 09:47:52",
      "content": "<p>Check the&nbsp;<a href=\"https://www.kaggle.com/c/seizure-prediction/rules\">rules</a>, it says&nbsp;right after 'Competition Timeline' that training on individual subjects is allowed - i.e. use the same algorithm for all subjects - but that we can not use the label directly.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "53165",
      "postDate": "09/05/2014 10:08:13",
      "content": "<p>Scyfer,</p>\n<p>I've read the rules. When they say that this</p>\n<p style=\"padding-left: 30px\"># Not allowed:<br>if 'Dog_1' then foo()<br>if 'Patient_1' then bar()</p>\n<p>and</p>\n<p style=\"padding-left: 30px\">#Also Allowed<br>for subject=1:N<br> train(subject)<br> test(subject)</p>\n<p>I understand that I can use the same technique with different parameters for the 7 subjects, but I'm not allowed to use different techniques, am I right?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "53167",
      "postDate": "09/05/2014 10:37:07",
      "content": "<p>Oke, I think I understand it, but it's a little bit vague. Cuz, lets say in your implementation you train multiple classifiers and take the best and you do that for each patient, than it's the same technique.</p>\n<p>But my conclusion is that this is allowed.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "53168",
      "postDate": "09/05/2014 10:39:41",
      "content": "<p>That's how I understood it. As long as the parameters aren't hardcoded per patient that should be fine.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "53169",
      "postDate": "09/05/2014 10:49:18",
      "content": "<p>Ok, got it.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 53164,
      "author_name": "scyfer",
      "author_url": "",
      "post_date": "09/05/2014 09:47:52",
      "content": "<p>Check the&nbsp;<a href=\"https://www.kaggle.com/c/seizure-prediction/rules\">rules</a>, it says&nbsp;right after 'Competition Timeline' that training on individual subjects is allowed - i.e. use the same algorithm for all subjects - but that we can not use the label directly.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 53165,
      "author_name": "joseleiva",
      "author_url": "",
      "post_date": "09/05/2014 10:08:13",
      "content": "<p>Scyfer,</p>\n<p>I've read the rules. When they say that this</p>\n<p style=\"padding-left: 30px\"># Not allowed:<br>if 'Dog_1' then foo()<br>if 'Patient_1' then bar()</p>\n<p>and</p>\n<p style=\"padding-left: 30px\">#Also Allowed<br>for subject=1:N<br> train(subject)<br> test(subject)</p>\n<p>I understand that I can use the same technique with different parameters for the 7 subjects, but I'm not allowed to use different techniques, am I right?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 53167,
      "author_name": "pbsciencelab",
      "author_url": "",
      "post_date": "09/05/2014 10:37:07",
      "content": "<p>Oke, I think I understand it, but it's a little bit vague. Cuz, lets say in your implementation you train multiple classifiers and take the best and you do that for each patient, than it's the same technique.</p>\n<p>But my conclusion is that this is allowed.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 53168,
      "author_name": "scyfer",
      "author_url": "",
      "post_date": "09/05/2014 10:39:41",
      "content": "<p>That's how I understood it. As long as the parameters aren't hardcoded per patient that should be fine.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 53169,
      "author_name": "joseleiva",
      "author_url": "",
      "post_date": "09/05/2014 10:49:18",
      "content": "<p>Ok, got it.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "53163": "",
    "53164": "",
    "53165": "",
    "53167": "",
    "53168": "",
    "53169": ""
  },
  "source": "meta"
}