{
  "id": 7887,
  "title": "Logistic Regression with Scikit Learn",
  "url": "/competitions/decoding-the-human-brain/discussion/7887",
  "author_name": "",
  "post_date": "2014-04-26T20:01:04.837Z",
  "votes": null,
  "comment_count": 2,
  "views": 1252,
  "content": "<p>Running my own version of the benchmark.&nbsp; I've got 16 gigs.&nbsp; Before zscoring, it trained rather quickly.&nbsp; After zscoring - this is slow as molasses.&nbsp; I haven't actually had the patience to let it complete a training pass (or rather, I feared it may have gone into the for-all-intents-and-purposes infinite computation time regime).&nbsp; I've got a 4Ghz octo-core processor.&nbsp; How long would you expect this would take to complete a single training pass?&nbsp; I'm assuming the zscoring has totally changed the interaction with the tolerance, trying to grok a simple way of determining to what degree.&nbsp;</p>\n\n<p>Any suggestions?</p>",
  "messages": [
    {
      "id": "43101",
      "postDate": "04/26/2014 20:01:04",
      "content": "<p>Running my own version of the benchmark.&nbsp; I've got 16 gigs.&nbsp; Before zscoring, it trained rather quickly.&nbsp; After zscoring - this is slow as molasses.&nbsp; I haven't actually had the patience to let it complete a training pass (or rather, I feared it may have gone into the for-all-intents-and-purposes infinite computation time regime).&nbsp; I've got a 4Ghz octo-core processor.&nbsp; How long would you expect this would take to complete a single training pass?&nbsp; I'm assuming the zscoring has totally changed the interaction with the tolerance, trying to grok a simple way of determining to what degree.&nbsp;</p>\n\n<p>Any suggestions?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "43102",
      "postDate": "04/26/2014 20:15:13",
      "content": "<p>Annnd - yes - I set the tolerance to 1 and it trains much faster, gives a reasonable xval.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "43104",
      "postDate": "04/26/2014 20:28:39",
      "content": "<p>Annnd it's totally underfitting.&nbsp; I set the tolerance to 1/10 annnnd now I'm waiting forever again.&nbsp; Some wisdom would be nice here...</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 43102,
      "author_name": "phillipadkins",
      "author_url": "",
      "post_date": "04/26/2014 20:15:13",
      "content": "<p>Annnd - yes - I set the tolerance to 1 and it trains much faster, gives a reasonable xval.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 43104,
      "author_name": "phillipadkins",
      "author_url": "",
      "post_date": "04/26/2014 20:28:39",
      "content": "<p>Annnd it's totally underfitting.&nbsp; I set the tolerance to 1/10 annnnd now I'm waiting forever again.&nbsp; Some wisdom would be nice here...</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "43101": "",
    "43102": "",
    "43104": ""
  },
  "source": "meta"
}