{
  "id": 12533,
  "title": "How low can it go?",
  "url": "/competitions/malware-classification/discussion/12533",
  "author_name": "",
  "post_date": "2015-02-16T17:20:02.290Z",
  "votes": 1,
  "comment_count": 9,
  "views": 2631,
  "content": "<p>Any guesses?&nbsp;</p>",
  "messages": [
    {
      "id": "64350",
      "postDate": "02/16/2015 17:20:02",
      "content": "<p>Any guesses?&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64352",
      "postDate": "02/16/2015 17:46:58",
      "content": "<p>60 days</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64353",
      "postDate": "02/16/2015 17:47:26",
      "content": "<p>i meant the score.. :D</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64354",
      "postDate": "02/16/2015 17:57:17",
      "content": "<p>That's trickier. &nbsp;:-)</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64361",
      "postDate": "02/16/2015 18:55:22",
      "content": "<p>I think &lt;0.005 :P</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64365",
      "postDate": "02/16/2015 19:53:09",
      "content": "<p>For this particular set up, with 9 classes, only 30% of test data for public score and using multi-log-loss as metric, then my money is on 0.00417096 as the minimum achievable</p>\n<p>How I got that number? after a bunch of probably very twisted and wrong mathematics :D</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64730",
      "postDate": "02/22/2015 08:51:55",
      "content": "<p>already 0.007 :D</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "66191",
      "postDate": "03/14/2015 17:07:35",
      "content": "<p>[quote=NxGTR;64365]</p>\n<p>For this particular set up, with 9 classes, only 30% of test data for public score and using multi-log-loss as metric, then my money is on 0.00417096 as the minimum achievable</p>\n<p>How I got that number? after a bunch of probably very twisted and wrong mathematics :D</p>\n<p>[/quote]</p>\n\n<p>We are approaching your figure.... :)</p>\n<p>Still a month to go..</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "66192",
      "postDate": "03/14/2015 17:10:56",
      "content": "<p>given the amount of data in this competition, its too easy to overfit...&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "66195",
      "postDate": "03/14/2015 17:34:27",
      "content": "<p>[quote=Stergios;66191]<br>We are approaching your figure.... :)<br>Still a month to go..<br>[/quote]</p>\n<p>Yes indeed, however even at my not-so-shocking 0.011x, I am already worried about overfitting, I built a kind-of-simple model to reach 0.012x, but reaching 0.011x required me a much more complex model for very minimal gain.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 64352,
      "author_name": "inversion",
      "author_url": "",
      "post_date": "02/16/2015 17:46:58",
      "content": "<p>60 days</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64353,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "02/16/2015 17:47:26",
      "content": "<p>i meant the score.. :D</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64354,
      "author_name": "inversion",
      "author_url": "",
      "post_date": "02/16/2015 17:57:17",
      "content": "<p>That's trickier. &nbsp;:-)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64361,
      "author_name": "jiweiliu",
      "author_url": "",
      "post_date": "02/16/2015 18:55:22",
      "content": "<p>I think &lt;0.005 :P</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64365,
      "author_name": "carloshuertas",
      "author_url": "",
      "post_date": "02/16/2015 19:53:09",
      "content": "<p>For this particular set up, with 9 classes, only 30% of test data for public score and using multi-log-loss as metric, then my money is on 0.00417096 as the minimum achievable</p>\n<p>How I got that number? after a bunch of probably very twisted and wrong mathematics :D</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64730,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "02/22/2015 08:51:55",
      "content": "<p>already 0.007 :D</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 66191,
      "author_name": "asterios",
      "author_url": "",
      "post_date": "03/14/2015 17:07:35",
      "content": "<p>[quote=NxGTR;64365]</p>\n<p>For this particular set up, with 9 classes, only 30% of test data for public score and using multi-log-loss as metric, then my money is on 0.00417096 as the minimum achievable</p>\n<p>How I got that number? after a bunch of probably very twisted and wrong mathematics :D</p>\n<p>[/quote]</p>\n\n<p>We are approaching your figure.... :)</p>\n<p>Still a month to go..</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 66192,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "03/14/2015 17:10:56",
      "content": "<p>given the amount of data in this competition, its too easy to overfit...&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 66195,
      "author_name": "carloshuertas",
      "author_url": "",
      "post_date": "03/14/2015 17:34:27",
      "content": "<p>[quote=Stergios;66191]<br>We are approaching your figure.... :)<br>Still a month to go..<br>[/quote]</p>\n<p>Yes indeed, however even at my not-so-shocking 0.011x, I am already worried about overfitting, I built a kind-of-simple model to reach 0.012x, but reaching 0.011x required me a much more complex model for very minimal gain.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "64350": "",
    "64352": "",
    "64353": "",
    "64354": "",
    "64361": "",
    "64365": "",
    "64730": "",
    "66191": "",
    "66192": "",
    "66195": ""
  },
  "source": "meta"
}