{
  "id": 22182,
  "title": "Congrats to the \"Devil Team\"",
  "url": "/competitions/avito-duplicate-ads-detection/discussion/22182",
  "author_name": "",
  "post_date": "2016-07-12T03:32:32.967Z",
  "votes": 8,
  "comment_count": 2,
  "views": 696,
  "content": "<p>Congratulations to the Devil Team! </p>\n\n<p>Winning with a gap 0.5% in AUC in such a contest is really outstanding. Even more impressive was the strategy to hold hands still and to submit a leap improvement in the last hour of the competition. Very cool, indeed! </p>\n\n<p>The contest demonstrated that machine learning is more than algorithms, parameter tuning and brute force computing.  It is more than applying many different models, bagging, ensembling, stacking and meta-modelling - which is kind of boring activity. </p>\n\n<p><strong>Machine learning starts and ends with human brains understanding the data, feeding the machines, and interpreting the results - at least for some more years.</strong> </p>\n\n<p>Thanks to Avito and Kaggle for organizing this competition. </p>\n\n<p>Peter Borrmann, The Quants</p>",
  "messages": [
    {
      "id": "126741",
      "postDate": "07/12/2016 03:32:32",
      "content": "<p>Congratulations to the Devil Team! </p>\n\n<p>Winning with a gap 0.5% in AUC in such a contest is really outstanding. Even more impressive was the strategy to hold hands still and to submit a leap improvement in the last hour of the competition. Very cool, indeed! </p>\n\n<p>The contest demonstrated that machine learning is more than algorithms, parameter tuning and brute force computing.  It is more than applying many different models, bagging, ensembling, stacking and meta-modelling - which is kind of boring activity. </p>\n\n<p><strong>Machine learning starts and ends with human brains understanding the data, feeding the machines, and interpreting the results - at least for some more years.</strong> </p>\n\n<p>Thanks to Avito and Kaggle for organizing this competition. </p>\n\n<p>Peter Borrmann, The Quants</p>",
      "rawMarkdown": "Congratulations to the Devil Team! \r\n\r\nWinning with a gap 0.5% in AUC in such a contest is really outstanding. Even more impressive was the strategy to hold hands still and to submit a leap improvement in the last hour of the competition. Very cool, indeed! \r\n\r\nThe contest demonstrated that machine learning is more than algorithms, parameter tuning and brute force computing.  It is more than applying many different models, bagging, ensembling, stacking and meta-modelling - which is kind of boring activity. \r\n\r\n**Machine learning starts and ends with human brains understanding the data, feeding the machines, and interpreting the results - at least for some more years.** \r\n\r\nThanks to Avito and Kaggle for organizing this competition. \r\n\r\nPeter Borrmann, The Quants",
      "votes": null
    },
    {
      "id": "126768",
      "postDate": "07/12/2016 07:27:33",
      "content": "<p>Congratulation to u1234x1234 and Snatislav! So great job!</p>\n\n<p>And I really love current friendly battle for TOP1. Gilberto, Marios and Stas have different approach, but they are so smart. And all of them deserve&#8203; status TOP1.</p>",
      "rawMarkdown": "Congratulation to u1234x1234 and Snatislav! So great job!\r\n\r\nAnd I really love current friendly battle for TOP1. Gilberto, Marios and Stas have different approach, but they are so smart. And all of them deserve status TOP1.",
      "votes": null
    },
    {
      "id": "126779",
      "postDate": "07/12/2016 09:23:20",
      "content": "<p>Well said Peter. Congratulations to the amazing scores of both &quot;Devil Team&quot; and &quot;TheQuants&quot; -- you guys are an inspiration for the rest of us mortals :)</p>",
      "rawMarkdown": "Well said Peter. Congratulations to the amazing scores of both \"Devil Team\" and \"TheQuants\" -- you guys are an inspiration for the rest of us mortals :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 126768,
      "author_name": "dowakin",
      "author_url": "",
      "post_date": "07/12/2016 07:27:33",
      "content": "<p>Congratulation to u1234x1234 and Snatislav! So great job!</p>\n\n<p>And I really love current friendly battle for TOP1. Gilberto, Marios and Stas have different approach, but they are so smart. And all of them deserve&#8203; status TOP1.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 126779,
      "author_name": "yilihome",
      "author_url": "",
      "post_date": "07/12/2016 09:23:20",
      "content": "<p>Well said Peter. Congratulations to the amazing scores of both &quot;Devil Team&quot; and &quot;TheQuants&quot; -- you guys are an inspiration for the rest of us mortals :)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "126741": "Congratulations to the Devil Team! \r\n\r\nWinning with a gap 0.5% in AUC in such a contest is really outstanding. Even more impressive was the strategy to hold hands still and to submit a leap improvement in the last hour of the competition. Very cool, indeed! \r\n\r\nThe contest demonstrated that machine learning is more than algorithms, parameter tuning and brute force computing.  It is more than applying many different models, bagging, ensembling, stacking and meta-modelling - which is kind of boring activity. \r\n\r\n**Machine learning starts and ends with human brains understanding the data, feeding the machines, and interpreting the results - at least for some more years.** \r\n\r\nThanks to Avito and Kaggle for organizing this competition. \r\n\r\nPeter Borrmann, The Quants",
    "126768": "Congratulation to u1234x1234 and Snatislav! So great job!\r\n\r\nAnd I really love current friendly battle for TOP1. Gilberto, Marios and Stas have different approach, but they are so smart. And all of them deserve status TOP1.",
    "126779": "Well said Peter. Congratulations to the amazing scores of both \"Devil Team\" and \"TheQuants\" -- you guys are an inspiration for the rest of us mortals :)"
  },
  "source": "meta"
}