{
  "id": 188636,
  "title": "Here comes the shake up!",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/188636",
  "author_name": "",
  "post_date": "2020-10-04T10:59:32.353732900Z",
  "votes": 7,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Hi all, </p>\n<p>2 days left and from the looks of the close scores lots of people are using at least in part the multi quantile model from <a href=\"https://www.kaggle.com/ulrich07\" target=\"_blank\">@ulrich07</a> and <a href=\"https://www.kaggle.com/reighns\" target=\"_blank\">@reighns</a> effnet. </p>\n<p>I’ve read a few comments that suggest these original notebooks are over fitting the public LB (not a criticism at all of those two legends who I’m sure have gone on to improve massively on those already awesome notebooks - thanks for sharing to both!).</p>\n<p>Are people expecting a big shake up? Personally, even though I’ve trained my own version of reighns’s effnet on some more ctscan features (and finally found my error! ;) ) I’m still expecting to drop massively from current silver spot to about 500th in the private LB. </p>\n<p>What do you think will happen?</p>",
  "messages": [
    {
      "id": "1036863",
      "postDate": "10/04/2020 10:59:32",
      "content": "<p>Hi all, </p>\n<p>2 days left and from the looks of the close scores lots of people are using at least in part the multi quantile model from <a href=\"https://www.kaggle.com/ulrich07\" target=\"_blank\">@ulrich07</a> and <a href=\"https://www.kaggle.com/reighns\" target=\"_blank\">@reighns</a> effnet. </p>\n<p>I’ve read a few comments that suggest these original notebooks are over fitting the public LB (not a criticism at all of those two legends who I’m sure have gone on to improve massively on those already awesome notebooks - thanks for sharing to both!).</p>\n<p>Are people expecting a big shake up? Personally, even though I’ve trained my own version of reighns’s effnet on some more ctscan features (and finally found my error! ;) ) I’m still expecting to drop massively from current silver spot to about 500th in the private LB. </p>\n<p>What do you think will happen?</p>",
      "rawMarkdown": "Hi all, \n\n2 days left and from the looks of the close scores lots of people are using at least in part the multi quantile model from @ulrich07 and @reighns effnet. \n\nI’ve read a few comments that suggest these original notebooks are over fitting the public LB (not a criticism at all of those two legends who I’m sure have gone on to improve massively on those already awesome notebooks - thanks for sharing to both!).\n\nAre people expecting a big shake up? Personally, even though I’ve trained my own version of reighns’s effnet on some more ctscan features (and finally found my error! ;) ) I’m still expecting to drop massively from current silver spot to about 500th in the private LB. \n\nWhat do you think will happen?",
      "votes": null
    },
    {
      "id": "1036883",
      "postDate": "10/04/2020 11:28:31",
      "content": "<p>I'm expecting a shake up close to M5 and LANL</p>",
      "rawMarkdown": "I'm expecting a shake up close to M5 and LANL",
      "votes": null
    },
    {
      "id": "1037098",
      "postDate": "10/04/2020 15:53:29",
      "content": "<p>If I'm just thinking about myself, I really hope there is a biiig shakeup (1420 places to be precise). I've never put so much effort into something with so little (apparent) return. This is a good test of not tying your emotions to public lb.</p>",
      "rawMarkdown": "If I'm just thinking about myself, I really hope there is a biiig shakeup (1420 places to be precise). I've never put so much effort into something with so little (apparent) return. This is a good test of not tying your emotions to public lb.",
      "votes": null
    },
    {
      "id": "1037122",
      "postDate": "10/04/2020 16:16:38",
      "content": "<p>\"This leaderboard is calculated with approximately 15% of the test data. The final results will be based on the other 85%, so the final standings may be different.\"</p>\n<p>As only 15% is there in test data<br>\ntrust your CV<br>\nas long as CV is good, that's the set best!</p>\n<p>All the best!</p>",
      "rawMarkdown": "\"This leaderboard is calculated with approximately 15% of the test data. The final results will be based on the other 85%, so the final standings may be different.\"\n\nAs only 15% is there in test data\ntrust your CV\nas long as CV is good, that's the set best!\n\nAll the best!",
      "votes": null
    },
    {
      "id": "1037145",
      "postDate": "10/04/2020 16:31:46",
      "content": "<p>I love your optimism ;)</p>\n<p>But I really don't know what you mean to say by \"little (apparent) return\".</p>\n<p>If you've put efforts, you'll get definitely get something in return. If not a medal, I'll consider this competition as a really good opportunity to work with real-life noisy data.</p>",
      "rawMarkdown": "I love your optimism ;)\n\nBut I really don't know what you mean to say by \"little (apparent) return\".\n\nIf you've put efforts, you'll get definitely get something in return. If not a medal, I'll consider this competition as a really good opportunity to work with real-life noisy data.",
      "votes": null
    },
    {
      "id": "1037217",
      "postDate": "10/04/2020 17:36:28",
      "content": "<p>I mean that if there is return, it's not apparent, because my public LB standing hasn't budged since my first few submissions.</p>\n<p>But yes, I agree with you that the learnings are clearly good return.</p>",
      "rawMarkdown": "I mean that if there is return, it's not apparent, because my public LB standing hasn't budged since my first few submissions.\n\nBut yes, I agree with you that the learnings are clearly good return.",
      "votes": null
    },
    {
      "id": "1037283",
      "postDate": "10/04/2020 19:27:47",
      "content": "<p>oh, i liked LANL shake-up 😃</p>\n<p>but yeah, could be of similar strength.</p>",
      "rawMarkdown": "oh, i liked LANL shake-up 😃\n\nbut yeah, could be of similar strength.",
      "votes": null
    },
    {
      "id": "1037396",
      "postDate": "10/05/2020 01:06:03",
      "content": "<p>I've never written so much code with only knowledge in return. Our best Public Model is definitely a massively overfitted one. So your hopes can become true in the end.</p>",
      "rawMarkdown": "I've never written so much code with only knowledge in return. Our best Public Model is definitely a massively overfitted one. So your hopes can become true in the end.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1036883,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "10/04/2020 11:28:31",
      "content": "<p>I'm expecting a shake up close to M5 and LANL</p>",
      "votes": null,
      "replies": [
        {
          "id": 1037283,
          "author_name": "ilu000",
          "author_url": "",
          "post_date": "10/04/2020 19:27:47",
          "content": "<p>oh, i liked LANL shake-up 😃</p>\n<p>but yeah, could be of similar strength.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1037098,
      "author_name": "alexandersoare",
      "author_url": "",
      "post_date": "10/04/2020 15:53:29",
      "content": "<p>If I'm just thinking about myself, I really hope there is a biiig shakeup (1420 places to be precise). I've never put so much effort into something with so little (apparent) return. This is a good test of not tying your emotions to public lb.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1037145,
          "author_name": "aadhavvignesh",
          "author_url": "",
          "post_date": "10/04/2020 16:31:46",
          "content": "<p>I love your optimism ;)</p>\n<p>But I really don't know what you mean to say by \"little (apparent) return\".</p>\n<p>If you've put efforts, you'll get definitely get something in return. If not a medal, I'll consider this competition as a really good opportunity to work with real-life noisy data.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1037217,
          "author_name": "alexandersoare",
          "author_url": "",
          "post_date": "10/04/2020 17:36:28",
          "content": "<p>I mean that if there is return, it's not apparent, because my public LB standing hasn't budged since my first few submissions.</p>\n<p>But yes, I agree with you that the learnings are clearly good return.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1037396,
          "author_name": "ronaldokun",
          "author_url": "",
          "post_date": "10/05/2020 01:06:03",
          "content": "<p>I've never written so much code with only knowledge in return. Our best Public Model is definitely a massively overfitted one. So your hopes can become true in the end.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1037122,
      "author_name": "kmldas",
      "author_url": "",
      "post_date": "10/04/2020 16:16:38",
      "content": "<p>\"This leaderboard is calculated with approximately 15% of the test data. The final results will be based on the other 85%, so the final standings may be different.\"</p>\n<p>As only 15% is there in test data<br>\ntrust your CV<br>\nas long as CV is good, that's the set best!</p>\n<p>All the best!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1036863": "Hi all, \n\n2 days left and from the looks of the close scores lots of people are using at least in part the multi quantile model from @ulrich07 and @reighns effnet. \n\nI’ve read a few comments that suggest these original notebooks are over fitting the public LB (not a criticism at all of those two legends who I’m sure have gone on to improve massively on those already awesome notebooks - thanks for sharing to both!).\n\nAre people expecting a big shake up? Personally, even though I’ve trained my own version of reighns’s effnet on some more ctscan features (and finally found my error! ;) ) I’m still expecting to drop massively from current silver spot to about 500th in the private LB. \n\nWhat do you think will happen?",
    "1036883": "I'm expecting a shake up close to M5 and LANL",
    "1037098": "If I'm just thinking about myself, I really hope there is a biiig shakeup (1420 places to be precise). I've never put so much effort into something with so little (apparent) return. This is a good test of not tying your emotions to public lb.",
    "1037122": "\"This leaderboard is calculated with approximately 15% of the test data. The final results will be based on the other 85%, so the final standings may be different.\"\n\nAs only 15% is there in test data\ntrust your CV\nas long as CV is good, that's the set best!\n\nAll the best!",
    "1037145": "I love your optimism ;)\n\nBut I really don't know what you mean to say by \"little (apparent) return\".\n\nIf you've put efforts, you'll get definitely get something in return. If not a medal, I'll consider this competition as a really good opportunity to work with real-life noisy data.",
    "1037217": "I mean that if there is return, it's not apparent, because my public LB standing hasn't budged since my first few submissions.\n\nBut yes, I agree with you that the learnings are clearly good return.",
    "1037283": "oh, i liked LANL shake-up 😃\n\nbut yeah, could be of similar strength.",
    "1037396": "I've never written so much code with only knowledge in return. Our best Public Model is definitely a massively overfitted one. So your hopes can become true in the end."
  },
  "source": "meta"
}