{
  "id": 169151,
  "title": "Congratulations & Foreseeing the Shakeup",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/169151",
  "author_name": "Qishen Ha",
  "post_date": "2020-07-23T04:01:42.044000",
  "votes": 46,
  "comment_count": 13,
  "views": 0,
  "content": "<p>Congratulations for all winners! I'm glad to see that finnally many experienced people end up remaining in the gold zone in this shakeup game!</p>\n\n<p>I stopped working on this competition 2 month ago for a very simple observation:</p>\n\n<ul>\n<li>I submit a single model and got public LB 0.87</li>\n<li>Then I submit the 5-fold version ensemble and got public LB 0.85 (Though all of these models got around CV 0.90x)</li>\n<li>Public testset contains only around 400~ samples and private testset is no that better (600~ samples)</li>\n</ul>\n\n<p>That means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition and at last moment I choose my early submissions of this competition (with much worse CV/LB than my public one) and got private LB 0.88x to act like that I didn't work on this game 🙌 </p>\n\n<p>And I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking, while didn't got any medal!</p>\n\n<p>Congratulations again and thank you for reading!</p>",
  "messages": [
    {
      "id": 940629,
      "postDate": "2020-07-23T04:01:42.043Z",
      "content": "<p>Congratulations for all winners! I'm glad to see that finnally many experienced people end up remaining in the gold zone in this shakeup game!</p>\n\n<p>I stopped working on this competition 2 month ago for a very simple observation:</p>\n\n<ul>\n<li>I submit a single model and got public LB 0.87</li>\n<li>Then I submit the 5-fold version ensemble and got public LB 0.85 (Though all of these models got around CV 0.90x)</li>\n<li>Public testset contains only around 400~ samples and private testset is no that better (600~ samples)</li>\n</ul>\n\n<p>That means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition and at last moment I choose my early submissions of this competition (with much worse CV/LB than my public one) and got private LB 0.88x to act like that I didn't work on this game 🙌 </p>\n\n<p>And I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking, while didn't got any medal!</p>\n\n<p>Congratulations again and thank you for reading!</p>",
      "rawMarkdown": "Congratulations for all winners! I'm glad to see that finnally many experienced people end up remaining in the gold zone in this shakeup game!\n\nI stopped working on this competition 2 month ago for a very simple observation:\n\n* I submit a single model and got public LB 0.87\n* Then I submit the 5-fold version ensemble and got public LB 0.85 (Though all of these models got around CV 0.90x)\n* Public testset contains only around 400~ samples and private testset is no that better (600~ samples)\n\nThat means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition and at last moment I choose my early submissions of this competition (with much worse CV/LB than my public one) and got private LB 0.88x to act like that I didn't work on this game 🙌 \n\nAnd I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking, while didn't got any medal!\n\nCongratulations again and thank you for reading!",
      "votes": 46
    },
    {
      "id": 940690,
      "postDate": "2020-07-23T04:58:56.883Z",
      "content": "<p>Thank you for your sharings. Now I understand why you weren’t submitting new submissions. You and lafoss are the winners of this competition in my opinion. </p>",
      "rawMarkdown": "Thank you for your sharings. Now I understand why you weren’t submitting new submissions. You and lafoss are the winners of this competition in my opinion. ",
      "votes": 12,
      "replies": [
        {
          "id": 940824,
          "postDate": "2020-07-23T05:59:03.643Z",
          "content": "<p>Thanks! 🙌 </p>",
          "rawMarkdown": "Thanks! 🙌 ",
          "votes": 2
        }
      ]
    },
    {
      "id": 940644,
      "postDate": "2020-07-23T04:14:31.870Z",
      "content": "<p><strong>And I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking</strong> that's  the biggest success for you in this competition,it is about the \"strongest single public model of this competition\" and you had that and also you could keep that as private but you didn't,you shared your knowledge and that takes courage,thank you a ton for your awesome kernel,it helped a lot of people in this competition and i am sure a lot of researcher will get huge benefit from your awesome kernel\nthank you a lot <a href=\"/haqishen\">@haqishen</a>  san for sharing wisdom\nyour codes are poetry :) pleasure reading those :) </p>",
      "rawMarkdown": "**And I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking** that's  the biggest success for you in this competition,it is about the \"strongest single public model of this competition\" and you had that and also you could keep that as private but you didn't,you shared your knowledge and that takes courage,thank you a ton for your awesome kernel,it helped a lot of people in this competition and i am sure a lot of researcher will get huge benefit from your awesome kernel\nthank you a lot @haqishen  san for sharing wisdom\nyour codes are poetry :) pleasure reading those :) ",
      "votes": 5,
      "replies": [
        {
          "id": 940654,
          "postDate": "2020-07-23T04:22:58.307Z",
          "content": "<p>Thanks for you kind words ;)</p>",
          "rawMarkdown": "Thanks for you kind words ;)",
          "votes": 2
        }
      ]
    },
    {
      "id": 943955,
      "postDate": "2020-07-24T17:58:16.053Z",
      "content": "<p><a href=\"/haqishen\">@haqishen</a> Thank you for your notebook it really helped me out, with getting the idea of what was to be done</p>",
      "rawMarkdown": "@haqishen Thank you for your notebook it really helped me out, with getting the idea of what was to be done",
      "votes": 1
    },
    {
      "id": 941989,
      "postDate": "2020-07-23T14:31:04.943Z",
      "content": "<p>Thank you very much <a href=\"/haqishen\">@haqishen</a> your notebook was another game changer in this competition. Thanks for sharing your knowledge and helping others to learn. 🙌 </p>",
      "rawMarkdown": "Thank you very much @haqishen your notebook was another game changer in this competition. Thanks for sharing your knowledge and helping others to learn. 🙌 ",
      "votes": 1
    },
    {
      "id": 941881,
      "postDate": "2020-07-23T13:23:39.720Z",
      "content": "<p>Although you didn't win, but your contribution was great. Your model help many achieved the target that you couldn't , I think that's the win for you. \nThanks </p>",
      "rawMarkdown": "Although you didn't win, but your contribution was great. Your model help many achieved the target that you couldn't , I think that's the win for you. \nThanks ",
      "votes": 1
    },
    {
      "id": 941706,
      "postDate": "2020-07-23T11:14:47.583Z",
      "content": "<p>I have learned a lot from your great notebook - many thanks!</p>",
      "rawMarkdown": "I have learned a lot from your great notebook - many thanks!",
      "votes": 1
    },
    {
      "id": 941247,
      "postDate": "2020-07-23T06:35:55.783Z",
      "content": "<p>Thank you so much! For me, iafoss's tiling image is game changer 1 and your bce loss is game changer 2. 😄 </p>",
      "rawMarkdown": "Thank you so much! For me, iafoss's tiling image is game changer 1 and your bce loss is game changer 2. 😄 ",
      "votes": 1
    },
    {
      "id": 941032,
      "postDate": "2020-07-23T06:04:20.930Z",
      "content": "<p>Thank you for your contribution and sharing. </p>",
      "rawMarkdown": "Thank you for your contribution and sharing. ",
      "votes": 1
    },
    {
      "id": 941745,
      "postDate": "2020-07-23T11:41:10.550Z",
      "content": "<p>&gt; That means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition</p>\n\n<p>GrandMaster Wisdom ^^</p>",
      "rawMarkdown": "&gt; That means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition\n\nGrandMaster Wisdom ^^",
      "votes": 2
    },
    {
      "id": 941240,
      "postDate": "2020-07-23T06:31:37.583Z",
      "content": "<p>Thank you <a href=\"/haqishen\">@haqishen</a> for your sharing here and also in Alaska2 on modifying EfficientNet architecture (had that ended sooner might have liked to try some thing with that info here!).  Your kernel did well with TTA variations and combinations, just did not pick any of them!!  But it is possible some in medals got there with your kernel.  </p>\n\n<p>You and <a href=\"/iafoss\">@iafoss</a> deserve some special award/recognition for your contributions here! </p>",
      "rawMarkdown": "Thank you @haqishen for your sharing here and also in Alaska2 on modifying EfficientNet architecture (had that ended sooner might have liked to try some thing with that info here!).  Your kernel did well with TTA variations and combinations, just did not pick any of them!!  But it is possible some in medals got there with your kernel.  \n\nYou and @iafoss deserve some special award/recognition for your contributions here! ",
      "votes": 2
    },
    {
      "id": 943671,
      "postDate": "2020-07-24T14:08:44.130Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 940690,
      "author_name": "Sedat",
      "author_url": "",
      "post_date": "2020-07-23T04:58:56.883000",
      "content": "<p>Thank you for your sharings. Now I understand why you weren’t submitting new submissions. You and lafoss are the winners of this competition in my opinion. </p>",
      "votes": 12,
      "replies": [
        {
          "id": 940824,
          "author_name": "Qishen Ha",
          "author_url": "",
          "post_date": "2020-07-23T05:59:03.643000",
          "content": "<p>Thanks! 🙌 </p>",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 940644,
      "author_name": "Mobassir",
      "author_url": "",
      "post_date": "2020-07-23T04:14:31.870000",
      "content": "<p><strong>And I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking</strong> that's  the biggest success for you in this competition,it is about the \"strongest single public model of this competition\" and you had that and also you could keep that as private but you didn't,you shared your knowledge and that takes courage,thank you a ton for your awesome kernel,it helped a lot of people in this competition and i am sure a lot of researcher will get huge benefit from your awesome kernel\nthank you a lot <a href=\"/haqishen\">@haqishen</a>  san for sharing wisdom\nyour codes are poetry :) pleasure reading those :) </p>",
      "votes": 5,
      "replies": [
        {
          "id": 940654,
          "author_name": "Qishen Ha",
          "author_url": "",
          "post_date": "2020-07-23T04:22:58.307000",
          "content": "<p>Thanks for you kind words ;)</p>",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 943955,
      "author_name": "Digvijay Yadav",
      "author_url": "",
      "post_date": "2020-07-24T17:58:16.053000",
      "content": "<p><a href=\"/haqishen\">@haqishen</a> Thank you for your notebook it really helped me out, with getting the idea of what was to be done</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 941989,
      "author_name": "ratan rohith",
      "author_url": "",
      "post_date": "2020-07-23T14:31:04.943000",
      "content": "<p>Thank you very much <a href=\"/haqishen\">@haqishen</a> your notebook was another game changer in this competition. Thanks for sharing your knowledge and helping others to learn. 🙌 </p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 941881,
      "author_name": "RAHUL SINGH INDA",
      "author_url": "",
      "post_date": "2020-07-23T13:23:39.720000",
      "content": "<p>Although you didn't win, but your contribution was great. Your model help many achieved the target that you couldn't , I think that's the win for you. \nThanks </p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 941706,
      "author_name": "Ivan L",
      "author_url": "",
      "post_date": "2020-07-23T11:14:47.583000",
      "content": "<p>I have learned a lot from your great notebook - many thanks!</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 941247,
      "author_name": "Shiyuan Zeng",
      "author_url": "",
      "post_date": "2020-07-23T06:35:55.783000",
      "content": "<p>Thank you so much! For me, iafoss's tiling image is game changer 1 and your bce loss is game changer 2. 😄 </p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 941032,
      "author_name": "Vee",
      "author_url": "",
      "post_date": "2020-07-23T06:04:20.930000",
      "content": "<p>Thank you for your contribution and sharing. </p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 941745,
      "author_name": "Benjamin Dubreu",
      "author_url": "",
      "post_date": "2020-07-23T11:41:10.550000",
      "content": "<p>&gt; That means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition</p>\n\n<p>GrandMaster Wisdom ^^</p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 941240,
      "author_name": "something4kag",
      "author_url": "",
      "post_date": "2020-07-23T06:31:37.583000",
      "content": "<p>Thank you <a href=\"/haqishen\">@haqishen</a> for your sharing here and also in Alaska2 on modifying EfficientNet architecture (had that ended sooner might have liked to try some thing with that info here!).  Your kernel did well with TTA variations and combinations, just did not pick any of them!!  But it is possible some in medals got there with your kernel.  </p>\n\n<p>You and <a href=\"/iafoss\">@iafoss</a> deserve some special award/recognition for your contributions here! </p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 943671,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-24T14:08:44.130000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "940629": "Congratulations for all winners! I'm glad to see that finnally many experienced people end up remaining in the gold zone in this shakeup game!\n\nI stopped working on this competition 2 month ago for a very simple observation:\n\n* I submit a single model and got public LB 0.87\n* Then I submit the 5-fold version ensemble and got public LB 0.85 (Though all of these models got around CV 0.90x)\n* Public testset contains only around 400~ samples and private testset is no that better (600~ samples)\n\nThat means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition and at last moment I choose my early submissions of this competition (with much worse CV/LB than my public one) and got private LB 0.88x to act like that I didn't work on this game 🙌 \n\nAnd I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking, while didn't got any medal!\n\nCongratulations again and thank you for reading!",
    "940690": "Thank you for your sharings. Now I understand why you weren’t submitting new submissions. You and lafoss are the winners of this competition in my opinion. ",
    "940644": "**And I'm quite glad to see that my public kernel didn't shake up that much on public/private ranking** that's  the biggest success for you in this competition,it is about the \"strongest single public model of this competition\" and you had that and also you could keep that as private but you didn't,you shared your knowledge and that takes courage,thank you a ton for your awesome kernel,it helped a lot of people in this competition and i am sure a lot of researcher will get huge benefit from your awesome kernel\nthank you a lot @haqishen  san for sharing wisdom\nyour codes are poetry :) pleasure reading those :) ",
    "943955": "@haqishen Thank you for your notebook it really helped me out, with getting the idea of what was to be done",
    "941989": "Thank you very much @haqishen your notebook was another game changer in this competition. Thanks for sharing your knowledge and helping others to learn. 🙌 ",
    "941881": "Although you didn't win, but your contribution was great. Your model help many achieved the target that you couldn't , I think that's the win for you. \nThanks ",
    "941706": "I have learned a lot from your great notebook - many thanks!",
    "941247": "Thank you so much! For me, iafoss's tiling image is game changer 1 and your bce loss is game changer 2. 😄 ",
    "941032": "Thank you for your contribution and sharing. ",
    "941745": "&gt; That means we cannot trust LB score while even cannot trust CV score!\nSo I leave this competition\n\nGrandMaster Wisdom ^^",
    "941240": "Thank you @haqishen for your sharing here and also in Alaska2 on modifying EfficientNet architecture (had that ended sooner might have liked to try some thing with that info here!).  Your kernel did well with TTA variations and combinations, just did not pick any of them!!  But it is possible some in medals got there with your kernel.  \n\nYou and @iafoss deserve some special award/recognition for your contributions here! ",
    "943671": ""
  }
}