{
  "id": 365191,
  "title": "How much we can achieve in principle (after the end of the competition) 0.817, 0.818 ... ? ",
  "url": "/competitions/open-problems-multimodal/discussion/365191",
  "author_name": "",
  "post_date": "2022-11-10T07:34:25.880686900Z",
  "votes": 9,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Senkin13 and \"Tmp\" broke the limit 0.816 yesterday ! The greatest congratulations to them ! </p>\n<p>Still, my feeling is that the challenge is difficult also technically - and so may be just having more time/computational resources/ideas one may go even further. </p>\n<p>The question is how far ? 0.817 ? or may be even beyond ?  </p>\n<p>Hopefully we can at least make some judgement after the competition end - and some top solutions might become public.<br>\nThus analyzing them and probably blending them -  one may hope to get some improvement over the top1 solution, I would not expect big improvement, but cannot totally exclude the chance to achieve say 0.817.<br>\nSo any your  remarks/comments/ideas - welcome ! </p>",
  "messages": [
    {
      "id": "2024030",
      "postDate": "11/10/2022 07:34:25",
      "content": "<p>Senkin13 and \"Tmp\" broke the limit 0.816 yesterday ! The greatest congratulations to them ! </p>\n<p>Still, my feeling is that the challenge is difficult also technically - and so may be just having more time/computational resources/ideas one may go even further. </p>\n<p>The question is how far ? 0.817 ? or may be even beyond ?  </p>\n<p>Hopefully we can at least make some judgement after the competition end - and some top solutions might become public.<br>\nThus analyzing them and probably blending them -  one may hope to get some improvement over the top1 solution, I would not expect big improvement, but cannot totally exclude the chance to achieve say 0.817.<br>\nSo any your  remarks/comments/ideas - welcome ! </p>",
      "rawMarkdown": "Senkin13 and \"Tmp\" broke the limit 0.816 yesterday ! The greatest congratulations to them ! \n\nStill, my feeling is that the challenge is difficult also technically - and so may be just having more time/computational resources/ideas one may go even further. \n\nThe question is how far ? 0.817 ? or may be even beyond ?  \n\nHopefully we can at least make some judgement after the competition end - and some top solutions might become public.\nThus analyzing them and probably blending them -  one may hope to get some improvement over the top1 solution, I would not expect big improvement, but cannot totally exclude the chance to achieve say 0.817.\nSo any your  remarks/comments/ideas - welcome !",
      "votes": null
    },
    {
      "id": "2024634",
      "postDate": "11/10/2022 16:38:41",
      "content": "<p>Interesting question, but of course the real issue is what happens on the private LB. How much shakeup will we see? Are the top solutions on the public LB badly overfitting, or will they also perform well on the private test set, which differs significantly from the public.</p>",
      "rawMarkdown": "Interesting question, but of course the real issue is what happens on the private LB. How much shakeup will we see? Are the top solutions on the public LB badly overfitting, or will they also perform well on the private test set, which differs significantly from the public.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2024634,
      "author_name": "dslate",
      "author_url": "",
      "post_date": "11/10/2022 16:38:41",
      "content": "<p>Interesting question, but of course the real issue is what happens on the private LB. How much shakeup will we see? Are the top solutions on the public LB badly overfitting, or will they also perform well on the private test set, which differs significantly from the public.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2024030": "Senkin13 and \"Tmp\" broke the limit 0.816 yesterday ! The greatest congratulations to them ! \n\nStill, my feeling is that the challenge is difficult also technically - and so may be just having more time/computational resources/ideas one may go even further. \n\nThe question is how far ? 0.817 ? or may be even beyond ?  \n\nHopefully we can at least make some judgement after the competition end - and some top solutions might become public.\nThus analyzing them and probably blending them -  one may hope to get some improvement over the top1 solution, I would not expect big improvement, but cannot totally exclude the chance to achieve say 0.817.\nSo any your  remarks/comments/ideas - welcome !",
    "2024634": "Interesting question, but of course the real issue is what happens on the private LB. How much shakeup will we see? Are the top solutions on the public LB badly overfitting, or will they also perform well on the private test set, which differs significantly from the public."
  },
  "source": "meta"
}