{
  "id": 115001,
  "title": "Is it possible to reach 0.70+?",
  "url": "/competitions/understanding_cloud_organization/discussion/115001",
  "author_name": "",
  "post_date": "2019-10-30T16:28:10.558777Z",
  "votes": 3,
  "comment_count": 9,
  "views": 0,
  "content": "<p>We still have around three weeks to go. How high do you think the top LB score will be? </p>",
  "messages": [
    {
      "id": "661719",
      "postDate": "10/30/2019 16:28:10",
      "content": "<p>We still have around three weeks to go. How high do you think the top LB score will be? </p>",
      "rawMarkdown": "We still have around three weeks to go. How high do you think the top LB score will be?",
      "votes": null
    },
    {
      "id": "661741",
      "postDate": "10/30/2019 16:51:17",
      "content": "<p>Kagglers will never say something is impossible. It's possible either by overfitting the data or using a genuine approach.   My cross-validation says it's ~0.60. Though validation schema is different for different people.</p>",
      "rawMarkdown": "Kagglers will never say something is impossible. It's possible either by overfitting the data or using a genuine approach.   My cross-validation says it's ~0.60. Though validation schema is different for different people.",
      "votes": null
    },
    {
      "id": "661745",
      "postDate": "10/30/2019 16:56:13",
      "content": "<p>Well... I don't think so. Based on LB, score seems to be converging to 0.680, so that's my bet.\nBut it's always possible that someone find something new and get a good boost. I just hope we don't have a big shake up in LB. I look forward to reading winners solutions, have learned a lot in this competition</p>",
      "rawMarkdown": "Well... I don't think so. Based on LB, score seems to be converging to 0.680, so that's my bet.\nBut it's always possible that someone find something new and get a good boost. I just hope we don't have a big shake up in LB. I look forward to reading winners solutions, have learned a lot in this competition",
      "votes": null
    },
    {
      "id": "661760",
      "postDate": "10/30/2019 17:16:09",
      "content": "<p>i can see huge shakeup is coming</p>",
      "rawMarkdown": "i can see huge shakeup is coming",
      "votes": null
    },
    {
      "id": "661768",
      "postDate": "10/30/2019 17:25:07",
      "content": "<p>I think the shakeup will be slighter than Steel because there is no private test set in this competition. But there is still a risk on overfitting the 25% of test data...</p>",
      "rawMarkdown": "I think the shakeup will be slighter than Steel because there is no private test set in this competition. But there is still a risk on overfitting the 25% of test data...",
      "votes": null
    },
    {
      "id": "661808",
      "postDate": "10/30/2019 18:30:47",
      "content": "<p>I don't think 0.70+ is possible because of label noise, I would guess no higher than 0.680 for public LB and probably 0.670 for the private LB</p>",
      "rawMarkdown": "I don't think 0.70+ is possible because of label noise, I would guess no higher than 0.680 for public LB and probably 0.670 for the private LB",
      "votes": null
    },
    {
      "id": "662029",
      "postDate": "10/31/2019 02:17:51",
      "content": "<p>Is the noise of the label very serious? I don’t know some specific features of different clouds after observing some images.</p>",
      "rawMarkdown": "Is the noise of the label very serious? I don’t know some specific features of different clouds after observing some images.",
      "votes": null
    },
    {
      "id": "662051",
      "postDate": "10/31/2019 03:23:27",
      "content": "<p>Well, friend ,  I  gotta  to  say  everything is possible , after  two  years , I  think   everyone  could  reach that   ,   there  will  be   baseline  model   called   SuperMan_model.py   published  by  Google    and   get  LB 0.80  very   easily</p>",
      "rawMarkdown": "Well, friend ,  I  gotta  to  say  everything is possible , after  two  years , I  think   everyone  could  reach that   ,   there  will  be   baseline  model   called   SuperMan_model.py   published  by  Google    and   get  LB 0.80  very   easily",
      "votes": null
    },
    {
      "id": "662098",
      "postDate": "10/31/2019 04:40:31",
      "content": "<p>Well, I am waiting for EfficientSuperNetB100😉</p>",
      "rawMarkdown": "Well, I am waiting for EfficientSuperNetB100😉",
      "votes": null
    },
    {
      "id": "664631",
      "postDate": "11/04/2019 02:00:08",
      "content": "<p>For me, a score of 0.7 is hopeless, but a score of 0.675 is still hopeful, only for 25% of the test data, and the last 75% of the test data is unknown.</p>",
      "rawMarkdown": "For me, a score of 0.7 is hopeless, but a score of 0.675 is still hopeful, only for 25% of the test data, and the last 75% of the test data is unknown.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 661741,
      "author_name": "sgalib",
      "author_url": "",
      "post_date": "10/30/2019 16:51:17",
      "content": "<p>Kagglers will never say something is impossible. It's possible either by overfitting the data or using a genuine approach.   My cross-validation says it's ~0.60. Though validation schema is different for different people.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 661745,
      "author_name": "igormunizims",
      "author_url": "",
      "post_date": "10/30/2019 16:56:13",
      "content": "<p>Well... I don't think so. Based on LB, score seems to be converging to 0.680, so that's my bet.\nBut it's always possible that someone find something new and get a good boost. I just hope we don't have a big shake up in LB. I look forward to reading winners solutions, have learned a lot in this competition</p>",
      "votes": null,
      "replies": [
        {
          "id": 661760,
          "author_name": "mobassir",
          "author_url": "",
          "post_date": "10/30/2019 17:16:09",
          "content": "<p>i can see huge shakeup is coming</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 661768,
          "author_name": "gogo827jz",
          "author_url": "",
          "post_date": "10/30/2019 17:25:07",
          "content": "<p>I think the shakeup will be slighter than Steel because there is no private test set in this competition. But there is still a risk on overfitting the 25% of test data...</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 661808,
      "author_name": "mnpinto",
      "author_url": "",
      "post_date": "10/30/2019 18:30:47",
      "content": "<p>I don't think 0.70+ is possible because of label noise, I would guess no higher than 0.680 for public LB and probably 0.670 for the private LB</p>",
      "votes": null,
      "replies": [
        {
          "id": 662029,
          "author_name": "shentaowang",
          "author_url": "",
          "post_date": "10/31/2019 02:17:51",
          "content": "<p>Is the noise of the label very serious? I don’t know some specific features of different clouds after observing some images.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 662051,
      "author_name": "xujingzhao",
      "author_url": "",
      "post_date": "10/31/2019 03:23:27",
      "content": "<p>Well, friend ,  I  gotta  to  say  everything is possible , after  two  years , I  think   everyone  could  reach that   ,   there  will  be   baseline  model   called   SuperMan_model.py   published  by  Google    and   get  LB 0.80  very   easily</p>",
      "votes": null,
      "replies": [
        {
          "id": 662098,
          "author_name": "gogo827jz",
          "author_url": "",
          "post_date": "10/31/2019 04:40:31",
          "content": "<p>Well, I am waiting for EfficientSuperNetB100😉</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 664631,
      "author_name": "zhangeng",
      "author_url": "",
      "post_date": "11/04/2019 02:00:08",
      "content": "<p>For me, a score of 0.7 is hopeless, but a score of 0.675 is still hopeful, only for 25% of the test data, and the last 75% of the test data is unknown.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "661719": "We still have around three weeks to go. How high do you think the top LB score will be?",
    "661741": "Kagglers will never say something is impossible. It's possible either by overfitting the data or using a genuine approach.   My cross-validation says it's ~0.60. Though validation schema is different for different people.",
    "661745": "Well... I don't think so. Based on LB, score seems to be converging to 0.680, so that's my bet.\nBut it's always possible that someone find something new and get a good boost. I just hope we don't have a big shake up in LB. I look forward to reading winners solutions, have learned a lot in this competition",
    "661760": "i can see huge shakeup is coming",
    "661768": "I think the shakeup will be slighter than Steel because there is no private test set in this competition. But there is still a risk on overfitting the 25% of test data...",
    "661808": "I don't think 0.70+ is possible because of label noise, I would guess no higher than 0.680 for public LB and probably 0.670 for the private LB",
    "662029": "Is the noise of the label very serious? I don’t know some specific features of different clouds after observing some images.",
    "662051": "Well, friend ,  I  gotta  to  say  everything is possible , after  two  years , I  think   everyone  could  reach that   ,   there  will  be   baseline  model   called   SuperMan_model.py   published  by  Google    and   get  LB 0.80  very   easily",
    "662098": "Well, I am waiting for EfficientSuperNetB100😉",
    "664631": "For me, a score of 0.7 is hopeless, but a score of 0.675 is still hopeful, only for 25% of the test data, and the last 75% of the test data is unknown."
  },
  "source": "meta"
}