{
  "id": 304155,
  "title": "Public LB and Private LB",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/304155",
  "author_name": "aquaright",
  "post_date": "2022-01-31T02:34:52.962000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I've been reading some of the posts by <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> and I just wanted to clarify a couple of things. </p>\n<p>This is an excerpt from the lb page: <code>This leaderboard is calculated with approximately 25% of the test data. The final results will be based on the other 75%, so the final standings may be different.</code></p>\n<p>Based on the wording, it sounds like public lb and private lb are from the same distribution, but <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> mentioned in one of his posts that f2 could drastically go down depending on the shakeup between difference in empty images between public lb and private lb. </p>\n<p>I think there's definitely a difference in train since many report a gap between public lb and cv. </p>\n<p>I'm currently just trying to gauge the 'img_size * 2.5' method.</p>",
  "messages": [
    {
      "id": 1669860,
      "postDate": "2022-01-31T02:34:52.963Z",
      "content": "<p>I've been reading some of the posts by <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> and I just wanted to clarify a couple of things. </p>\n<p>This is an excerpt from the lb page: <code>This leaderboard is calculated with approximately 25% of the test data. The final results will be based on the other 75%, so the final standings may be different.</code></p>\n<p>Based on the wording, it sounds like public lb and private lb are from the same distribution, but <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> mentioned in one of his posts that f2 could drastically go down depending on the shakeup between difference in empty images between public lb and private lb. </p>\n<p>I think there's definitely a difference in train since many report a gap between public lb and cv. </p>\n<p>I'm currently just trying to gauge the 'img_size * 2.5' method.</p>",
      "rawMarkdown": "I've been reading some of the posts by @hengck23 and I just wanted to clarify a couple of things. \n\nThis is an excerpt from the lb page: `This leaderboard is calculated with approximately 25% of the test data. The final results will be based on the other 75%, so the final standings may be different.`\n\nBased on the wording, it sounds like public lb and private lb are from the same distribution, but @hengck23 mentioned in one of his posts that f2 could drastically go down depending on the shakeup between difference in empty images between public lb and private lb. \n\nI think there's definitely a difference in train since many report a gap between public lb and cv. \n\nI'm currently just trying to gauge the 'img_size * 2.5' method.",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1669860": "I've been reading some of the posts by @hengck23 and I just wanted to clarify a couple of things. \n\nThis is an excerpt from the lb page: `This leaderboard is calculated with approximately 25% of the test data. The final results will be based on the other 75%, so the final standings may be different.`\n\nBased on the wording, it sounds like public lb and private lb are from the same distribution, but @hengck23 mentioned in one of his posts that f2 could drastically go down depending on the shakeup between difference in empty images between public lb and private lb. \n\nI think there's definitely a difference in train since many report a gap between public lb and cv. \n\nI'm currently just trying to gauge the 'img_size * 2.5' method."
  }
}