{
  "id": 206797,
  "title": "Some noobie questions",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/206797",
  "author_name": "",
  "post_date": "2020-12-26T15:15:03.430456800Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello everyone! I'm kinda new to ml competitions. I have a couple of perhaps kinda silly questions:</p>\n<ol>\n<li>What does CV and LB stand for? </li>\n<li>I tried to submit a pre-calculated csv (commented out everything except a cp command) but I got an error. Is it because it had nothing for the private test data? </li>\n</ol>",
  "messages": [
    {
      "id": "1127480",
      "postDate": "12/26/2020 15:15:03",
      "content": "<p>Hello everyone! I'm kinda new to ml competitions. I have a couple of perhaps kinda silly questions:</p>\n<ol>\n<li>What does CV and LB stand for? </li>\n<li>I tried to submit a pre-calculated csv (commented out everything except a cp command) but I got an error. Is it because it had nothing for the private test data? </li>\n</ol>",
      "rawMarkdown": "Hello everyone! I'm kinda new to ml competitions. I have a couple of perhaps kinda silly questions:\n1. What does CV and LB stand for? \n2. I tried to submit a pre-calculated csv (commented out everything except a cp command) but I got an error. Is it because it had nothing for the private test data?",
      "votes": null
    },
    {
      "id": "1127517",
      "postDate": "12/26/2020 15:51:54",
      "content": "<p>CV = Cross Validation. Local results on your validation data. Typically you run your training algorithm 5 times, using 80% of the data for training and the other 20% for validation. Average the 5 validation results to get your CV. Sometimes CV will be used to must mean one run, rather than an average of all 5.</p>\n<p>LB - Leaderboard Results - official results from the Kaggle Leaderboard</p>\n<p>If you over-fit your algorithm to the training data, your CV will be better than your LB. The goal is to have them close. And, in general, always trust your CV over the LB.</p>\n<p>You cannot pre-calculate the submission.csv. It must be run in the committed notebook against the hidden test data.</p>\n<p>-Rich</p>",
      "rawMarkdown": "CV = Cross Validation. Local results on your validation data. Typically you run your training algorithm 5 times, using 80% of the data for training and the other 20% for validation. Average the 5 validation results to get your CV. Sometimes CV will be used to must mean one run, rather than an average of all 5.\n\nLB - Leaderboard Results - official results from the Kaggle Leaderboard\n\nIf you over-fit your algorithm to the training data, your CV will be better than your LB. The goal is to have them close. And, in general, always trust your CV over the LB.\n\nYou cannot pre-calculate the submission.csv. It must be run in the committed notebook against the hidden test data.\n\n-Rich",
      "votes": null
    },
    {
      "id": "1128313",
      "postDate": "12/27/2020 10:18:28",
      "content": "<p>Thanks for the answer!<br>\nAlthough, it seems that I can actually pre-calculate it, there are notebooks that help with that.<br>\nAlthough that's just for convenience for testing, not as a final submission.</p>",
      "rawMarkdown": "Thanks for the answer!\nAlthough, it seems that I can actually pre-calculate it, there are notebooks that help with that.\nAlthough that's just for convenience for testing, not as a final submission.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1127517,
      "author_name": "richardepstein",
      "author_url": "",
      "post_date": "12/26/2020 15:51:54",
      "content": "<p>CV = Cross Validation. Local results on your validation data. Typically you run your training algorithm 5 times, using 80% of the data for training and the other 20% for validation. Average the 5 validation results to get your CV. Sometimes CV will be used to must mean one run, rather than an average of all 5.</p>\n<p>LB - Leaderboard Results - official results from the Kaggle Leaderboard</p>\n<p>If you over-fit your algorithm to the training data, your CV will be better than your LB. The goal is to have them close. And, in general, always trust your CV over the LB.</p>\n<p>You cannot pre-calculate the submission.csv. It must be run in the committed notebook against the hidden test data.</p>\n<p>-Rich</p>",
      "votes": null,
      "replies": [
        {
          "id": 1128313,
          "author_name": "antoni4040",
          "author_url": "",
          "post_date": "12/27/2020 10:18:28",
          "content": "<p>Thanks for the answer!<br>\nAlthough, it seems that I can actually pre-calculate it, there are notebooks that help with that.<br>\nAlthough that's just for convenience for testing, not as a final submission.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1127480": "Hello everyone! I'm kinda new to ml competitions. I have a couple of perhaps kinda silly questions:\n1. What does CV and LB stand for? \n2. I tried to submit a pre-calculated csv (commented out everything except a cp command) but I got an error. Is it because it had nothing for the private test data?",
    "1127517": "CV = Cross Validation. Local results on your validation data. Typically you run your training algorithm 5 times, using 80% of the data for training and the other 20% for validation. Average the 5 validation results to get your CV. Sometimes CV will be used to must mean one run, rather than an average of all 5.\n\nLB - Leaderboard Results - official results from the Kaggle Leaderboard\n\nIf you over-fit your algorithm to the training data, your CV will be better than your LB. The goal is to have them close. And, in general, always trust your CV over the LB.\n\nYou cannot pre-calculate the submission.csv. It must be run in the committed notebook against the hidden test data.\n\n-Rich",
    "1128313": "Thanks for the answer!\nAlthough, it seems that I can actually pre-calculate it, there are notebooks that help with that.\nAlthough that's just for convenience for testing, not as a final submission."
  },
  "source": "meta"
}