{
  "id": 75496,
  "title": "Can I use the Public Leaderboard Score in my training ?",
  "url": "/competitions/humpback-whale-identification/discussion/75496",
  "author_name": "",
  "post_date": "2018-12-22T11:35:56.589812700Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I have submited several submissions, and each submission have a public score. I want to know if I can use this score as a parameter in my next training ?\nSomebody told me that the score is the result of the test data, if I use the score in my next training, It's equivalent to use the \"information from test set\". This is not permitted.</p>",
  "messages": [
    {
      "id": "443767",
      "postDate": "12/22/2018 11:35:56",
      "content": "<p>I have submited several submissions, and each submission have a public score. I want to know if I can use this score as a parameter in my next training ?\nSomebody told me that the score is the result of the test data, if I use the score in my next training, It's equivalent to use the \"information from test set\". This is not permitted.</p>",
      "rawMarkdown": "I have submited several submissions, and each submission have a public score. I want to know if I can use this score as a parameter in my next training ?\nSomebody told me that the score is the result of the test data, if I use the score in my next training, It's equivalent to use the \"information from test set\". This is not permitted.",
      "votes": null
    },
    {
      "id": "443815",
      "postDate": "12/22/2018 13:43:06",
      "content": "<p>You use it as a guidance but probably should not use it explicitly as a source of data. </p>",
      "rawMarkdown": "You use it as a guidance but probably should not use it explicitly as a source of data.",
      "votes": null
    },
    {
      "id": "443843",
      "postDate": "12/22/2018 14:56:41",
      "content": "<p>It would not be against the rules, but it could be unwise as you might find yourself overfitting to the public LB.</p>",
      "rawMarkdown": "It would not be against the rules, but it could be unwise as you might find yourself overfitting to the public LB.",
      "votes": null
    },
    {
      "id": "444507",
      "postDate": "12/24/2018 06:39:16",
      "content": "<p>Yes you can. But as William said, public LB is not trustable, especially when your score is convergent. An alternative way is to make use of your local score to substitute public score. By the way, according to your high score, focusing on the overfitting things may be more important : )</p>",
      "rawMarkdown": "Yes you can. But as William said, public LB is not trustable, especially when your score is convergent. An alternative way is to make use of your local score to substitute public score. By the way, according to your high score, focusing on the overfitting things may be more important : )",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 443815,
      "author_name": "zoujie",
      "author_url": "",
      "post_date": "12/22/2018 13:43:06",
      "content": "<p>You use it as a guidance but probably should not use it explicitly as a source of data. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 443843,
      "author_name": "hortonhearsafoo",
      "author_url": "",
      "post_date": "12/22/2018 14:56:41",
      "content": "<p>It would not be against the rules, but it could be unwise as you might find yourself overfitting to the public LB.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 444507,
      "author_name": "yiheng",
      "author_url": "",
      "post_date": "12/24/2018 06:39:16",
      "content": "<p>Yes you can. But as William said, public LB is not trustable, especially when your score is convergent. An alternative way is to make use of your local score to substitute public score. By the way, according to your high score, focusing on the overfitting things may be more important : )</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "443767": "I have submited several submissions, and each submission have a public score. I want to know if I can use this score as a parameter in my next training ?\nSomebody told me that the score is the result of the test data, if I use the score in my next training, It's equivalent to use the \"information from test set\". This is not permitted.",
    "443815": "You use it as a guidance but probably should not use it explicitly as a source of data.",
    "443843": "It would not be against the rules, but it could be unwise as you might find yourself overfitting to the public LB.",
    "444507": "Yes you can. But as William said, public LB is not trustable, especially when your score is convergent. An alternative way is to make use of your local score to substitute public score. By the way, according to your high score, focusing on the overfitting things may be more important : )"
  },
  "source": "meta"
}