{
  "id": 174483,
  "title": "Recommendations on Final submissions --Reduce overfitting??",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/174483",
  "author_name": "Kamal Das",
  "post_date": "2020-08-13T17:16:29.405000",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>HI All,</p>\n<p>checking in what you all do to reduce over fitting, and your methods to select the best final models, as the public and private scores often vary</p>\n<p>I am getting better scores in my validation in some models that in the public leader board; and lower in others… do you trust your validation score or the LB? </p>\n<p>Recommendations on how to select Final submissions are welcome… been active in competitions only for a month or so, appreciate your guidance and inputs! </p>\n<p>Thanks in advance!</p>",
  "messages": [
    {
      "id": 971334,
      "postDate": "2020-08-15T12:05:27.350Z",
      "content": "<p>Thanks, <a href=\"/kmldas\">@kmldas</a> for a great piece of advice.2 more days to go will try to implement.</p>",
      "rawMarkdown": "Thanks, @kmldas for a great piece of advice.2 more days to go will try to implement.",
      "votes": 1,
      "replies": [
        {
          "id": 971992,
          "postDate": "2020-08-16T06:30:43.750Z",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/vin1234\" target=\"_blank\">@vin1234</a> looking for advice myself</p>\n<p>This competition allows for 3 selections, besides the highest ranked, I am opting for 2 others which are not in my top 15 (in my Public LB scores), as the model was different and my CV is better…. hoping it works! Will know in a couple of days! </p>\n<p>Hoping CV is better indicator than Public LB</p>",
          "rawMarkdown": "Hi @vin1234 looking for advice myself\n\nThis competition allows for 3 selections, besides the highest ranked, I am opting for 2 others which are not in my top 15 (in my Public LB scores), as the model was different and my CV is better.... hoping it works! Will know in a couple of days! \n\nHoping CV is better indicator than Public LB\n",
          "votes": 1
        },
        {
          "id": 972038,
          "postDate": "2020-08-16T07:28:28.763Z",
          "content": "<p>Fingers crossed… and hope for the best…</p>",
          "rawMarkdown": "Fingers crossed... and hope for the best...",
          "votes": 1
        }
      ]
    },
    {
      "id": 969432,
      "postDate": "2020-08-13T17:16:29.407Z",
      "content": "<p>HI All,</p>\n<p>checking in what you all do to reduce over fitting, and your methods to select the best final models, as the public and private scores often vary</p>\n<p>I am getting better scores in my validation in some models that in the public leader board; and lower in others… do you trust your validation score or the LB? </p>\n<p>Recommendations on how to select Final submissions are welcome… been active in competitions only for a month or so, appreciate your guidance and inputs! </p>\n<p>Thanks in advance!</p>",
      "rawMarkdown": "HI All,\n\nchecking in what you all do to reduce over fitting, and your methods to select the best final models, as the public and private scores often vary\n\nI am getting better scores in my validation in some models that in the public leader board; and lower in others... do you trust your validation score or the LB? \n\nRecommendations on how to select Final submissions are welcome... been active in competitions only for a month or so, appreciate your guidance and inputs! \n\nThanks in advance!",
      "votes": 1
    },
    {
      "id": 974612,
      "postDate": "2020-08-18T02:00:47.950Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 971334,
      "author_name": "VinayVikram",
      "author_url": "",
      "post_date": "2020-08-15T12:05:27.350000",
      "content": "<p>Thanks, <a href=\"/kmldas\">@kmldas</a> for a great piece of advice.2 more days to go will try to implement.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 971992,
          "author_name": "Kamal Das",
          "author_url": "",
          "post_date": "2020-08-16T06:30:43.750000",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/vin1234\" target=\"_blank\">@vin1234</a> looking for advice myself</p>\n<p>This competition allows for 3 selections, besides the highest ranked, I am opting for 2 others which are not in my top 15 (in my Public LB scores), as the model was different and my CV is better…. hoping it works! Will know in a couple of days! </p>\n<p>Hoping CV is better indicator than Public LB</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 972038,
          "author_name": "VinayVikram",
          "author_url": "",
          "post_date": "2020-08-16T07:28:28.763000",
          "content": "<p>Fingers crossed… and hope for the best…</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 974612,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-08-18T02:00:47.950000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "971334": "Thanks, @kmldas for a great piece of advice.2 more days to go will try to implement.",
    "969432": "HI All,\n\nchecking in what you all do to reduce over fitting, and your methods to select the best final models, as the public and private scores often vary\n\nI am getting better scores in my validation in some models that in the public leader board; and lower in others... do you trust your validation score or the LB? \n\nRecommendations on how to select Final submissions are welcome... been active in competitions only for a month or so, appreciate your guidance and inputs! \n\nThanks in advance!",
    "974612": ""
  }
}