{
  "id": 220594,
  "title": "Lessons Learned!",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/220594",
  "author_name": "",
  "post_date": "2021-02-19T00:18:06.173735Z",
  "votes": 12,
  "comment_count": 3,
  "views": 0,
  "content": "<p>This was my first kaggle competition and probably the biggest lesson I have learnt is to use Cross Validation. In this competition I did not use cross validation because I was scared of having to spend 5x as long experimenting with new ideas.</p>\n<p>During the competition, I was at spot 260 on the public leaderboard. Towards the end of the competition I was beginning to have difficulties evaluating my models because there was so much inconsistency between my single split local score and the public leaderboard. </p>\n<p>At the end, I dropped down to spot 665 in the final rankings because I was not able to robustly assess my models. Lesson learned: </p>\n<p>Don't cheap out on Cross Validation my friends!</p>",
  "messages": [
    {
      "id": "1209536",
      "postDate": "02/19/2021 00:18:06",
      "content": "<p>This was my first kaggle competition and probably the biggest lesson I have learnt is to use Cross Validation. In this competition I did not use cross validation because I was scared of having to spend 5x as long experimenting with new ideas.</p>\n<p>During the competition, I was at spot 260 on the public leaderboard. Towards the end of the competition I was beginning to have difficulties evaluating my models because there was so much inconsistency between my single split local score and the public leaderboard. </p>\n<p>At the end, I dropped down to spot 665 in the final rankings because I was not able to robustly assess my models. Lesson learned: </p>\n<p>Don't cheap out on Cross Validation my friends!</p>",
      "rawMarkdown": "This was my first kaggle competition and probably the biggest lesson I have learnt is to use Cross Validation. In this competition I did not use cross validation because I was scared of having to spend 5x as long experimenting with new ideas.\n\nDuring the competition, I was at spot 260 on the public leaderboard. Towards the end of the competition I was beginning to have difficulties evaluating my models because there was so much inconsistency between my single split local score and the public leaderboard. \n\nAt the end, I dropped down to spot 665 in the final rankings because I was not able to robustly assess my models. Lesson learned: \n\nDon't cheap out on Cross Validation my friends!",
      "votes": null
    },
    {
      "id": "1209550",
      "postDate": "02/19/2021 00:27:05",
      "content": "<p>Cross validation, ensemble, tta… and shake up. I feel that novice has a long way to catch up.</p>",
      "rawMarkdown": "Cross validation, ensemble, tta... and shake up. I feel that novice has a long way to catch up.",
      "votes": null
    },
    {
      "id": "1209642",
      "postDate": "02/19/2021 01:42:50",
      "content": "<p>Same with me.</p>",
      "rawMarkdown": "Same with me.",
      "votes": null
    },
    {
      "id": "1210108",
      "postDate": "02/19/2021 07:53:52",
      "content": "<p>yeah, before this competition I have never use CV in computer vision, don't know about ensemble and tta. pretty sad but lesson learned. hope get better tomorrow</p>",
      "rawMarkdown": "yeah, before this competition I have never use CV in computer vision, don't know about ensemble and tta. pretty sad but lesson learned. hope get better tomorrow",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1209550,
      "author_name": "komakizzz",
      "author_url": "",
      "post_date": "02/19/2021 00:27:05",
      "content": "<p>Cross validation, ensemble, tta… and shake up. I feel that novice has a long way to catch up.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1210108,
          "author_name": "houoinkyoma",
          "author_url": "",
          "post_date": "02/19/2021 07:53:52",
          "content": "<p>yeah, before this competition I have never use CV in computer vision, don't know about ensemble and tta. pretty sad but lesson learned. hope get better tomorrow</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1209642,
      "author_name": "electro",
      "author_url": "",
      "post_date": "02/19/2021 01:42:50",
      "content": "<p>Same with me.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1209536": "This was my first kaggle competition and probably the biggest lesson I have learnt is to use Cross Validation. In this competition I did not use cross validation because I was scared of having to spend 5x as long experimenting with new ideas.\n\nDuring the competition, I was at spot 260 on the public leaderboard. Towards the end of the competition I was beginning to have difficulties evaluating my models because there was so much inconsistency between my single split local score and the public leaderboard. \n\nAt the end, I dropped down to spot 665 in the final rankings because I was not able to robustly assess my models. Lesson learned: \n\nDon't cheap out on Cross Validation my friends!",
    "1209550": "Cross validation, ensemble, tta... and shake up. I feel that novice has a long way to catch up.",
    "1209642": "Same with me.",
    "1210108": "yeah, before this competition I have never use CV in computer vision, don't know about ensemble and tta. pretty sad but lesson learned. hope get better tomorrow"
  },
  "source": "meta"
}