{
  "id": 207739,
  "title": "What validation split are you using?",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/207739",
  "author_name": "",
  "post_date": "2020-12-31T04:28:22.736324Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Just want to hear everyone's opinions and what conclusions you have drawn from your experiments so far! I've tried only 0.2 and 0.1 and I've found that 0.1 gave me better results, but is more prone to overfitting. Guess that means I should implement heavier regularization!</p>",
  "messages": [
    {
      "id": "1133274",
      "postDate": "12/31/2020 04:28:22",
      "content": "<p>Just want to hear everyone's opinions and what conclusions you have drawn from your experiments so far! I've tried only 0.2 and 0.1 and I've found that 0.1 gave me better results, but is more prone to overfitting. Guess that means I should implement heavier regularization!</p>",
      "rawMarkdown": "Just want to hear everyone's opinions and what conclusions you have drawn from your experiments so far! I've tried only 0.2 and 0.1 and I've found that 0.1 gave me better results, but is more prone to overfitting. Guess that means I should implement heavier regularization!",
      "votes": null
    },
    {
      "id": "1133597",
      "postDate": "12/31/2020 11:15:11",
      "content": "<p>It is more productive to use cross validation with 5 folds, in the case of this particular competition, stratified by label cross validation</p>",
      "rawMarkdown": "It is more productive to use cross validation with 5 folds, in the case of this particular competition, stratified by label cross validation",
      "votes": null
    },
    {
      "id": "1134310",
      "postDate": "01/01/2021 06:07:52",
      "content": "<p>Ah, so in that case it would be a 0.2 stratified cv split. Got it, thanks!</p>",
      "rawMarkdown": "Ah, so in that case it would be a 0.2 stratified cv split. Got it, thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1133597,
      "author_name": "vladvdv",
      "author_url": "",
      "post_date": "12/31/2020 11:15:11",
      "content": "<p>It is more productive to use cross validation with 5 folds, in the case of this particular competition, stratified by label cross validation</p>",
      "votes": null,
      "replies": [
        {
          "id": 1134310,
          "author_name": "junyingsg",
          "author_url": "",
          "post_date": "01/01/2021 06:07:52",
          "content": "<p>Ah, so in that case it would be a 0.2 stratified cv split. Got it, thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1133274": "Just want to hear everyone's opinions and what conclusions you have drawn from your experiments so far! I've tried only 0.2 and 0.1 and I've found that 0.1 gave me better results, but is more prone to overfitting. Guess that means I should implement heavier regularization!",
    "1133597": "It is more productive to use cross validation with 5 folds, in the case of this particular competition, stratified by label cross validation",
    "1134310": "Ah, so in that case it would be a 0.2 stratified cv split. Got it, thanks!"
  },
  "source": "meta"
}