{
  "id": 133305,
  "title": "Can anyone guide me? This is my first competition",
  "url": "/competitions/bengaliai-cv19/discussion/133305",
  "author_name": "",
  "post_date": "2020-03-02T01:50:09.723075300Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I have trained my models with 4 stratified folds with data augmentation for ~150 epochs and CV score is around ~0.99 but LB is 0.95 only. Why is this gap so much? Am i missing anything serious? </p>",
  "messages": [
    {
      "id": "760977",
      "postDate": "03/02/2020 01:50:09",
      "content": "<p>I have trained my models with 4 stratified folds with data augmentation for ~150 epochs and CV score is around ~0.99 but LB is 0.95 only. Why is this gap so much? Am i missing anything serious? </p>",
      "rawMarkdown": "I have trained my models with 4 stratified folds with data augmentation for ~150 epochs and CV score is around ~0.99 but LB is 0.95 only. Why is this gap so much? Am i missing anything serious?",
      "votes": null
    },
    {
      "id": "761024",
      "postDate": "03/02/2020 03:31:07",
      "content": "<p>How are you calculating your validation score? Are you using <code>(2*root recall macro + vowel recall macro + consonant recall macro) /4.0</code>?</p>",
      "rawMarkdown": "How are you calculating your validation score? Are you using `(2*root recall macro + vowel recall macro + consonant recall macro) /4.0`?",
      "votes": null
    },
    {
      "id": "761080",
      "postDate": "03/02/2020 05:22:21",
      "content": "<p>No i just used the default categorical crossentropy 😅.\nDo i need to def a loss function and implement it? </p>",
      "rawMarkdown": "No i just used the default categorical crossentropy 😅.\nDo i need to def a loss function and implement it?",
      "votes": null
    },
    {
      "id": "762242",
      "postDate": "03/03/2020 11:05:17",
      "content": "<p>take one from sklearn - recall_score </p>",
      "rawMarkdown": "take one from sklearn - recall_score",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 761024,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "03/02/2020 03:31:07",
      "content": "<p>How are you calculating your validation score? Are you using <code>(2*root recall macro + vowel recall macro + consonant recall macro) /4.0</code>?</p>",
      "votes": null,
      "replies": [
        {
          "id": 761080,
          "author_name": "chittalpatel",
          "author_url": "",
          "post_date": "03/02/2020 05:22:21",
          "content": "<p>No i just used the default categorical crossentropy 😅.\nDo i need to def a loss function and implement it? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 762242,
          "author_name": "kupchanski",
          "author_url": "",
          "post_date": "03/03/2020 11:05:17",
          "content": "<p>take one from sklearn - recall_score </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "760977": "I have trained my models with 4 stratified folds with data augmentation for ~150 epochs and CV score is around ~0.99 but LB is 0.95 only. Why is this gap so much? Am i missing anything serious?",
    "761024": "How are you calculating your validation score? Are you using `(2*root recall macro + vowel recall macro + consonant recall macro) /4.0`?",
    "761080": "No i just used the default categorical crossentropy 😅.\nDo i need to def a loss function and implement it?",
    "762242": "take one from sklearn - recall_score"
  },
  "source": "meta"
}