{
  "id": 177241,
  "title": "Is F1-score too low for Initial epochs for everyone?",
  "url": "/competitions/birdsong-recognition/discussion/177241",
  "author_name": "",
  "post_date": "2020-08-25T08:33:58.118791600Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello everyone , <br>\nI am really sorry If this question has already been asked and answered , I will really appreciate if someone directs me there and I will delete this thread </p>\n<p>As I am training my baseline , I observe my f1 score to be very low for first two epochs and in fact it decreases with decrease in loss , it's really wierd , however I saw some public kernels experiencing the same , so I just wanted to confirm if that's the case with everyone or I might be doing something wrong</p>\n<p>Thanks</p>",
  "messages": [
    {
      "id": "984715",
      "postDate": "08/25/2020 08:33:58",
      "content": "<p>Hello everyone , <br>\nI am really sorry If this question has already been asked and answered , I will really appreciate if someone directs me there and I will delete this thread </p>\n<p>As I am training my baseline , I observe my f1 score to be very low for first two epochs and in fact it decreases with decrease in loss , it's really wierd , however I saw some public kernels experiencing the same , so I just wanted to confirm if that's the case with everyone or I might be doing something wrong</p>\n<p>Thanks</p>",
      "rawMarkdown": "Hello everyone , \nI am really sorry If this question has already been asked and answered , I will really appreciate if someone directs me there and I will delete this thread \n\nAs I am training my baseline , I observe my f1 score to be very low for first two epochs and in fact it decreases with decrease in loss , it's really wierd , however I saw some public kernels experiencing the same , so I just wanted to confirm if that's the case with everyone or I might be doing something wrong\n\nThanks",
      "votes": null
    },
    {
      "id": "984772",
      "postDate": "08/25/2020 09:24:11",
      "content": "<p>I have actually observed a similar trend as well. It does however start to go back up again after first few epochs (around 5-10) and then occasionally drops as training progresses.</p>",
      "rawMarkdown": "I have actually observed a similar trend as well. It does however start to go back up again after first few epochs (around 5-10) and then occasionally drops as training progresses.",
      "votes": null
    },
    {
      "id": "984783",
      "postDate": "08/25/2020 09:32:37",
      "content": "<p>Thanks <a href=\"https://www.kaggle.com/syborg\" target=\"_blank\">@syborg</a> for the reply <br>\nI guess the problem is tough to learn and hence it takes some epochs for the model to get better</p>",
      "rawMarkdown": "Thanks @syborg for the reply \nI guess the problem is tough to learn and hence it takes some epochs for the model to get better",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 984772,
      "author_name": "syborg",
      "author_url": "",
      "post_date": "08/25/2020 09:24:11",
      "content": "<p>I have actually observed a similar trend as well. It does however start to go back up again after first few epochs (around 5-10) and then occasionally drops as training progresses.</p>",
      "votes": null,
      "replies": [
        {
          "id": 984783,
          "author_name": "tanulsingh077",
          "author_url": "",
          "post_date": "08/25/2020 09:32:37",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/syborg\" target=\"_blank\">@syborg</a> for the reply <br>\nI guess the problem is tough to learn and hence it takes some epochs for the model to get better</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "984715": "Hello everyone , \nI am really sorry If this question has already been asked and answered , I will really appreciate if someone directs me there and I will delete this thread \n\nAs I am training my baseline , I observe my f1 score to be very low for first two epochs and in fact it decreases with decrease in loss , it's really wierd , however I saw some public kernels experiencing the same , so I just wanted to confirm if that's the case with everyone or I might be doing something wrong\n\nThanks",
    "984772": "I have actually observed a similar trend as well. It does however start to go back up again after first few epochs (around 5-10) and then occasionally drops as training progresses.",
    "984783": "Thanks @syborg for the reply \nI guess the problem is tough to learn and hence it takes some epochs for the model to get better"
  },
  "source": "meta"
}