{
  "id": 123753,
  "title": "Individual loss or combined loss? ",
  "url": "/competitions/bengaliai-cv19/discussion/123753",
  "author_name": "",
  "post_date": "2019-12-30T06:15:45.850950900Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi fellow mates. I could see most of the kernels were combining the loss of three components and backpropogating it together. I also tried  the same with densenet. But will it not make sense if we do three networks for three components?</p>\n\n<p>Kindly let me know your thoughts.</p>",
  "messages": [
    {
      "id": "706250",
      "postDate": "12/30/2019 06:15:45",
      "content": "<p>Hi fellow mates. I could see most of the kernels were combining the loss of three components and backpropogating it together. I also tried  the same with densenet. But will it not make sense if we do three networks for three components?</p>\n\n<p>Kindly let me know your thoughts.</p>",
      "rawMarkdown": "Hi fellow mates. I could see most of the kernels were combining the loss of three components and backpropogating it together. I also tried  the same with densenet. But will it not make sense if we do three networks for three components?\n\nKindly let me know your thoughts.",
      "votes": null
    },
    {
      "id": "706263",
      "postDate": "12/30/2019 06:46:09",
      "content": "<p>I tried that way since I am a newbie and have lack of knowledge to deal with these losses more technically. There could be better way used already by who are matured in this area.</p>",
      "rawMarkdown": "I tried that way since I am a newbie and have lack of knowledge to deal with these losses more technically. There could be better way used already by who are matured in this area.",
      "votes": null
    },
    {
      "id": "706306",
      "postDate": "12/30/2019 08:22:54",
      "content": "<p>yeah..i think here individual network is being used for each components. <a href=\"https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-cnn\">https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-cnn</a></p>",
      "rawMarkdown": "yeah..i think here individual network is being used for each components. https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-cnn",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 706263,
      "author_name": "hanjoonchoe",
      "author_url": "",
      "post_date": "12/30/2019 06:46:09",
      "content": "<p>I tried that way since I am a newbie and have lack of knowledge to deal with these losses more technically. There could be better way used already by who are matured in this area.</p>",
      "votes": null,
      "replies": [
        {
          "id": 706306,
          "author_name": "nandhuelan",
          "author_url": "",
          "post_date": "12/30/2019 08:22:54",
          "content": "<p>yeah..i think here individual network is being used for each components. <a href=\"https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-cnn\">https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-cnn</a></p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "706250": "Hi fellow mates. I could see most of the kernels were combining the loss of three components and backpropogating it together. I also tried  the same with densenet. But will it not make sense if we do three networks for three components?\n\nKindly let me know your thoughts.",
    "706263": "I tried that way since I am a newbie and have lack of knowledge to deal with these losses more technically. There could be better way used already by who are matured in this area.",
    "706306": "yeah..i think here individual network is being used for each components. https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-cnn"
  },
  "source": "meta"
}