{
  "id": 85268,
  "title": "Normalization is all you need",
  "url": "/competitions/vsb-power-line-fault-detection/discussion/85268",
  "author_name": "",
  "post_date": "2019-03-22T15:31:49.980065Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Instead of using the Adam optimizer I used the Adam optimizer with Weight Normalization.\nThis didn't only help me to survive the shake-up but also speed-up the training of my model.\nI think adding normalization techniques to my model makes it robust to the change made between the test set in private and public leaderboard. </p>",
  "messages": [
    {
      "id": "496793",
      "postDate": "03/22/2019 15:31:49",
      "content": "<p>Instead of using the Adam optimizer I used the Adam optimizer with Weight Normalization.\nThis didn't only help me to survive the shake-up but also speed-up the training of my model.\nI think adding normalization techniques to my model makes it robust to the change made between the test set in private and public leaderboard. </p>",
      "rawMarkdown": "Instead of using the Adam optimizer I used the Adam optimizer with Weight Normalization.\nThis didn't only help me to survive the shake-up but also speed-up the training of my model.\nI think adding normalization techniques to my model makes it robust to the change made between the test set in private and public leaderboard.",
      "votes": null
    },
    {
      "id": "497038",
      "postDate": "03/22/2019 22:05:37",
      "content": "<p>How different are your local matthews correlation coefficient, public LB and private LB? I want to know the concrete scores.</p>",
      "rawMarkdown": "How different are your local matthews correlation coefficient, public LB and private LB? I want to know the concrete scores.",
      "votes": null
    },
    {
      "id": "497463",
      "postDate": "03/23/2019 15:37:11",
      "content": "<p><a href=\"/yatzhash\">@yatzhash</a>  :  Public score was 0.59 and private score was 0.62</p>",
      "rawMarkdown": "yatzhash  :  Public score was 0.59 and private score was 0.62",
      "votes": null
    },
    {
      "id": "504367",
      "postDate": "03/31/2019 13:40:34",
      "content": "<p>I'm sorry for my late reply. Thank you. </p>",
      "rawMarkdown": "I'm sorry for my late reply. Thank you.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 497038,
      "author_name": "yatzhash",
      "author_url": "",
      "post_date": "03/22/2019 22:05:37",
      "content": "<p>How different are your local matthews correlation coefficient, public LB and private LB? I want to know the concrete scores.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 497463,
      "author_name": "azizbenothman",
      "author_url": "",
      "post_date": "03/23/2019 15:37:11",
      "content": "<p><a href=\"/yatzhash\">@yatzhash</a>  :  Public score was 0.59 and private score was 0.62</p>",
      "votes": null,
      "replies": [
        {
          "id": 504367,
          "author_name": "yatzhash",
          "author_url": "",
          "post_date": "03/31/2019 13:40:34",
          "content": "<p>I'm sorry for my late reply. Thank you. </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "496793": "Instead of using the Adam optimizer I used the Adam optimizer with Weight Normalization.\nThis didn't only help me to survive the shake-up but also speed-up the training of my model.\nI think adding normalization techniques to my model makes it robust to the change made between the test set in private and public leaderboard.",
    "497038": "How different are your local matthews correlation coefficient, public LB and private LB? I want to know the concrete scores.",
    "497463": "yatzhash  :  Public score was 0.59 and private score was 0.62",
    "504367": "I'm sorry for my late reply. Thank you."
  },
  "source": "meta"
}