{
  "id": 459744,
  "title": "155th solution",
  "url": "/competitions/child-mind-institute-detect-sleep-states/discussion/459744",
  "author_name": "ya9do",
  "post_date": "2023-12-06T13:10:06.617000",
  "votes": 9,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Thank you to organizers and congrats to all participants.</p>\n<p>I started working after I found <a href=\"https://www.kaggle.com/tubotubo\" target=\"_blank\">@tubotubo</a>'s public notebook &amp; his github repo.<br>\n<a href=\"https://www.kaggle.com/code/tubotubo/cmi-submit\" target=\"_blank\">https://www.kaggle.com/code/tubotubo/cmi-submit</a><br>\n<a href=\"https://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states\" target=\"_blank\">https://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states</a><br>\nFirst time for me to tackle on Deep learnning techninque instead of GBDT, I learned so many things.</p>\n<h2>My solution</h2>\n<p>Basically my solution comes from tubotubo 's public notebook.<br>\nDifferences are following;</p>\n<ul>\n<li>Use top-5 models instead of best model on an epoch.</li>\n<li>Ensembling on variety of K-Folds</li>\n</ul>\n<p>I found validation score fluctuates on learning, also private data is 75% of test dataset, I decided to ensemble as much as I can.</p>\n<h2>A Tip tried but not worked</h2>\n<p>Here is a tip I couldn't find on the discussion. This is not an effective tip. I'd appreciate any feedbacks.</p>\n<ul>\n<li>Mean/Standardization on each series_id, instead of whole dataset<br>\nI wanted my model to learn the difference btwn awake/sleep instead of each seires_id's difference.<br>\nSo I applied mean/standardization to anglez/enmo on each series_id, but it didn't work as I expected.</li>\n</ul>",
  "messages": [
    {
      "id": 2551021,
      "postDate": "2023-12-06T13:10:06.617Z",
      "content": "<p>Thank you to organizers and congrats to all participants.</p>\n<p>I started working after I found <a href=\"https://www.kaggle.com/tubotubo\" target=\"_blank\">@tubotubo</a>'s public notebook &amp; his github repo.<br>\n<a href=\"https://www.kaggle.com/code/tubotubo/cmi-submit\" target=\"_blank\">https://www.kaggle.com/code/tubotubo/cmi-submit</a><br>\n<a href=\"https://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states\" target=\"_blank\">https://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states</a><br>\nFirst time for me to tackle on Deep learnning techninque instead of GBDT, I learned so many things.</p>\n<h2>My solution</h2>\n<p>Basically my solution comes from tubotubo 's public notebook.<br>\nDifferences are following;</p>\n<ul>\n<li>Use top-5 models instead of best model on an epoch.</li>\n<li>Ensembling on variety of K-Folds</li>\n</ul>\n<p>I found validation score fluctuates on learning, also private data is 75% of test dataset, I decided to ensemble as much as I can.</p>\n<h2>A Tip tried but not worked</h2>\n<p>Here is a tip I couldn't find on the discussion. This is not an effective tip. I'd appreciate any feedbacks.</p>\n<ul>\n<li>Mean/Standardization on each series_id, instead of whole dataset<br>\nI wanted my model to learn the difference btwn awake/sleep instead of each seires_id's difference.<br>\nSo I applied mean/standardization to anglez/enmo on each series_id, but it didn't work as I expected.</li>\n</ul>",
      "rawMarkdown": "Thank you to organizers and congrats to all participants.\n\nI started working after I found @tubotubo's public notebook & his github repo.\nhttps://www.kaggle.com/code/tubotubo/cmi-submit\nhttps://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states\nFirst time for me to tackle on Deep learnning techninque instead of GBDT, I learned so many things.\n\n## My solution\nBasically my solution comes from tubotubo 's public notebook.\nDifferences are following;\n- Use top-5 models instead of best model on an epoch.\n- Ensembling on variety of K-Folds\n\nI found validation score fluctuates on learning, also private data is 75% of test dataset, I decided to ensemble as much as I can.\n\n## A Tip tried but not worked\nHere is a tip I couldn't find on the discussion. This is not an effective tip. I'd appreciate any feedbacks.\n- Mean/Standardization on each series_id, instead of whole dataset\nI wanted my model to learn the difference btwn awake/sleep instead of each seires_id's difference.\nSo I applied mean/standardization to anglez/enmo on each series_id, but it didn't work as I expected.",
      "votes": 9
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2551021": "Thank you to organizers and congrats to all participants.\n\nI started working after I found @tubotubo's public notebook & his github repo.\nhttps://www.kaggle.com/code/tubotubo/cmi-submit\nhttps://github.com/tubo213/kaggle-child-mind-institute-detect-sleep-states\nFirst time for me to tackle on Deep learnning techninque instead of GBDT, I learned so many things.\n\n## My solution\nBasically my solution comes from tubotubo 's public notebook.\nDifferences are following;\n- Use top-5 models instead of best model on an epoch.\n- Ensembling on variety of K-Folds\n\nI found validation score fluctuates on learning, also private data is 75% of test dataset, I decided to ensemble as much as I can.\n\n## A Tip tried but not worked\nHere is a tip I couldn't find on the discussion. This is not an effective tip. I'd appreciate any feedbacks.\n- Mean/Standardization on each series_id, instead of whole dataset\nI wanted my model to learn the difference btwn awake/sleep instead of each seires_id's difference.\nSo I applied mean/standardization to anglez/enmo on each series_id, but it didn't work as I expected."
  }
}