{
  "id": 233080,
  "title": "Post processing, Ensembling - Overfiiting... avoiding the dreaded shakeup!!",
  "url": "/competitions/indoor-location-navigation/discussion/233080",
  "author_name": "Kamal Das",
  "post_date": "2021-04-17T06:16:19.478000",
  "votes": 9,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi All,</p>\n<p>Most of us are using Post processing (multiple in my case); ensembling with own (and perhaps top public models)</p>\n<p>\"This leaderboard is calculated with approximately 15% of the test data.  The final results will be based on the other 85%, so the final standings may be different.\"</p>\n<p>15% is low and I fear I am overfitting. </p>\n<p>Given the huge amount of Post processing I am not sure how to look at my CV and gauge if my results are more generalised or overfitting… I think i am massively overfitting to the public LB/15%</p>\n<p>what is your recommendation on avoiding this and avoiding the nasty shakeup? like this meme keeps reminding me…</p>\n<p><a href=\"https://storage.googleapis.com/kagglesdsdata/datasets/1277000/2128116/lb.JPG?X-Goog-Algorithm=GOOG4-RSA-SHA256&amp;X-Goog-Credential=databundle-worker-v2%40kaggle-161607.iam.gserviceaccount.com%2F20210419%2Fauto%2Fstorage%2Fgoog4_request&amp;X-Goog-Date=20210419T162539Z&amp;X-Goog-Expires=172799&amp;X-Goog-SignedHeaders=host&amp;X-Goog-Signature=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\" target=\"_blank\">image</a></p>\n<p>Link if image does not show: <a href=\"https://www.kaggle.com/kmldas/gif-for-discussion?select=lb.JPG\" target=\"_blank\">https://www.kaggle.com/kmldas/gif-for-discussion?select=lb.JPG</a></p>",
  "messages": [
    {
      "id": 1276108,
      "postDate": "2021-04-17T06:16:19.480Z",
      "content": "<p>Hi All,</p>\n<p>Most of us are using Post processing (multiple in my case); ensembling with own (and perhaps top public models)</p>\n<p>\"This leaderboard is calculated with approximately 15% of the test data.  The final results will be based on the other 85%, so the final standings may be different.\"</p>\n<p>15% is low and I fear I am overfitting. </p>\n<p>Given the huge amount of Post processing I am not sure how to look at my CV and gauge if my results are more generalised or overfitting… I think i am massively overfitting to the public LB/15%</p>\n<p>what is your recommendation on avoiding this and avoiding the nasty shakeup? like this meme keeps reminding me…</p>\n<p><a href=\"https://storage.googleapis.com/kagglesdsdata/datasets/1277000/2128116/lb.JPG?X-Goog-Algorithm=GOOG4-RSA-SHA256&amp;X-Goog-Credential=databundle-worker-v2%40kaggle-161607.iam.gserviceaccount.com%2F20210419%2Fauto%2Fstorage%2Fgoog4_request&amp;X-Goog-Date=20210419T162539Z&amp;X-Goog-Expires=172799&amp;X-Goog-SignedHeaders=host&amp;X-Goog-Signature=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\" target=\"_blank\">image</a></p>\n<p>Link if image does not show: <a href=\"https://www.kaggle.com/kmldas/gif-for-discussion?select=lb.JPG\" target=\"_blank\">https://www.kaggle.com/kmldas/gif-for-discussion?select=lb.JPG</a></p>",
      "rawMarkdown": "Hi All,\n\nMost of us are using Post processing (multiple in my case); ensembling with own (and perhaps top public models)\n\n\"This leaderboard is calculated with approximately 15% of the test data.  The final results will be based on the other 85%, so the final standings may be different.\"\n\n15% is low and I fear I am overfitting. \n\nGiven the huge amount of Post processing I am not sure how to look at my CV and gauge if my results are more generalised or overfitting... I think i am massively overfitting to the public LB/15%\n\nwhat is your recommendation on avoiding this and avoiding the nasty shakeup? like this meme keeps reminding me...\n\n[image](https://storage.googleapis.com/kagglesdsdata/datasets/1277000/2128116/lb.JPG?X-Goog-Algorithm=GOOG4-RSA-SHA256&X-Goog-Credential=databundle-worker-v2%40kaggle-161607.iam.gserviceaccount.com%2F20210419%2Fauto%2Fstorage%2Fgoog4_request&X-Goog-Date=20210419T162539Z&X-Goog-Expires=172799&X-Goog-SignedHeaders=host&X-Goog-Signature=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)\n\nLink if image does not show: https://www.kaggle.com/kmldas/gif-for-discussion?select=lb.JPG",
      "votes": 9
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1276108": "Hi All,\n\nMost of us are using Post processing (multiple in my case); ensembling with own (and perhaps top public models)\n\n\"This leaderboard is calculated with approximately 15% of the test data.  The final results will be based on the other 85%, so the final standings may be different.\"\n\n15% is low and I fear I am overfitting. \n\nGiven the huge amount of Post processing I am not sure how to look at my CV and gauge if my results are more generalised or overfitting... I think i am massively overfitting to the public LB/15%\n\nwhat is your recommendation on avoiding this and avoiding the nasty shakeup? like this meme keeps reminding me...\n\n[image](https://storage.googleapis.com/kagglesdsdata/datasets/1277000/2128116/lb.JPG?X-Goog-Algorithm=GOOG4-RSA-SHA256&X-Goog-Credential=databundle-worker-v2%40kaggle-161607.iam.gserviceaccount.com%2F20210419%2Fauto%2Fstorage%2Fgoog4_request&X-Goog-Date=20210419T162539Z&X-Goog-Expires=172799&X-Goog-SignedHeaders=host&X-Goog-Signature=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)\n\nLink if image does not show: https://www.kaggle.com/kmldas/gif-for-discussion?select=lb.JPG"
  }
}