{ "id": 322596, "title": "Phones Mean does not work for test data", "url": "/competitions/smartphone-decimeter-2022/discussion/322596", "author_name": "", "post_date": "2022-05-03T05:58:46.788031600Z", "votes": 14, "comment_count": 1, "views": 0, "content": "

I'm excited about my first Google Smartphone Decimeter Challenge in a year!
\nI hope we can fight and grow together😆

\n

In this competition, it appears that test data was produced by using only one cell phone per measurement. (However, train data is measured by multiple phones)

\n

Therefore, \"phones mean\", which simply raises the score in the previous competition, was not effective in the test.

\n

But when I did it in train, the score improved from 4.357 to 3.376.
\nHere is the notebook I checked: GSDC2 - Phones Mean (but test isn't improved)

\n

I hope this will be of some help.

\n

Reference

\n", "messages": [ { "id": "1775521", "postDate": "05/03/2022 05:58:46", "content": "

I'm excited about my first Google Smartphone Decimeter Challenge in a year!
\nI hope we can fight and grow together😆

\n

In this competition, it appears that test data was produced by using only one cell phone per measurement. (However, train data is measured by multiple phones)

\n

Therefore, \"phones mean\", which simply raises the score in the previous competition, was not effective in the test.

\n

But when I did it in train, the score improved from 4.357 to 3.376.
\nHere is the notebook I checked: GSDC2 - Phones Mean (but test isn't improved)

\n

I hope this will be of some help.

\n

Reference

\n", "rawMarkdown": "I'm excited about my first Google Smartphone Decimeter Challenge in a year!\nI hope we can fight and grow together😆\n\nIn this competition, it appears that test data was produced by using only one cell phone per measurement. (However, train data is measured by multiple phones)\n\nTherefore, [\"phones mean\"](https://www.kaggle.com/code/t88take/gsdc-phones-mean-prediction), which simply raises the score in the previous competition, was not effective in the test.\n\nBut when I did it in train, the score improved from 4.357 to 3.376.\nHere is the notebook I checked: [GSDC2 - Phones Mean (but test isn't improved)](https://www.kaggle.com/code/columbia2131/gsdc2-phones-mean-but-test-isn-t-improved)\n\nI hope this will be of some help.\n\n\n### Reference\n* [GSDC phones mean prediction](https://www.kaggle.com/code/t88take/gsdc-phones-mean-prediction) created by @t88take", "votes": null }, { "id": "1805168", "postDate": "05/29/2022 21:10:19", "content": "

if you cannot mean the rover, how about mean of (rover, base) pairs

", "rawMarkdown": "if you cannot mean the rover, how about mean of (rover, base) pairs", "votes": null } ], "comments": [ { "id": 1805168, "author_name": "hengck23", "author_url": "", "post_date": "05/29/2022 21:10:19", "content": "

if you cannot mean the rover, how about mean of (rover, base) pairs

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1775521": "I'm excited about my first Google Smartphone Decimeter Challenge in a year!\nI hope we can fight and grow together😆\n\nIn this competition, it appears that test data was produced by using only one cell phone per measurement. (However, train data is measured by multiple phones)\n\nTherefore, [\"phones mean\"](https://www.kaggle.com/code/t88take/gsdc-phones-mean-prediction), which simply raises the score in the previous competition, was not effective in the test.\n\nBut when I did it in train, the score improved from 4.357 to 3.376.\nHere is the notebook I checked: [GSDC2 - Phones Mean (but test isn't improved)](https://www.kaggle.com/code/columbia2131/gsdc2-phones-mean-but-test-isn-t-improved)\n\nI hope this will be of some help.\n\n\n### Reference\n* [GSDC phones mean prediction](https://www.kaggle.com/code/t88take/gsdc-phones-mean-prediction) created by @t88take", "1805168": "if you cannot mean the rover, how about mean of (rover, base) pairs" }, "source": "meta" }