{ "id": 77911, "title": "Nice overview paper on model evaluation and model selection", "url": "/competitions/elo-merchant-category-recommendation/discussion/77911", "author_name": "Yifan Xie", "post_date": "2019-01-17T15:18:10.747000", "votes": 46, "comment_count": 11, "views": 0, "content": "

Like many other competitions, we can see there is a lot of discussions on model evaluation in this competition - validation strategies, choice of \"k\" in K-fold, feature selection, you name it. I came across the following recent paper that gives a very good overview, and cover some statistical aspect of cross-validation that is not often discussed here:

\n\n

Here it is: Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning, by Sebastian Raschka, November 2018

\n\n

Hopefully this helps many of us to clarify our thought and make more insightful decisions - especially someone like me who usually overfit things :)

", "messages": [ { "id": 457504, "postDate": "2019-01-17T15:18:10.747Z", "content": "

Like many other competitions, we can see there is a lot of discussions on model evaluation in this competition - validation strategies, choice of \"k\" in K-fold, feature selection, you name it. I came across the following recent paper that gives a very good overview, and cover some statistical aspect of cross-validation that is not often discussed here:

\n\n

Here it is: Model Evaluation, Model Selection, and Algorithm Selection in Machine Learning, by Sebastian Raschka, November 2018

\n\n

Hopefully this helps many of us to clarify our thought and make more insightful decisions - especially someone like me who usually overfit things :)

", "rawMarkdown": "Like many other competitions, we can see there is a lot of discussions on model evaluation in this competition - validation strategies, choice of \"k\" in K-fold, feature selection, you name it. I came across the following recent paper that gives a very good overview, and cover some statistical aspect of cross-validation that is not often discussed here:\n\nHere it is: [Model Evaluation, Model Selection, and Algorithm Selection in Machine Learnin][1]g, by Sebastian Raschka, November 2018\n\nHopefully this helps many of us to clarify our thought and make more insightful decisions - especially someone like me who usually overfit things :) \n\n [1]: https://arxiv.org/pdf/1811.12808.pdf", "votes": 46 }, { "id": 650763, "postDate": "2019-10-16T16:55:51.900Z", "content": "

Thanks for sharing .\n@psvishnu check this link...

", "rawMarkdown": "Thanks for sharing .\n@psvishnu check this link...", "votes": 1, "replies": [ { "id": 650858, "postDate": "2019-10-16T19:01:32.157Z", "content": "

@tarunkoutharapu it looks interesting, next on my read list

", "rawMarkdown": "@tarunkoutharapu it looks interesting, next on my read list" } ] }, { "id": 457988, "postDate": "2019-01-18T13:39:46.557Z", "content": "

Thanks u for sharing :)\nIt is very helpful for me!

", "rawMarkdown": "Thanks u for sharing :)\nIt is very helpful for me! ", "votes": 2 }, { "id": 466525, "postDate": "2019-02-05T14:57:08.687Z", "content": "

Thanks for sharing. But none of those approaches are helping to get reliable CV ( CV rmse is not correlated to test rmse )for me. Can you please share your personal approach or gives some tips?

", "rawMarkdown": "Thanks for sharing. But none of those approaches are helping to get reliable CV ( CV rmse is not correlated to test rmse )for me. Can you please share your personal approach or gives some tips?" }, { "id": 457747, "postDate": "2019-01-18T02:57:42.630Z", "content": "

Nice job

", "rawMarkdown": "Nice job" }, { "id": 457685, "postDate": "2019-01-17T23:47:43.080Z", "content": "

thinks!

", "rawMarkdown": "thinks!" }, { "id": 458144, "postDate": "2019-01-18T22:14:30.023Z", "rawMarkdown": "", "isDeleted": true }, { "id": 457519, "postDate": "2019-01-17T15:48:40.840Z", "content": "

Thanks for sharing :)

", "rawMarkdown": "Thanks for sharing :)", "votes": 1 }, { "id": 1518581, "postDate": "2021-09-20T19:17:44.670Z", "content": "

Thanks for sharing

", "rawMarkdown": "Thanks for sharing" }, { "id": 466738, "postDate": "2019-02-05T21:58:38.440Z", "content": "

Thanks for sharing!

", "rawMarkdown": "Thanks for sharing!" }, { "id": 458407, "postDate": "2019-01-19T14:59:41.887Z", "content": "

Thanks for Sharing!!

", "rawMarkdown": "Thanks for Sharing!!\n" } ], "comments": [ { "id": 650763, "author_name": "Tarun K", "author_url": "", "post_date": "2019-10-16T16:55:51.900000", "content": "

Thanks for sharing .\n@psvishnu check this link...

", "votes": 1, "replies": [ { "id": 650858, "author_name": "psv", "author_url": "", "post_date": "2019-10-16T19:01:32.157000", "content": "

@tarunkoutharapu it looks interesting, next on my read list

", "votes": 0, "replies": [] } ] }, { "id": 457988, "author_name": "SeongYunByeon", "author_url": "", "post_date": "2019-01-18T13:39:46.557000", "content": "

Thanks u for sharing :)\nIt is very helpful for me!

", "votes": 2, "replies": [] }, { "id": 466525, "author_name": "Salamat", "author_url": "", "post_date": "2019-02-05T14:57:08.687000", "content": "

Thanks for sharing. But none of those approaches are helping to get reliable CV ( CV rmse is not correlated to test rmse )for me. Can you please share your personal approach or gives some tips?

", "votes": 0, "replies": [] }, { "id": 457747, "author_name": "Jianfei Wang", "author_url": "", "post_date": "2019-01-18T02:57:42.630000", "content": "

Nice job

", "votes": 0, "replies": [] }, { "id": 457685, "author_name": "Zhuang Jia", "author_url": "", "post_date": "2019-01-17T23:47:43.080000", "content": "

thinks!

", "votes": 0, "replies": [] }, { "id": 458144, "author_name": "", "author_url": "", "post_date": "2019-01-18T22:14:30.023000", "content": "", "votes": 0, "replies": [] }, { "id": 457519, "author_name": "Manoj", "author_url": "", "post_date": "2019-01-17T15:48:40.840000", "content": "

Thanks for sharing :)

", "votes": 1, "replies": [] }, { "id": 1518581, "author_name": "Konduri Sai Somanadha Sastry", "author_url": "", "post_date": "2021-09-20T19:17:44.670000", "content": "

Thanks for sharing

", "votes": 0, "replies": [] }, { "id": 466738, "author_name": "Guto P", "author_url": "", "post_date": "2019-02-05T21:58:38.440000", "content": "

Thanks for sharing!

", "votes": 0, "replies": [] }, { "id": 458407, "author_name": "Pranav Nema", "author_url": "", "post_date": "2019-01-19T14:59:41.887000", "content": "

Thanks for Sharing!!

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "457504": "Like many other competitions, we can see there is a lot of discussions on model evaluation in this competition - validation strategies, choice of \"k\" in K-fold, feature selection, you name it. I came across the following recent paper that gives a very good overview, and cover some statistical aspect of cross-validation that is not often discussed here:\n\nHere it is: [Model Evaluation, Model Selection, and Algorithm Selection in Machine Learnin][1]g, by Sebastian Raschka, November 2018\n\nHopefully this helps many of us to clarify our thought and make more insightful decisions - especially someone like me who usually overfit things :) \n\n [1]: https://arxiv.org/pdf/1811.12808.pdf", "650763": "Thanks for sharing .\n@psvishnu check this link...", "457988": "Thanks u for sharing :)\nIt is very helpful for me! ", "466525": "Thanks for sharing. But none of those approaches are helping to get reliable CV ( CV rmse is not correlated to test rmse )for me. Can you please share your personal approach or gives some tips?", "457747": "Nice job", "457685": "thinks!", "458144": "", "457519": "Thanks for sharing :)", "1518581": "Thanks for sharing", "466738": "Thanks for sharing!", "458407": "Thanks for Sharing!!\n" } }