{ "id": 93911, "title": "The race for (overfitting) LANL", "url": "/competitions/LANL-Earthquake-Prediction/discussion/93911", "author_name": "", "post_date": "2019-05-31T03:48:07.115407300Z", "votes": 28, "comment_count": 18, "views": 0, "content": "

I made a quick kernel showing how some of the top public LB teams arrived at their current position with an interactive plot. I can update it in the coming days as scores change if people find it fun.\n\"\"\nhttps://www.kaggle.com/robikscube/the-race-for-lanl

", "messages": [ { "id": "540106", "postDate": "05/31/2019 03:48:07", "content": "

I made a quick kernel showing how some of the top public LB teams arrived at their current position with an interactive plot. I can update it in the coming days as scores change if people find it fun.\n\"\"\nhttps://www.kaggle.com/robikscube/the-race-for-lanl

", "rawMarkdown": "I made a quick kernel showing how some of the top public LB teams arrived at their current position with an interactive plot. I can update it in the coming days as scores change if people find it fun.\n![](https://i.imgur.com/DHR598A.gif)\nhttps://www.kaggle.com/robikscube/the-race-for-lanl", "votes": null }, { "id": "540116", "postDate": "05/31/2019 04:01:29", "content": "

Great work rob!

\n\n

Keep spreading your magic.

", "rawMarkdown": "Great work rob!\n\nKeep spreading your magic.", "votes": null }, { "id": "540152", "postDate": "05/31/2019 05:27:43", "content": "

The real magic will be figuring out how to plot everyoneโ€™s movement from public LB to private LB after the impending shakeup. ๐Ÿ˜†

", "rawMarkdown": "The real magic will be figuring out how to plot everyoneโ€™s movement from public LB to private LB after the impending shakeup. ๐Ÿ˜†", "votes": null }, { "id": "540207", "postDate": "05/31/2019 06:37:55", "content": "

I would call it the race to overfitting.

", "rawMarkdown": "I would call it the race to overfitting.", "votes": null }, { "id": "540323", "postDate": "05/31/2019 11:06:31", "content": "

nice graph!

", "rawMarkdown": "nice graph!", "votes": null }, { "id": "540324", "postDate": "05/31/2019 11:06:54", "content": "

Cool plot

", "rawMarkdown": "Cool plot", "votes": null }, { "id": "540336", "postDate": "05/31/2019 11:16:55", "content": "

Make a nice little gif.

", "rawMarkdown": "Make a nice little [gif](https://www.kaggle.com/c/microsoft-malware-prediction/discussion/84388#latest-492728).", "votes": null }, { "id": "540514", "postDate": "05/31/2019 15:40:10", "content": "

You're probably right- for some. It will be fun after the competition is over to see if we can identify correlation between public LB submission trends and final private LB scores. Either way it's fun to see how people have arrived at their current public LB positions. It seems like many stall out at the 1.29 mark - but you've managed to crack it ๐Ÿ˜„

", "rawMarkdown": "You're probably right- for some. It will be fun after the competition is over to see if we can identify correlation between public LB submission trends and final private LB scores. Either way it's fun to see how people have arrived at their current public LB positions. It seems like many stall out at the 1.29 mark - but you've managed to crack it ๐Ÿ˜„", "votes": null }, { "id": "540522", "postDate": "05/31/2019 15:48:33", "content": "

I updated the post to include a gif version of the plot

", "rawMarkdown": "I updated the post to include a gif version of the plot", "votes": null }, { "id": "540541", "postDate": "05/31/2019 16:03:52", "content": "
\n

but you've managed to crack it ๐Ÿ˜„

\n
\n\n

Because we overfit :)

", "rawMarkdown": "> but you've managed to crack it ๐Ÿ˜„ \n\nBecause we overfit :)", "votes": null }, { "id": "540584", "postDate": "05/31/2019 17:32:42", "content": "

Wow amazing...... ๐Ÿ‘ Nice animation.....

", "rawMarkdown": "Wow amazing...... ๐Ÿ‘ Nice animation.....", "votes": null }, { "id": "540629", "postDate": "05/31/2019 19:13:45", "content": "

I updated the notebook with the latest data and added a few plots.

\n\n

I thought this one was especially interesting, each faint line represents a team. Can you guess what the scores of public kernels were? Who else thinks we will see a bunch move down to the top kernel score in the next few days?

\n\n

Also note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.

\n\n

\"scores\"

", "rawMarkdown": "I updated the notebook with the latest data and added a few plots.\n\nI thought this one was especially interesting, each faint line represents a team. Can you guess what the scores of public kernels were? Who else thinks we will see a bunch move down to the top kernel score in the next few days?\n\nAlso note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.\n\n![scores](https://i.imgur.com/mcvMHKm.png)", "votes": null }, { "id": "540765", "postDate": "06/01/2019 03:09:41", "content": "

good job ,it's really interesting XD

", "rawMarkdown": "good job ,it's really interesting XD", "votes": null }, { "id": "541336", "postDate": "06/02/2019 08:57:26", "content": "
\n

Also note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.

\n
\n\n

Yeah, I noticed this kernel and after blending with it got 1.381.

", "rawMarkdown": "> Also note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.\n\nYeah, I noticed this kernel and after blending with it got 1.381.", "votes": null }, { "id": "541888", "postDate": "06/03/2019 06:50:27", "content": "

wow~๏ผŒso cool

", "rawMarkdown": "wow~๏ผŒso cool", "votes": null }, { "id": "542783", "postDate": "06/04/2019 04:57:53", "content": "

sometimes being lazy saves from overfitting :)

", "rawMarkdown": "sometimes being lazy saves from overfitting :)", "votes": null }, { "id": "544436", "postDate": "06/05/2019 14:06:09", "content": "

Cool! Really interesting stuff .

", "rawMarkdown": "Cool! Really interesting stuff .", "votes": null }, { "id": "544442", "postDate": "06/05/2019 14:13:14", "content": "

I agree with that, but I am not sure about why you say it here. Do you think we were lazy hence did not overfit or , to the contrary, that we should have been more lazy?

", "rawMarkdown": "I agree with that, but I am not sure about why you say it here. Do you think we were lazy hence did not overfit or , to the contrary, that we should have been more lazy?", "votes": null }, { "id": "546197", "postDate": "06/06/2019 10:48:53", "content": "

This was my philosophical reaction to the title of this discussion. I am having great time learning the top solutions of this competition, including yours. Thanks a lot for sharing it.

", "rawMarkdown": "This was my philosophical reaction to the title of this discussion. I am having great time learning the top solutions of this competition, including yours. Thanks a lot for sharing it.", "votes": null } ], "comments": [ { "id": 540116, "author_name": "interneuron", "author_url": "", "post_date": "05/31/2019 04:01:29", "content": "

Great work rob!

\n\n

Keep spreading your magic.

", "votes": null, "replies": [ { "id": 540152, "author_name": "robikscube", "author_url": "", "post_date": "05/31/2019 05:27:43", "content": "

The real magic will be figuring out how to plot everyoneโ€™s movement from public LB to private LB after the impending shakeup. ๐Ÿ˜†

", "votes": null, "replies": [] } ] }, { "id": 540207, "author_name": "cpmpml", "author_url": "", "post_date": "05/31/2019 06:37:55", "content": "

I would call it the race to overfitting.

", "votes": null, "replies": [ { "id": 540514, "author_name": "robikscube", "author_url": "", "post_date": "05/31/2019 15:40:10", "content": "

You're probably right- for some. It will be fun after the competition is over to see if we can identify correlation between public LB submission trends and final private LB scores. Either way it's fun to see how people have arrived at their current public LB positions. It seems like many stall out at the 1.29 mark - but you've managed to crack it ๐Ÿ˜„

", "votes": null, "replies": [] }, { "id": 540541, "author_name": "cpmpml", "author_url": "", "post_date": "05/31/2019 16:03:52", "content": "
\n

but you've managed to crack it ๐Ÿ˜„

\n
\n\n

Because we overfit :)

", "votes": null, "replies": [] } ] }, { "id": 540323, "author_name": "frtgnn", "author_url": "", "post_date": "05/31/2019 11:06:31", "content": "

nice graph!

", "votes": null, "replies": [] }, { "id": 540324, "author_name": "cdeotte", "author_url": "", "post_date": "05/31/2019 11:06:54", "content": "

Cool plot

", "votes": null, "replies": [] }, { "id": 540336, "author_name": "teeyee314", "author_url": "", "post_date": "05/31/2019 11:16:55", "content": "

Make a nice little gif.

", "votes": null, "replies": [ { "id": 540522, "author_name": "robikscube", "author_url": "", "post_date": "05/31/2019 15:48:33", "content": "

I updated the post to include a gif version of the plot

", "votes": null, "replies": [] } ] }, { "id": 540584, "author_name": "himusoni", "author_url": "", "post_date": "05/31/2019 17:32:42", "content": "

Wow amazing...... ๐Ÿ‘ Nice animation.....

", "votes": null, "replies": [] }, { "id": 540629, "author_name": "robikscube", "author_url": "", "post_date": "05/31/2019 19:13:45", "content": "

I updated the notebook with the latest data and added a few plots.

\n\n

I thought this one was especially interesting, each faint line represents a team. Can you guess what the scores of public kernels were? Who else thinks we will see a bunch move down to the top kernel score in the next few days?

\n\n

Also note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.

\n\n

\"scores\"

", "votes": null, "replies": [ { "id": 541336, "author_name": "sergeyzlobin", "author_url": "", "post_date": "06/02/2019 08:57:26", "content": "
\n

Also note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.

\n
\n\n

Yeah, I noticed this kernel and after blending with it got 1.381.

", "votes": null, "replies": [] } ] }, { "id": 540765, "author_name": "takeiy", "author_url": "", "post_date": "06/01/2019 03:09:41", "content": "

good job ,it's really interesting XD

", "votes": null, "replies": [] }, { "id": 541888, "author_name": "meistermorxrc", "author_url": "", "post_date": "06/03/2019 06:50:27", "content": "

wow~๏ผŒso cool

", "votes": null, "replies": [] }, { "id": 542783, "author_name": "vladk17", "author_url": "", "post_date": "06/04/2019 04:57:53", "content": "

sometimes being lazy saves from overfitting :)

", "votes": null, "replies": [ { "id": 544442, "author_name": "cpmpml", "author_url": "", "post_date": "06/05/2019 14:13:14", "content": "

I agree with that, but I am not sure about why you say it here. Do you think we were lazy hence did not overfit or , to the contrary, that we should have been more lazy?

", "votes": null, "replies": [] }, { "id": 546197, "author_name": "vladk17", "author_url": "", "post_date": "06/06/2019 10:48:53", "content": "

This was my philosophical reaction to the title of this discussion. I am having great time learning the top solutions of this competition, including yours. Thanks a lot for sharing it.

", "votes": null, "replies": [] } ] }, { "id": 544436, "author_name": "sriharipramod", "author_url": "", "post_date": "06/05/2019 14:06:09", "content": "

Cool! Really interesting stuff .

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "540106": "I made a quick kernel showing how some of the top public LB teams arrived at their current position with an interactive plot. I can update it in the coming days as scores change if people find it fun.\n![](https://i.imgur.com/DHR598A.gif)\nhttps://www.kaggle.com/robikscube/the-race-for-lanl", "540116": "Great work rob!\n\nKeep spreading your magic.", "540152": "The real magic will be figuring out how to plot everyoneโ€™s movement from public LB to private LB after the impending shakeup. ๐Ÿ˜†", "540207": "I would call it the race to overfitting.", "540323": "nice graph!", "540324": "Cool plot", "540336": "Make a nice little [gif](https://www.kaggle.com/c/microsoft-malware-prediction/discussion/84388#latest-492728).", "540514": "You're probably right- for some. It will be fun after the competition is over to see if we can identify correlation between public LB submission trends and final private LB scores. Either way it's fun to see how people have arrived at their current public LB positions. It seems like many stall out at the 1.29 mark - but you've managed to crack it ๐Ÿ˜„", "540522": "I updated the post to include a gif version of the plot", "540541": "> but you've managed to crack it ๐Ÿ˜„ \n\nBecause we overfit :)", "540584": "Wow amazing...... ๐Ÿ‘ Nice animation.....", "540629": "I updated the notebook with the latest data and added a few plots.\n\nI thought this one was especially interesting, each faint line represents a team. Can you guess what the scores of public kernels were? Who else thinks we will see a bunch move down to the top kernel score in the next few days?\n\nAlso note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.\n\n![scores](https://i.imgur.com/mcvMHKm.png)", "540765": "good job ,it's really interesting XD", "541336": "> Also note the faint line jump down to ~1.38. I think that was due to a kernel that was only public for a few hours.\n\nYeah, I noticed this kernel and after blending with it got 1.381.", "541888": "wow~๏ผŒso cool", "542783": "sometimes being lazy saves from overfitting :)", "544436": "Cool! Really interesting stuff .", "544442": "I agree with that, but I am not sure about why you say it here. Do you think we were lazy hence did not overfit or , to the contrary, that we should have been more lazy?", "546197": "This was my philosophical reaction to the title of this discussion. I am having great time learning the top solutions of this competition, including yours. Thanks a lot for sharing it." }, "source": "meta" }