{ "id": 105939, "title": "Last minute surprise?", "url": "/competitions/champs-scalar-coupling/discussion/105939", "author_name": "", "post_date": "2019-08-27T09:34:23.710266800Z", "votes": 3, "comment_count": 29, "views": 0, "content": "

Some teams that were near the top when they submitted, haven't submitted for a while, like Tony Y., DL Guys, and Jaechang Lim. Will these jump to prize zone at the last minute?

\n\n

I hope not ;)

", "messages": [ { "id": "608948", "postDate": "08/27/2019 09:34:23", "content": "

Some teams that were near the top when they submitted, haven't submitted for a while, like Tony Y., DL Guys, and Jaechang Lim. Will these jump to prize zone at the last minute?

\n\n

I hope not ;)

", "rawMarkdown": "Some teams that were near the top when they submitted, haven't submitted for a while, like Tony Y., DL Guys, and Jaechang Lim. Will these jump to prize zone at the last minute?\n\nI hope not ;)", "votes": null }, { "id": "608960", "postDate": "08/27/2019 09:40:51", "content": "

Don’t do that,the submission maybe failed as some kinds of reasons.

", "rawMarkdown": "Don’t do that,the submission maybe failed as some kinds of reasons.", "votes": null }, { "id": "609086", "postDate": "08/27/2019 12:03:28", "content": "

I think last minute surprise is probably probable in this kind of probably stable-CV-LB competition.

", "rawMarkdown": "I think last minute surprise is probably probable in this kind of probably stable-CV-LB competition.", "votes": null }, { "id": "609105", "postDate": "08/27/2019 12:17:17", "content": "

At this point I am more intrigued about private vs public LB shuffle.\nHow big can it be in terms of LMAE? 0.01 0.001?

", "rawMarkdown": "At this point I am more intrigued about private vs public LB shuffle.\nHow big can it be in terms of LMAE? 0.01 0.001?", "votes": null }, { "id": "609112", "postDate": "08/27/2019 12:20:37", "content": "

I predict that no more than 2 teams in top 20 will shuffle more than 3 places.

", "rawMarkdown": "I predict that no more than 2 teams in top 20 will shuffle more than 3 places.", "votes": null }, { "id": "609113", "postDate": "08/27/2019 12:21:39", "content": "

I hope my last-minute surprise will work. Still struggling on the test set. My train & validation score in LGBM is almost -11 + with minimum variance, but I don't know these values error occurs in the test set. :-p

", "rawMarkdown": "I hope my last-minute surprise will work. Still struggling on the test set. My train & validation score in LGBM is almost -11 + with minimum variance, but I don't know these values error occurs in the test set. :-p", "votes": null }, { "id": "609116", "postDate": "08/27/2019 12:27:01", "content": "
\n

I am more intrigued about private vs public LB shuffle.

\n
\n\n

Read: will our lead over 2nd team on public LB be enough to stay 1st in private LB? ;)

", "rawMarkdown": "> I am more intrigued about private vs public LB shuffle.\n\nRead: will our lead over 2nd team on public LB be enough to stay 1st in private LB? ;)", "votes": null }, { "id": "609120", "postDate": "08/27/2019 12:29:43", "content": "

I really hope that nothing spectacular happens ;)

", "rawMarkdown": "I really hope that nothing spectacular happens ;)", "votes": null }, { "id": "609123", "postDate": "08/27/2019 12:30:47", "content": "

Public LB is not frozen yet!

", "rawMarkdown": "Public LB is not frozen yet!", "votes": null }, { "id": "609126", "postDate": "08/27/2019 12:32:50", "content": "

@cpmpml please don't say it out loud like that ;) we all know their creepy feelings

", "rawMarkdown": "cpmpml please don't say it out loud like that ;) we all know their creepy feelings", "votes": null }, { "id": "609144", "postDate": "08/27/2019 12:47:33", "content": "

There are no advantages to submit just before the deadline. Considering the risk of incidents such as power failure, server malfunction, participants should submit final submission as soon as possible.

", "rawMarkdown": "There are no advantages to submit just before the deadline. Considering the risk of incidents such as power failure, server malfunction, participants should submit final submission as soon as possible.", "votes": null }, { "id": "609184", "postDate": "08/27/2019 13:26:17", "content": "

You forgot: Kaggle notebooks malfunction ;)

", "rawMarkdown": "You forgot: Kaggle notebooks malfunction ;)", "votes": null }, { "id": "609185", "postDate": "08/27/2019 13:27:19", "content": "

I don't see how you cen be better than -5-6 given this is the uncertainty in the data we have. You must be leaking target value somehow.

", "rawMarkdown": "I don't see how you cen be better than -5-6 given this is the uncertainty in the data we have. You must be leaking target value somehow.", "votes": null }, { "id": "609192", "postDate": "08/27/2019 13:39:05", "content": "

It's just that I am jealous ;)

", "rawMarkdown": "It's just that I am jealous ;)", "votes": null }, { "id": "609439", "postDate": "08/27/2019 18:04:19", "content": "
\n

At this point I am more intrigued about private vs public LB shuffle.\n How big can it be in terms of LMAE? 0.01 0.001?

\n
\n\n

You can approximate it. Just sample test set size rows from your OOF predictions. Then take 5% 95% percentiles. The difference should say how much it can shuffle at most.

", "rawMarkdown": "> At this point I am more intrigued about private vs public LB shuffle.\nHow big can it be in terms of LMAE? 0.01 0.001?\n\nYou can approximate it. Just sample test set size rows from your OOF predictions. Then take 5% 95% percentiles. The difference should say how much it can shuffle at most.", "votes": null }, { "id": "609468", "postDate": "08/27/2019 18:44:33", "content": "

I think you missed the smile :D

", "rawMarkdown": "I think you missed the smile :D", "votes": null }, { "id": "609474", "postDate": "08/27/2019 18:59:46", "content": "

I did ;)

", "rawMarkdown": "I did ;)", "votes": null }, { "id": "609511", "postDate": "08/27/2019 19:46:09", "content": "

On Instacart, the 1st place came along with his winning submission in the last five minutes of the competition.

\n\n

On Mercari, @grjhuard got ahead us with his single submission in the last hour ( And this was kernel competition , hence with risk of malfunction)

", "rawMarkdown": "On Instacart, the 1st place came along with his winning submission in the last five minutes of the competition. \n\nOn Mercari, @grjhuard got ahead us with his single submission in the last hour ( And this was kernel competition , hence with risk of malfunction)", "votes": null }, { "id": "609664", "postDate": "08/28/2019 01:28:29", "content": "

surprised!

\n\n

\"\"

", "rawMarkdown": "surprised!\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F486218%2F612988b0054b220c2091caa1e719e4e5%2FQQ20190828-0927062x.png?generation=1566955661485896&alt=media)", "votes": null }, { "id": "609699", "postDate": "08/28/2019 02:37:22", "content": "

The most surprised would be hybrid team. A week ago they were out of gold zone as I remember...

", "rawMarkdown": "The most surprised would be hybrid team. A week ago they were out of gold zone as I remember...", "votes": null }, { "id": "609743", "postDate": "08/28/2019 04:10:43", "content": "

You can check this dicussion, https://www.kaggle.com/c/champs-scalar-coupling/discussion/106054#latest-609719. You will find that they are domain experts. :)

", "rawMarkdown": "You can check this dicussion, https://www.kaggle.com/c/champs-scalar-coupling/discussion/106054#latest-609719. You will find that they are domain experts. :)", "votes": null }, { "id": "609830", "postDate": "08/28/2019 06:24:30", "content": "

@cpmpml: You are correct, I found that I'm leaking the target. But still, I got the score on based on your assumption.

", "rawMarkdown": "cpmpml: You are correct, I found that I'm leaking the target. But still, I got the score on based on your assumption.", "votes": null }, { "id": "609931", "postDate": "08/28/2019 08:43:26", "content": "

Last minute surprised myself by adding -.1 to my single model...

", "rawMarkdown": "Last minute surprised myself by adding -.1 to my single model...", "votes": null }, { "id": "609971", "postDate": "08/28/2019 09:21:35", "content": "

If I was a cynic I would note that inversion posted a comment about shakeup in one of his other competitions! ;)

", "rawMarkdown": "If I was a cynic I would note that inversion posted a comment about shakeup in one of his other competitions! ;)", "votes": null }, { "id": "609984", "postDate": "08/28/2019 09:45:57", "content": "

Probably ;)

", "rawMarkdown": "Probably ;)", "votes": null }, { "id": "610090", "postDate": "08/28/2019 12:07:30", "content": "

good boost)

", "rawMarkdown": "good boost)", "votes": null }, { "id": "610094", "postDate": "08/28/2019 12:14:35", "content": "

Link?

", "rawMarkdown": "Link?", "votes": null }, { "id": "610134", "postDate": "08/28/2019 13:25:06", "content": "

With no submissions left :)

", "rawMarkdown": "With no submissions left :)", "votes": null }, { "id": "610556", "postDate": "08/28/2019 21:34:25", "content": "

Is the surprise happening? Waiting for the write-up after competition.

", "rawMarkdown": "Is the surprise happening? Waiting for the write-up after competition.", "votes": null }, { "id": "612355", "postDate": "08/29/2019 18:34:36", "content": "

agreed

", "rawMarkdown": "agreed", "votes": null } ], "comments": [ { "id": 608960, "author_name": "senkin13", "author_url": "", "post_date": "08/27/2019 09:40:51", "content": "

Don’t do that,the submission maybe failed as some kinds of reasons.

", "votes": null, "replies": [] }, { "id": 609086, "author_name": "khahuras", "author_url": "", "post_date": "08/27/2019 12:03:28", "content": "

I think last minute surprise is probably probable in this kind of probably stable-CV-LB competition.

", "votes": null, "replies": [ { "id": 609984, "author_name": "scirpus", "author_url": "", "post_date": "08/28/2019 09:45:57", "content": "

Probably ;)

", "votes": null, "replies": [] } ] }, { "id": 609105, "author_name": "antorsae", "author_url": "", "post_date": "08/27/2019 12:17:17", "content": "

At this point I am more intrigued about private vs public LB shuffle.\nHow big can it be in terms of LMAE? 0.01 0.001?

", "votes": null, "replies": [ { "id": 609112, "author_name": "khahuras", "author_url": "", "post_date": "08/27/2019 12:20:37", "content": "

I predict that no more than 2 teams in top 20 will shuffle more than 3 places.

", "votes": null, "replies": [] }, { "id": 609116, "author_name": "cpmpml", "author_url": "", "post_date": "08/27/2019 12:27:01", "content": "
\n

I am more intrigued about private vs public LB shuffle.

\n
\n\n

Read: will our lead over 2nd team on public LB be enough to stay 1st in private LB? ;)

", "votes": null, "replies": [] }, { "id": 609123, "author_name": "antorsae", "author_url": "", "post_date": "08/27/2019 12:30:47", "content": "

Public LB is not frozen yet!

", "votes": null, "replies": [] }, { "id": 609126, "author_name": "khahuras", "author_url": "", "post_date": "08/27/2019 12:32:50", "content": "

@cpmpml please don't say it out loud like that ;) we all know their creepy feelings

", "votes": null, "replies": [] }, { "id": 609192, "author_name": "cpmpml", "author_url": "", "post_date": "08/27/2019 13:39:05", "content": "

It's just that I am jealous ;)

", "votes": null, "replies": [] }, { "id": 609439, "author_name": "aerdem4", "author_url": "", "post_date": "08/27/2019 18:04:19", "content": "
\n

At this point I am more intrigued about private vs public LB shuffle.\n How big can it be in terms of LMAE? 0.01 0.001?

\n
\n\n

You can approximate it. Just sample test set size rows from your OOF predictions. Then take 5% 95% percentiles. The difference should say how much it can shuffle at most.

", "votes": null, "replies": [] }, { "id": 612355, "author_name": "medoalmasry", "author_url": "", "post_date": "08/29/2019 18:34:36", "content": "

agreed

", "votes": null, "replies": [] } ] }, { "id": 609113, "author_name": "projdev", "author_url": "", "post_date": "08/27/2019 12:21:39", "content": "

I hope my last-minute surprise will work. Still struggling on the test set. My train & validation score in LGBM is almost -11 + with minimum variance, but I don't know these values error occurs in the test set. :-p

", "votes": null, "replies": [ { "id": 609185, "author_name": "cpmpml", "author_url": "", "post_date": "08/27/2019 13:27:19", "content": "

I don't see how you cen be better than -5-6 given this is the uncertainty in the data we have. You must be leaking target value somehow.

", "votes": null, "replies": [] }, { "id": 609468, "author_name": "somnambwl", "author_url": "", "post_date": "08/27/2019 18:44:33", "content": "

I think you missed the smile :D

", "votes": null, "replies": [] }, { "id": 609474, "author_name": "cpmpml", "author_url": "", "post_date": "08/27/2019 18:59:46", "content": "

I did ;)

", "votes": null, "replies": [] }, { "id": 609830, "author_name": "projdev", "author_url": "", "post_date": "08/28/2019 06:24:30", "content": "

@cpmpml: You are correct, I found that I'm leaking the target. But still, I got the score on based on your assumption.

", "votes": null, "replies": [] } ] }, { "id": 609120, "author_name": "artgor", "author_url": "", "post_date": "08/27/2019 12:29:43", "content": "

I really hope that nothing spectacular happens ;)

", "votes": null, "replies": [] }, { "id": 609144, "author_name": "toshik", "author_url": "", "post_date": "08/27/2019 12:47:33", "content": "

There are no advantages to submit just before the deadline. Considering the risk of incidents such as power failure, server malfunction, participants should submit final submission as soon as possible.

", "votes": null, "replies": [ { "id": 609184, "author_name": "cpmpml", "author_url": "", "post_date": "08/27/2019 13:26:17", "content": "

You forgot: Kaggle notebooks malfunction ;)

", "votes": null, "replies": [] }, { "id": 609511, "author_name": "serigne", "author_url": "", "post_date": "08/27/2019 19:46:09", "content": "

On Instacart, the 1st place came along with his winning submission in the last five minutes of the competition.

\n\n

On Mercari, @grjhuard got ahead us with his single submission in the last hour ( And this was kernel competition , hence with risk of malfunction)

", "votes": null, "replies": [] } ] }, { "id": 609664, "author_name": "hadxu123", "author_url": "", "post_date": "08/28/2019 01:28:29", "content": "

surprised!

\n\n

\"\"

", "votes": null, "replies": [ { "id": 609743, "author_name": "youhanlee", "author_url": "", "post_date": "08/28/2019 04:10:43", "content": "

You can check this dicussion, https://www.kaggle.com/c/champs-scalar-coupling/discussion/106054#latest-609719. You will find that they are domain experts. :)

", "votes": null, "replies": [] } ] }, { "id": 609699, "author_name": "khahuras", "author_url": "", "post_date": "08/28/2019 02:37:22", "content": "

The most surprised would be hybrid team. A week ago they were out of gold zone as I remember...

", "votes": null, "replies": [] }, { "id": 609931, "author_name": "julianmukaj", "author_url": "", "post_date": "08/28/2019 08:43:26", "content": "

Last minute surprised myself by adding -.1 to my single model...

", "votes": null, "replies": [ { "id": 610090, "author_name": "jingobox", "author_url": "", "post_date": "08/28/2019 12:07:30", "content": "

good boost)

", "votes": null, "replies": [] }, { "id": 610134, "author_name": "julianmukaj", "author_url": "", "post_date": "08/28/2019 13:25:06", "content": "

With no submissions left :)

", "votes": null, "replies": [] } ] }, { "id": 609971, "author_name": "scirpus", "author_url": "", "post_date": "08/28/2019 09:21:35", "content": "

If I was a cynic I would note that inversion posted a comment about shakeup in one of his other competitions! ;)

", "votes": null, "replies": [ { "id": 610094, "author_name": "cpmpml", "author_url": "", "post_date": "08/28/2019 12:14:35", "content": "

Link?

", "votes": null, "replies": [] } ] }, { "id": 610556, "author_name": "vithng", "author_url": "", "post_date": "08/28/2019 21:34:25", "content": "

Is the surprise happening? Waiting for the write-up after competition.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "608948": "Some teams that were near the top when they submitted, haven't submitted for a while, like Tony Y., DL Guys, and Jaechang Lim. Will these jump to prize zone at the last minute?\n\nI hope not ;)", "608960": "Don’t do that,the submission maybe failed as some kinds of reasons.", "609086": "I think last minute surprise is probably probable in this kind of probably stable-CV-LB competition.", "609105": "At this point I am more intrigued about private vs public LB shuffle.\nHow big can it be in terms of LMAE? 0.01 0.001?", "609112": "I predict that no more than 2 teams in top 20 will shuffle more than 3 places.", "609113": "I hope my last-minute surprise will work. Still struggling on the test set. My train & validation score in LGBM is almost -11 + with minimum variance, but I don't know these values error occurs in the test set. :-p", "609116": "> I am more intrigued about private vs public LB shuffle.\n\nRead: will our lead over 2nd team on public LB be enough to stay 1st in private LB? ;)", "609120": "I really hope that nothing spectacular happens ;)", "609123": "Public LB is not frozen yet!", "609126": "cpmpml please don't say it out loud like that ;) we all know their creepy feelings", "609144": "There are no advantages to submit just before the deadline. Considering the risk of incidents such as power failure, server malfunction, participants should submit final submission as soon as possible.", "609184": "You forgot: Kaggle notebooks malfunction ;)", "609185": "I don't see how you cen be better than -5-6 given this is the uncertainty in the data we have. You must be leaking target value somehow.", "609192": "It's just that I am jealous ;)", "609439": "> At this point I am more intrigued about private vs public LB shuffle.\nHow big can it be in terms of LMAE? 0.01 0.001?\n\nYou can approximate it. Just sample test set size rows from your OOF predictions. Then take 5% 95% percentiles. The difference should say how much it can shuffle at most.", "609468": "I think you missed the smile :D", "609474": "I did ;)", "609511": "On Instacart, the 1st place came along with his winning submission in the last five minutes of the competition. \n\nOn Mercari, @grjhuard got ahead us with his single submission in the last hour ( And this was kernel competition , hence with risk of malfunction)", "609664": "surprised!\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F486218%2F612988b0054b220c2091caa1e719e4e5%2FQQ20190828-0927062x.png?generation=1566955661485896&alt=media)", "609699": "The most surprised would be hybrid team. A week ago they were out of gold zone as I remember...", "609743": "You can check this dicussion, https://www.kaggle.com/c/champs-scalar-coupling/discussion/106054#latest-609719. You will find that they are domain experts. :)", "609830": "cpmpml: You are correct, I found that I'm leaking the target. But still, I got the score on based on your assumption.", "609931": "Last minute surprised myself by adding -.1 to my single model...", "609971": "If I was a cynic I would note that inversion posted a comment about shakeup in one of his other competitions! ;)", "609984": "Probably ;)", "610090": "good boost)", "610094": "Link?", "610134": "With no submissions left :)", "610556": "Is the surprise happening? Waiting for the write-up after competition.", "612355": "agreed" }, "source": "meta" }