{ "id": 327765, "title": "GBDT or NN,which is the winner of this competition", "url": "/competitions/amex-default-prediction/discussion/327765", "author_name": "", "post_date": "2022-05-29T03:58:07.571713500Z", "votes": 25, "comment_count": 9, "views": 0, "content": "

In most tabular data competition, GBDT has better performace than nn but this competition's data is anonymized, big size, include sequence data,I think nn also has opportunity to challenge.I had a poll in twitter, about 70% of kagglers think GBDT will win,I expect to see the final winning solutions.

\n

\"Screen-Shot-2022-05-29-at-12-10-17.png\"

", "messages": [ { "id": "1804481", "postDate": "05/29/2022 03:58:07", "content": "

In most tabular data competition, GBDT has better performace than nn but this competition's data is anonymized, big size, include sequence data,I think nn also has opportunity to challenge.I had a poll in twitter, about 70% of kagglers think GBDT will win,I expect to see the final winning solutions.

\n

\"Screen-Shot-2022-05-29-at-12-10-17.png\"

", "rawMarkdown": "In most tabular data competition, GBDT has better performace than nn but this competition's data is anonymized, big size, include sequence data,I think nn also has opportunity to challenge.I had a poll in twitter, about 70% of kagglers think GBDT will win,I expect to see the final winning solutions.\n\n[![Screen-Shot-2022-05-29-at-12-10-17.png](https://i.postimg.cc/XqXgDtzZ/Screen-Shot-2022-05-29-at-12-10-17.png)](https://postimg.cc/PvsZCVhh)", "votes": null }, { "id": "1804618", "postDate": "05/29/2022 08:53:06", "content": "

the answer is obvious that an ensemble of both will win :)

", "rawMarkdown": "the answer is obvious that an ensemble of both will win :)", "votes": null }, { "id": "1804628", "postDate": "05/29/2022 09:20:12", "content": "

Haha, I think so, the large and anonymized data competition will also become NN's stage.

", "rawMarkdown": "Haha, I think so, the large and anonymized data competition will also become NN's stage.", "votes": null }, { "id": "1804785", "postDate": "05/29/2022 12:59:58", "content": "

The large amount of data leaves the door wide open for either to win. IMO you can never go wrong with decision trees. Past competitions have shown that decision trees are the main models used to win.

", "rawMarkdown": "The large amount of data leaves the door wide open for either to win. IMO you can never go wrong with decision trees. Past competitions have shown that decision trees are the main models used to win.", "votes": null }, { "id": "1804799", "postDate": "05/29/2022 13:20:48", "content": "

Rooting for the gbdt. But from the recent tabular data litterature it seems NN architectures can better squeeze the last 0.002 needed for the win.

", "rawMarkdown": "Rooting for the gbdt. But from the recent tabular data litterature it seems NN architectures can better squeeze the last 0.002 needed for the win.", "votes": null }, { "id": "1804813", "postDate": "05/29/2022 13:27:06", "content": "

ensemble of gbdt and nn maybe win, but I expect to see very strong single model of gbdt or nn that ensemble doesn't help

", "rawMarkdown": "ensemble of gbdt and nn maybe win, but I expect to see very strong single model of gbdt or nn that ensemble doesn't help", "votes": null }, { "id": "1804848", "postDate": "05/29/2022 13:54:47", "content": "

When would ensemble doesn't help, or why?

", "rawMarkdown": "When would ensemble doesn't help, or why?", "votes": null }, { "id": "1804857", "postDate": "05/29/2022 13:59:37", "content": "

Make Gbdt Great Again!😅

", "rawMarkdown": "Make Gbdt Great Again!😅", "votes": null }, { "id": "1804869", "postDate": "05/29/2022 14:07:09", "content": "

XGB is all you need 😂

", "rawMarkdown": "XGB is all you need 😂", "votes": null }, { "id": "1806299", "postDate": "05/31/2022 02:58:18", "content": "

nn for the win!

", "rawMarkdown": "nn for the win!", "votes": null } ], "comments": [ { "id": 1804618, "author_name": "raddar", "author_url": "", "post_date": "05/29/2022 08:53:06", "content": "

the answer is obvious that an ensemble of both will win :)

", "votes": null, "replies": [ { "id": 1804628, "author_name": "librauee", "author_url": "", "post_date": "05/29/2022 09:20:12", "content": "

Haha, I think so, the large and anonymized data competition will also become NN's stage.

", "votes": null, "replies": [] }, { "id": 1804813, "author_name": "senkin13", "author_url": "", "post_date": "05/29/2022 13:27:06", "content": "

ensemble of gbdt and nn maybe win, but I expect to see very strong single model of gbdt or nn that ensemble doesn't help

", "votes": null, "replies": [] }, { "id": 1804848, "author_name": "celiker", "author_url": "", "post_date": "05/29/2022 13:54:47", "content": "

When would ensemble doesn't help, or why?

", "votes": null, "replies": [] } ] }, { "id": 1804785, "author_name": "wuuthraad", "author_url": "", "post_date": "05/29/2022 12:59:58", "content": "

The large amount of data leaves the door wide open for either to win. IMO you can never go wrong with decision trees. Past competitions have shown that decision trees are the main models used to win.

", "votes": null, "replies": [] }, { "id": 1804799, "author_name": "lucasmorin", "author_url": "", "post_date": "05/29/2022 13:20:48", "content": "

Rooting for the gbdt. But from the recent tabular data litterature it seems NN architectures can better squeeze the last 0.002 needed for the win.

", "votes": null, "replies": [] }, { "id": 1804857, "author_name": "hydantess", "author_url": "", "post_date": "05/29/2022 13:59:37", "content": "

Make Gbdt Great Again!😅

", "votes": null, "replies": [] }, { "id": 1804869, "author_name": "mahluo", "author_url": "", "post_date": "05/29/2022 14:07:09", "content": "

XGB is all you need 😂

", "votes": null, "replies": [] }, { "id": 1806299, "author_name": "adewoleakorede", "author_url": "", "post_date": "05/31/2022 02:58:18", "content": "

nn for the win!

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1804481": "In most tabular data competition, GBDT has better performace than nn but this competition's data is anonymized, big size, include sequence data,I think nn also has opportunity to challenge.I had a poll in twitter, about 70% of kagglers think GBDT will win,I expect to see the final winning solutions.\n\n[![Screen-Shot-2022-05-29-at-12-10-17.png](https://i.postimg.cc/XqXgDtzZ/Screen-Shot-2022-05-29-at-12-10-17.png)](https://postimg.cc/PvsZCVhh)", "1804618": "the answer is obvious that an ensemble of both will win :)", "1804628": "Haha, I think so, the large and anonymized data competition will also become NN's stage.", "1804785": "The large amount of data leaves the door wide open for either to win. IMO you can never go wrong with decision trees. Past competitions have shown that decision trees are the main models used to win.", "1804799": "Rooting for the gbdt. But from the recent tabular data litterature it seems NN architectures can better squeeze the last 0.002 needed for the win.", "1804813": "ensemble of gbdt and nn maybe win, but I expect to see very strong single model of gbdt or nn that ensemble doesn't help", "1804848": "When would ensemble doesn't help, or why?", "1804857": "Make Gbdt Great Again!😅", "1804869": "XGB is all you need 😂", "1806299": "nn for the win!" }, "source": "meta" }