![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F16721372%2F8ae3c9c13447bae9aff8d4364d2220b7%2FCMI.png?generation=1734653824925389&alt=media\")
First of all, congratulations to the winners! It is very difficult to choose the final model, as each one behaves randomly.
\nI put a lot of work into this competition, but unfortunately I didn't get any medal here.
\nOne of my solution ranked between 61 and 69. I could have earned a medal, but it was hard for me to choose the final submission. It's all part of the competition.
Well I learn a lot from this and i am happy with this also.
\nHere the notebook Link : https://www.kaggle.com/code/abdmental01/optuna-optimizer
", "messages": [ { "id": "3076384", "postDate": "12/20/2024 00:22:10", "content": "First of all, congratulations to the winners! It is very difficult to choose the final model, as each one behaves randomly.
\nI put a lot of work into this competition, but unfortunately I didn't get any medal here.
\nOne of my solution ranked between 61 and 69. I could have earned a medal, but it was hard for me to choose the final submission. It's all part of the competition.
Well I learn a lot from this and i am happy with this also.
\nHere the notebook Link : https://www.kaggle.com/code/abdmental01/optuna-optimizer
", "rawMarkdown": "First of all, congratulations to the winners! It is very difficult to choose the final model, as each one behaves randomly.\n\nI put a lot of work into this competition, but unfortunately I didn't get any medal here.\nOne of my solution ranked between 61 and 69. I could have earned a medal, but it was hard for me to choose the final submission. It's all part of the competition.\n\n\n\nWell I learn a lot from this and i am happy with this also.\n\nHere the notebook Link : https://www.kaggle.com/code/abdmental01/optuna-optimizer", "votes": null }, { "id": "3076388", "postDate": "12/20/2024 00:28:54", "content": "You shouldn't have pick and just let the system automatically choose the best models for you.
", "rawMarkdown": "You shouldn't have pick and just let the system automatically choose the best models for you.", "votes": null }, { "id": "3076391", "postDate": "12/20/2024 00:29:57", "content": "The Kaggle auto system pick the best public score. and this model scores bad in public
", "rawMarkdown": "The Kaggle auto system pick the best public score. and this model scores bad in public", "votes": null }, { "id": "3076393", "postDate": "12/20/2024 00:32:39", "content": "It's important to identify 'lottery competition' early and avoid investing too much time. Well if you do it for learning it's another story. Then again, learning happens also in 'non-loterry' competitionsβ¦
", "rawMarkdown": "It's important to identify 'lottery competition' early and avoid investing too much time. Well if you do it for learning it's another story. Then again, learning happens also in 'non-loterry' competitions...", "votes": null }, { "id": "3076394", "postDate": "12/20/2024 00:32:53", "content": "Sorry man. We go again next time.
", "rawMarkdown": "Sorry man. We go again next time.", "votes": null }, { "id": "3076396", "postDate": "12/20/2024 00:33:21", "content": "First of all, congratulations to all the winners! ππ
\nThis has happened to all of us before. I'm sure in the next few hours, we'll hear about Kagglers whose notebooks scored above 0.50 but were not selected.
\nThe shake-up in the competition was massive, and it highlights how poor the choice of the metric was. π π€
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F17372370%2F5c3b814b30f2c29977a087b2a74fb148%2FCapture-compit.JPG?generation=1734654387524102&alt=media\")
Interestingly, one could get an easy silver medal just by forking this notebook
", "rawMarkdown": "Interestingly, one could get an easy silver medal just by forking [this](https://www.kaggle.com/code/yekenot/cmi-piu-deeptables-nn-cv-0-482) notebook", "votes": null }, { "id": "3076398", "postDate": "12/20/2024 00:34:23", "content": "The shake up was too much.
", "rawMarkdown": "The shake up was too much.", "votes": null }, { "id": "3076400", "postDate": "12/20/2024 00:36:17", "content": "Yes, but this is very hard to select the good model. total luck
", "rawMarkdown": "Yes, but this is very hard to select the good model. total luck", "votes": null }, { "id": "3076402", "postDate": "12/20/2024 00:37:46", "content": "I don't think this is a metric issue, but data distribution.
", "rawMarkdown": "I don't think this is a metric issue, but data distribution.", "votes": null }, { "id": "3076404", "postDate": "12/20/2024 00:37:47", "content": "I don't have experience with lottery competitions, and I will never forget this Shakeup π
", "rawMarkdown": "I don't have experience with lottery competitions, and I will never forget this Shakeup π", "votes": null }, { "id": "3076406", "postDate": "12/20/2024 00:38:57", "content": "I am in the same situation. ππ
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F21681918%2Fd3921733597c82f76f1bdcf59830a173%2Fkaggle_re.png?generation=1734655128036546&alt=media\")
\nI could get a bronze medal if I selected better model.
\nAre there any good way to find the better model ?
That's true too
", "rawMarkdown": "That's true too", "votes": null }, { "id": "3076408", "postDate": "12/20/2024 00:39:19", "content": "Yes you are right, in the silver Solution model, i just use different seed values for different Splits and it works. I read this in a thread and apply it.
", "rawMarkdown": "Yes you are right, in the silver Solution model, i just use different seed values for different Splits and it works. I read this in a thread and apply it.", "votes": null }, { "id": "3076410", "postDate": "12/20/2024 00:40:00", "content": "If there a way we also select the good models.
", "rawMarkdown": "If there a way we also select the good models.", "votes": null } ], "comments": [ { "id": 3076388, "author_name": "assignmeai", "author_url": "", "post_date": "12/20/2024 00:28:54", "content": "You shouldn't have pick and just let the system automatically choose the best models for you.
", "votes": null, "replies": [ { "id": 3076391, "author_name": "abdmental01", "author_url": "", "post_date": "12/20/2024 00:29:57", "content": "The Kaggle auto system pick the best public score. and this model scores bad in public
", "votes": null, "replies": [ { "id": 3076394, "author_name": "assignmeai", "author_url": "", "post_date": "12/20/2024 00:32:53", "content": "Sorry man. We go again next time.
", "votes": null, "replies": [] } ] } ] }, { "id": 3076393, "author_name": "shlomoron", "author_url": "", "post_date": "12/20/2024 00:32:39", "content": "It's important to identify 'lottery competition' early and avoid investing too much time. Well if you do it for learning it's another story. Then again, learning happens also in 'non-loterry' competitionsβ¦
", "votes": null, "replies": [ { "id": 3076398, "author_name": "assignmeai", "author_url": "", "post_date": "12/20/2024 00:34:23", "content": "The shake up was too much.
", "votes": null, "replies": [] }, { "id": 3076404, "author_name": "abdmental01", "author_url": "", "post_date": "12/20/2024 00:37:47", "content": "I don't have experience with lottery competitions, and I will never forget this Shakeup π
", "votes": null, "replies": [] } ] }, { "id": 3076396, "author_name": "rabieelkharoua", "author_url": "", "post_date": "12/20/2024 00:33:21", "content": "First of all, congratulations to all the winners! ππ
\nThis has happened to all of us before. I'm sure in the next few hours, we'll hear about Kagglers whose notebooks scored above 0.50 but were not selected.
\nThe shake-up in the competition was massive, and it highlights how poor the choice of the metric was. π π€
\n
I don't think this is a metric issue, but data distribution.
", "votes": null, "replies": [ { "id": 3076407, "author_name": "rabieelkharoua", "author_url": "", "post_date": "12/20/2024 00:38:58", "content": "That's true too
", "votes": null, "replies": [] }, { "id": 3076408, "author_name": "abdmental01", "author_url": "", "post_date": "12/20/2024 00:39:19", "content": "Yes you are right, in the silver Solution model, i just use different seed values for different Splits and it works. I read this in a thread and apply it.
", "votes": null, "replies": [] } ] } ] }, { "id": 3076397, "author_name": "yeoyunsianggeremie", "author_url": "", "post_date": "12/20/2024 00:33:46", "content": "Interestingly, one could get an easy silver medal just by forking this notebook
", "votes": null, "replies": [ { "id": 3076400, "author_name": "abdmental01", "author_url": "", "post_date": "12/20/2024 00:36:17", "content": "Yes, but this is very hard to select the good model. total luck
", "votes": null, "replies": [] } ] }, { "id": 3076406, "author_name": "kohdaiinagaki", "author_url": "", "post_date": "12/20/2024 00:38:57", "content": "I am in the same situation. ππ
\n
\nI could get a bronze medal if I selected better model.
\nAre there any good way to find the better model ?
If there a way we also select the good models.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3076384": "First of all, congratulations to the winners! It is very difficult to choose the final model, as each one behaves randomly.\n\nI put a lot of work into this competition, but unfortunately I didn't get any medal here.\nOne of my solution ranked between 61 and 69. I could have earned a medal, but it was hard for me to choose the final submission. It's all part of the competition.\n\n\n\nWell I learn a lot from this and i am happy with this also.\n\nHere the notebook Link : https://www.kaggle.com/code/abdmental01/optuna-optimizer", "3076388": "You shouldn't have pick and just let the system automatically choose the best models for you.", "3076391": "The Kaggle auto system pick the best public score. and this model scores bad in public", "3076393": "It's important to identify 'lottery competition' early and avoid investing too much time. Well if you do it for learning it's another story. Then again, learning happens also in 'non-loterry' competitions...", "3076394": "Sorry man. We go again next time.", "3076396": "First of all, congratulations to all the winners! ππ\n\nThis has happened to all of us before. I'm sure in the next few hours, we'll hear about Kagglers whose notebooks scored above 0.50 but were not selected.\n\nThe shake-up in the competition was massive, and it highlights how poor the choice of the metric was. π π€\n\n\n", "3076397": "Interestingly, one could get an easy silver medal just by forking [this](https://www.kaggle.com/code/yekenot/cmi-piu-deeptables-nn-cv-0-482) notebook", "3076398": "The shake up was too much.", "3076400": "Yes, but this is very hard to select the good model. total luck", "3076402": "I don't think this is a metric issue, but data distribution.", "3076404": "I don't have experience with lottery competitions, and I will never forget this Shakeup π", "3076406": "I am in the same situation. ππ\n\nI could get a bronze medal if I selected better model.\nAre there any good way to find the better model ?", "3076407": "That's true too", "3076408": "Yes you are right, in the silver Solution model, i just use different seed values for different Splits and it works. I read this in a thread and apply it.", "3076410": "If there a way we also select the good models." }, "source": "meta" }