{ "id": 599542, "title": "Best single model", "url": "/competitions/aeroclub-recsys-2025/discussion/599542", "author_name": "", "post_date": "2025-08-17T07:08:54.166497300Z", "votes": 6, "comment_count": 15, "views": 0, "content": "

Since there's a separate prize for best single model, I wanted to see what other folks implemented. For me the best single model was xgb (10 gkf mean) with private LB of 0.53807

\n

Not sure if that qualifies in top 5. 😐

", "messages": [ { "id": "3270704", "postDate": "08/17/2025 07:08:54", "content": "

Since there's a separate prize for best single model, I wanted to see what other folks implemented. For me the best single model was xgb (10 gkf mean) with private LB of 0.53807

\n

Not sure if that qualifies in top 5. 😐

", "rawMarkdown": "Since there's a separate prize for best single model, I wanted to see what other folks implemented. For me the best single model was xgb (10 gkf mean) with private LB of 0.53807 \n\nNot sure if that qualifies in top 5. 😐", "votes": null }, { "id": "3270707", "postDate": "08/17/2025 07:11:56", "content": "

our best is 0.53265, just 1 model trained without validation

", "rawMarkdown": "our best is 0.53265, just 1 model trained without validation", "votes": null }, { "id": "3270709", "postDate": "08/17/2025 07:13:48", "content": "

Super stuff! well done.

", "rawMarkdown": "Super stuff! well done.", "votes": null }, { "id": "3270711", "postDate": "08/17/2025 07:14:50", "content": "

a single xgb model train on all train data, public lb 0.52795 and private lb 0.53552

", "rawMarkdown": "a single xgb model train on all train data, public lb 0.52795 and private lb 0.53552", "votes": null }, { "id": "3270714", "postDate": "08/17/2025 07:16:58", "content": "

Super! In this competition, there's not a huge difference between single and ensembles.

", "rawMarkdown": "Super! In this competition, there's not a huge difference between single and ensembles.", "votes": null }, { "id": "3270715", "postDate": "08/17/2025 07:20:11", "content": "

I have 1 single model, 1 model train using all train data, but not sure if qualified for I did not submit using notebook, just submit csv file.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
ValidLBPB
5045406254156
", "rawMarkdown": "I have 1 single model, 1 model train using all train data, but not sure if qualified for I did not submit using notebook, just submit csv file.\n|Valid | LB | PB |\n|---| --- | --- |\n| 504 |54062 |54156|", "votes": null }, { "id": "3270716", "postDate": "08/17/2025 07:20:17", "content": "

My single model, that I cheked: 0.53056 on private:

\n

https://www.kaggle.com/code/mikhailgolubchik/sm-xgboost-single

", "rawMarkdown": "My single model, that I cheked: 0.53056 on private:\n\nhttps://www.kaggle.com/code/mikhailgolubchik/sm-xgboost-single", "votes": null }, { "id": "3270717", "postDate": "08/17/2025 07:20:39", "content": "

Agreed… my ensemble mostly just fits the public LB, so the gain over a single model is pretty minor😞

", "rawMarkdown": "Agreed... my ensemble mostly just fits the public LB, so the gain over a single model is pretty minor😞", "votes": null }, { "id": "3270720", "postDate": "08/17/2025 07:23:06", "content": "

Woah, don't think anyone's crossing this single model!

\n

I understood that you do not have to sub via notebook necessarily to qualify to single model prize. -> https://www.kaggle.com/competitions/aeroclub-recsys-2025/discussion/599183#3269711

", "rawMarkdown": "Woah, don't think anyone's crossing this single model! \n\nI understood that you do not have to sub via notebook necessarily to qualify to single model prize. -> https://www.kaggle.com/competitions/aeroclub-recsys-2025/discussion/599183#3269711", "votes": null }, { "id": "3270728", "postDate": "08/17/2025 07:41:46", "content": "

Yes my 41 models with optuna tune could not beat 5 top models wights 1 ensemble(late submit with PB 54725). Interestingly I asked gpt5 for sugeestion of last submission to choose 41 models optuna tune or 5 top models simple avg, gt5 suggested top5 models simple avg tough I still submitted 41 models ensemble result😂

", "rawMarkdown": "Yes my 41 models with optuna tune could not beat 5 top models wights 1 ensemble(late submit with PB 54725). Interestingly I asked gpt5 for sugeestion of last submission to choose 41 models optuna tune or 5 top models simple avg, gt5 suggested top5 models simple avg tough I still submitted 41 models ensemble result😂", "votes": null }, { "id": "3270733", "postDate": "08/17/2025 07:56:48", "content": "

My single XGBoost Model without any GKF: 0.51813,
\nPrivate Leaderboard score: 0.52544.

", "rawMarkdown": "My single XGBoost Model without any GKF: 0.51813, \nPrivate Leaderboard score: 0.52544.", "votes": null }, { "id": "3270757", "postDate": "08/17/2025 09:18:37", "content": "

hello, everyone. 0.52 privat score for single model LGBM. 0.527 for assemble of them with 2 sets of features\\parameters.

", "rawMarkdown": "hello, everyone. 0.52 privat score for single model LGBM. 0.527 for assemble of them with 2 sets of features\\parameters.", "votes": null }, { "id": "3270776", "postDate": "08/17/2025 10:10:11", "content": "

No worries @goldenlock. It will be considered. We understand memory limitations at the platform.

", "rawMarkdown": "No worries @goldenlock. It will be considered. We understand memory limitations at the platform.", "votes": null }, { "id": "3270792", "postDate": "08/17/2025 10:57:14", "content": "

My Best Single Model Score is 0.52098 for Public LB and 0.52823 for Private LB.
\nI used LightGBM training all data.

", "rawMarkdown": "My Best Single Model Score is 0.52098 for Public LB and 0.52823 for Private LB.\nI used LightGBM training all data.", "votes": null }, { "id": "3270993", "postDate": "08/17/2025 19:05:24", "content": "

@pheadrus btw your Single is 0.52777 (public) and 0.53807 (private)

", "rawMarkdown": "pheadrus btw your Single is 0.52777 (public) and 0.53807 (private)", "votes": null }, { "id": "3271099", "postDate": "08/18/2025 04:21:49", "content": "

yes, sorry for the confusion.

", "rawMarkdown": "yes, sorry for the confusion.", "votes": null } ], "comments": [ { "id": 3270707, "author_name": "insuperabile", "author_url": "", "post_date": "08/17/2025 07:11:56", "content": "

our best is 0.53265, just 1 model trained without validation

", "votes": null, "replies": [ { "id": 3270709, "author_name": "pheadrus", "author_url": "", "post_date": "08/17/2025 07:13:48", "content": "

Super stuff! well done.

", "votes": null, "replies": [] } ] }, { "id": 3270711, "author_name": "mango789", "author_url": "", "post_date": "08/17/2025 07:14:50", "content": "

a single xgb model train on all train data, public lb 0.52795 and private lb 0.53552

", "votes": null, "replies": [ { "id": 3270714, "author_name": "pheadrus", "author_url": "", "post_date": "08/17/2025 07:16:58", "content": "

Super! In this competition, there's not a huge difference between single and ensembles.

", "votes": null, "replies": [ { "id": 3270717, "author_name": "mango789", "author_url": "", "post_date": "08/17/2025 07:20:39", "content": "

Agreed… my ensemble mostly just fits the public LB, so the gain over a single model is pretty minor😞

", "votes": null, "replies": [ { "id": 3270728, "author_name": "goldenlock", "author_url": "", "post_date": "08/17/2025 07:41:46", "content": "

Yes my 41 models with optuna tune could not beat 5 top models wights 1 ensemble(late submit with PB 54725). Interestingly I asked gpt5 for sugeestion of last submission to choose 41 models optuna tune or 5 top models simple avg, gt5 suggested top5 models simple avg tough I still submitted 41 models ensemble result😂

", "votes": null, "replies": [] } ] } ] } ] }, { "id": 3270715, "author_name": "goldenlock", "author_url": "", "post_date": "08/17/2025 07:20:11", "content": "

I have 1 single model, 1 model train using all train data, but not sure if qualified for I did not submit using notebook, just submit csv file.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
ValidLBPB
5045406254156
", "votes": null, "replies": [ { "id": 3270720, "author_name": "pheadrus", "author_url": "", "post_date": "08/17/2025 07:23:06", "content": "

Woah, don't think anyone's crossing this single model!

\n

I understood that you do not have to sub via notebook necessarily to qualify to single model prize. -> https://www.kaggle.com/competitions/aeroclub-recsys-2025/discussion/599183#3269711

", "votes": null, "replies": [] }, { "id": 3270776, "author_name": "samvelkoch", "author_url": "", "post_date": "08/17/2025 10:10:11", "content": "

No worries @goldenlock. It will be considered. We understand memory limitations at the platform.

", "votes": null, "replies": [] } ] }, { "id": 3270716, "author_name": "mikhailgolubchik", "author_url": "", "post_date": "08/17/2025 07:20:17", "content": "

My single model, that I cheked: 0.53056 on private:

\n

https://www.kaggle.com/code/mikhailgolubchik/sm-xgboost-single

", "votes": null, "replies": [] }, { "id": 3270733, "author_name": "mhassaan1122", "author_url": "", "post_date": "08/17/2025 07:56:48", "content": "

My single XGBoost Model without any GKF: 0.51813,
\nPrivate Leaderboard score: 0.52544.

", "votes": null, "replies": [] }, { "id": 3270757, "author_name": "sergeyqt2024", "author_url": "", "post_date": "08/17/2025 09:18:37", "content": "

hello, everyone. 0.52 privat score for single model LGBM. 0.527 for assemble of them with 2 sets of features\\parameters.

", "votes": null, "replies": [] }, { "id": 3270792, "author_name": "xblade", "author_url": "", "post_date": "08/17/2025 10:57:14", "content": "

My Best Single Model Score is 0.52098 for Public LB and 0.52823 for Private LB.
\nI used LightGBM training all data.

", "votes": null, "replies": [] }, { "id": 3270993, "author_name": "samvelkoch", "author_url": "", "post_date": "08/17/2025 19:05:24", "content": "

@pheadrus btw your Single is 0.52777 (public) and 0.53807 (private)

", "votes": null, "replies": [ { "id": 3271099, "author_name": "pheadrus", "author_url": "", "post_date": "08/18/2025 04:21:49", "content": "

yes, sorry for the confusion.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3270704": "Since there's a separate prize for best single model, I wanted to see what other folks implemented. For me the best single model was xgb (10 gkf mean) with private LB of 0.53807 \n\nNot sure if that qualifies in top 5. 😐", "3270707": "our best is 0.53265, just 1 model trained without validation", "3270709": "Super stuff! well done.", "3270711": "a single xgb model train on all train data, public lb 0.52795 and private lb 0.53552", "3270714": "Super! In this competition, there's not a huge difference between single and ensembles.", "3270715": "I have 1 single model, 1 model train using all train data, but not sure if qualified for I did not submit using notebook, just submit csv file.\n|Valid | LB | PB |\n|---| --- | --- |\n| 504 |54062 |54156|", "3270716": "My single model, that I cheked: 0.53056 on private:\n\nhttps://www.kaggle.com/code/mikhailgolubchik/sm-xgboost-single", "3270717": "Agreed... my ensemble mostly just fits the public LB, so the gain over a single model is pretty minor😞", "3270720": "Woah, don't think anyone's crossing this single model! \n\nI understood that you do not have to sub via notebook necessarily to qualify to single model prize. -> https://www.kaggle.com/competitions/aeroclub-recsys-2025/discussion/599183#3269711", "3270728": "Yes my 41 models with optuna tune could not beat 5 top models wights 1 ensemble(late submit with PB 54725). Interestingly I asked gpt5 for sugeestion of last submission to choose 41 models optuna tune or 5 top models simple avg, gt5 suggested top5 models simple avg tough I still submitted 41 models ensemble result😂", "3270733": "My single XGBoost Model without any GKF: 0.51813, \nPrivate Leaderboard score: 0.52544.", "3270757": "hello, everyone. 0.52 privat score for single model LGBM. 0.527 for assemble of them with 2 sets of features\\parameters.", "3270776": "No worries @goldenlock. It will be considered. We understand memory limitations at the platform.", "3270792": "My Best Single Model Score is 0.52098 for Public LB and 0.52823 for Private LB.\nI used LightGBM training all data.", "3270993": "pheadrus btw your Single is 0.52777 (public) and 0.53807 (private)", "3271099": "yes, sorry for the confusion." }, "source": "meta" }