I find a notebook,https://www.kaggle.com/somayyehgholami/fork-of-lightgbm-catboost-ann-2505f2, which is public in the end of this competition, i think many team will get high score in PB in low cost. It is unfair for all team!
", "messages": [ { "id": 1394007, "postDate": "2021-07-20T05:22:26.413Z", "content": "Totally agree, but this has been a recurring situation in Kaggle competitions. People post high-scoring notebooks even much closer to the competition deadline. Also, not everything in public notebooks works on private LB, try to take from it whatever you can.
", "rawMarkdown": "Totally agree, but this has been a recurring situation in Kaggle competitions. People post high-scoring notebooks even much closer to the competition deadline. Also, not everything in public notebooks works on private LB, try to take from it whatever you can.", "votes": 5 }, { "id": 1393944, "postDate": "2021-07-20T03:45:05.883Z", "content": "I find a notebook,https://www.kaggle.com/somayyehgholami/fork-of-lightgbm-catboost-ann-2505f2, which is public in the end of this competition, i think many team will get high score in PB in low cost. It is unfair for all team!
", "rawMarkdown": " I find a notebook,https://www.kaggle.com/somayyehgholami/fork-of-lightgbm-catboost-ann-2505f2, which is public in the end of this competition, i think many team will get high score in PB in low cost. It is unfair for all team!", "votes": 2 }, { "id": 1394610, "postDate": "2021-07-20T13:13:34.850Z", "content": "Honestly, I don't think this is a problem. That notebook is bronze in the public LB, but I expect a big shake-up in the private LB.
\nActually, there are several variations on the same theme available in the public notebooks, differing largely by some tweaks to their parameters. For what it's worth, I think they are very good codes.
\nIt's easy to improve one's public score beyond this by playing with the parameters a little more, for example I got 1.3359 with minimal extra effort. It's obvious that many people will see this as overfitting. I certainly don't think that a difference of 0.001 in public LB has any relevance to the final evaluation scores.
", "rawMarkdown": "Honestly, I don't think this is a problem. That notebook is bronze in the public LB, but I expect a big shake-up in the private LB.\n\nActually, there are several variations on the same theme available in the public notebooks, differing largely by some tweaks to their parameters. For what it's worth, I think they are very good codes. \n\nIt's easy to improve one's public score beyond this by playing with the parameters a little more, for example I got 1.3359 with minimal extra effort. It's obvious that many people will see this as overfitting. I certainly don't think that a difference of 0.001 in public LB has any relevance to the final evaluation scores.", "votes": 3 }, { "id": 1394844, "postDate": "2021-07-20T17:04:32.553Z", "content": "I think that notebook was basically tweaking some parameters to overfit the public LB
", "rawMarkdown": "I think that notebook was basically tweaking some parameters to overfit the public LB ", "votes": 2 }, { "id": 1394136, "postDate": "2021-07-20T07:42:18.700Z", "rawMarkdown": "", "votes": -2, "isDeleted": true } ], "comments": [ { "id": 1394007, "author_name": "Chinta", "author_url": "", "post_date": "2021-07-20T05:22:26.413000", "content": "Totally agree, but this has been a recurring situation in Kaggle competitions. People post high-scoring notebooks even much closer to the competition deadline. Also, not everything in public notebooks works on private LB, try to take from it whatever you can.
", "votes": 5, "replies": [] }, { "id": 1394610, "author_name": "John Mitchell", "author_url": "", "post_date": "2021-07-20T13:13:34.850000", "content": "Honestly, I don't think this is a problem. That notebook is bronze in the public LB, but I expect a big shake-up in the private LB.
\nActually, there are several variations on the same theme available in the public notebooks, differing largely by some tweaks to their parameters. For what it's worth, I think they are very good codes.
\nIt's easy to improve one's public score beyond this by playing with the parameters a little more, for example I got 1.3359 with minimal extra effort. It's obvious that many people will see this as overfitting. I certainly don't think that a difference of 0.001 in public LB has any relevance to the final evaluation scores.
", "votes": 3, "replies": [] }, { "id": 1394844, "author_name": "Lu Bin Liu", "author_url": "", "post_date": "2021-07-20T17:04:32.553000", "content": "I think that notebook was basically tweaking some parameters to overfit the public LB
", "votes": 2, "replies": [] }, { "id": 1394136, "author_name": "", "author_url": "", "post_date": "2021-07-20T07:42:18.700000", "content": "", "votes": -2, "replies": [] } ], "raw_markdown_by_id": { "1394007": "Totally agree, but this has been a recurring situation in Kaggle competitions. People post high-scoring notebooks even much closer to the competition deadline. Also, not everything in public notebooks works on private LB, try to take from it whatever you can.", "1393944": " I find a notebook,https://www.kaggle.com/somayyehgholami/fork-of-lightgbm-catboost-ann-2505f2, which is public in the end of this competition, i think many team will get high score in PB in low cost. It is unfair for all team!", "1394610": "Honestly, I don't think this is a problem. That notebook is bronze in the public LB, but I expect a big shake-up in the private LB.\n\nActually, there are several variations on the same theme available in the public notebooks, differing largely by some tweaks to their parameters. For what it's worth, I think they are very good codes. \n\nIt's easy to improve one's public score beyond this by playing with the parameters a little more, for example I got 1.3359 with minimal extra effort. It's obvious that many people will see this as overfitting. I certainly don't think that a difference of 0.001 in public LB has any relevance to the final evaluation scores.", "1394844": "I think that notebook was basically tweaking some parameters to overfit the public LB ", "1394136": "" } }