{ "id": 167573, "title": "Public High Accuracy Notebooks is not good", "url": "/competitions/siim-isic-melanoma-classification/discussion/167573", "author_name": "Manh Lab", "post_date": "2020-07-17T05:49:49.746000", "votes": 17, "comment_count": 17, "views": 0, "content": "

I think many people like me, try hard on this competition, and have > 200 ratings. But if anyone public high LB notebooks, the LB is changed. That's terible!! You can give the method, not the SUBMISSION.CSV

", "messages": [ { "id": 932522, "postDate": "2020-07-17T05:49:49.747Z", "content": "

I think many people like me, try hard on this competition, and have > 200 ratings. But if anyone public high LB notebooks, the LB is changed. That's terible!! You can give the method, not the SUBMISSION.CSV

", "rawMarkdown": "I think many people like me, try hard on this competition, and have > 200 ratings. But if anyone public high LB notebooks, the LB is changed. That's terible!! You can give the method, not the SUBMISSION.CSV", "votes": 15 }, { "id": 934287, "postDate": "2020-07-18T11:02:04.210Z", "content": "

Kaggle should allow us to downvote kernels as well. This might discourage people of sharing blend of blends and inference kernels with a very high score.

", "rawMarkdown": "Kaggle should allow us to downvote kernels as well. This might discourage people of sharing blend of blends and inference kernels with a very high score.", "votes": 6, "replies": [ { "id": 934327, "postDate": "2020-07-18T11:35:32.227Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ] }, { "id": 934241, "postDate": "2020-07-18T10:35:05.190Z", "content": "

Yep definitely agree it's frustrating some times, I have been working a lot to get over the 0.94 baseline and when I did I saw that there's a public notebook that gives you better score by just ensembling a bunch of other submissions. I always remind myself that anyone who uses those notebooks to get higher lb won't get too far, he'll be stuck so it's one more reason to overtake them. I also agree with you the method/ indication should be given not the submission file. Good luck

", "rawMarkdown": "Yep definitely agree it's frustrating some times, I have been working a lot to get over the 0.94 baseline and when I did I saw that there's a public notebook that gives you better score by just ensembling a bunch of other submissions. I always remind myself that anyone who uses those notebooks to get higher lb won't get too far, he'll be stuck so it's one more reason to overtake them. I also agree with you the method/ indication should be given not the submission file. Good luck", "votes": 3, "replies": [ { "id": 934381, "postDate": "2020-07-18T12:02:58.083Z", "content": "

I see. I was working very much to reach 95.15 and POW! new notebook with higher LB and I down my position very quick

", "rawMarkdown": "I see. I was working very much to reach 95.15 and POW! new notebook with higher LB and I down my position very quick", "votes": 1 }, { "id": 934391, "postDate": "2020-07-18T12:07:46.043Z", "rawMarkdown": "", "isDeleted": true }, { "id": 934402, "postDate": "2020-07-18T12:13:38.367Z", "content": "

I only reach ~70 today. And i don't use that kernel to reach that position. Don't know why you think that!

", "rawMarkdown": "I only reach ~70 today. And i don't use that kernel to reach that position. Don't know why you think that! " }, { "id": 934405, "postDate": "2020-07-18T12:16:06.227Z", "rawMarkdown": "", "isDeleted": true }, { "id": 934410, "postDate": "2020-07-18T12:18:56.680Z", "content": "

@synked

", "rawMarkdown": "@synked " } ] }, { "id": 937266, "postDate": "2020-07-20T21:47:23.123Z", "content": "

I don't see any issue with a high LB notebook, if there is the method (architecture / algorithm etc).\nThe issue is when the high score is based only on blending to me.

", "rawMarkdown": "I don't see any issue with a high LB notebook, if there is the method (architecture / algorithm etc).\nThe issue is when the high score is based only on blending to me.", "votes": 1 }, { "id": 934293, "postDate": "2020-07-18T11:09:27.567Z", "content": "

Agree. Notebooks should be only used for providing some idea or hints and not providing answers.

", "rawMarkdown": "Agree. Notebooks should be only used for providing some idea or hints and not providing answers. ", "votes": 1 }, { "id": 934172, "postDate": "2020-07-18T09:26:56.283Z", "content": "

Full code to generate Submission.csv should not be provided.\nThe author can change the model parameters so that you still have to work for a good score!

", "rawMarkdown": "Full code to generate `Submission.csv` should not be provided.\nThe author can change the model parameters so that you still have to work for a good score!", "votes": 1 }, { "id": 932579, "postDate": "2020-07-17T06:48:55.547Z", "content": "

I agree. I see the public PB getting crowded with most getting the same score as the public notebook every time someone posts a high scoring LB notebook.

", "rawMarkdown": "I agree. I see the public PB getting crowded with most getting the same score as the public notebook every time someone posts a high scoring LB notebook. ", "votes": 1 }, { "id": 937274, "postDate": "2020-07-20T22:13:32.543Z", "content": "

This topic appears in every popular Kaggle competition :)

", "rawMarkdown": "This topic appears in every popular Kaggle competition :)" }, { "id": 934245, "postDate": "2020-07-18T10:36:31.283Z", "rawMarkdown": "", "votes": -3, "isDeleted": true }, { "id": 932600, "postDate": "2020-07-17T07:06:43.190Z", "rawMarkdown": "", "votes": 2, "isDeleted": true, "replies": [ { "id": 933642, "postDate": "2020-07-17T21:09:56.760Z", "content": "

Nop! Your notebook is very helpful! I am very intesting w when i read it.

", "rawMarkdown": "Nop! Your notebook is very helpful! I am very intesting w when i read it. ", "votes": -1 }, { "id": 933861, "postDate": "2020-07-18T04:53:40.810Z", "rawMarkdown": "", "isDeleted": true } ] } ], "comments": [ { "id": 934287, "author_name": "Seifeddine Fezzani", "author_url": "", "post_date": "2020-07-18T11:02:04.210000", "content": "

Kaggle should allow us to downvote kernels as well. This might discourage people of sharing blend of blends and inference kernels with a very high score.

", "votes": 6, "replies": [ { "id": 934327, "author_name": "", "author_url": "", "post_date": "2020-07-18T11:35:32.227000", "content": "", "votes": -1, "replies": [] } ] }, { "id": 934241, "author_name": "Aziz_Belaweid", "author_url": "", "post_date": "2020-07-18T10:35:05.190000", "content": "

Yep definitely agree it's frustrating some times, I have been working a lot to get over the 0.94 baseline and when I did I saw that there's a public notebook that gives you better score by just ensembling a bunch of other submissions. I always remind myself that anyone who uses those notebooks to get higher lb won't get too far, he'll be stuck so it's one more reason to overtake them. I also agree with you the method/ indication should be given not the submission file. Good luck

", "votes": 3, "replies": [ { "id": 934381, "author_name": "Manh Lab", "author_url": "", "post_date": "2020-07-18T12:02:58.083000", "content": "

I see. I was working very much to reach 95.15 and POW! new notebook with higher LB and I down my position very quick

", "votes": 1, "replies": [] }, { "id": 934391, "author_name": "", "author_url": "", "post_date": "2020-07-18T12:07:46.043000", "content": "", "votes": 0, "replies": [] }, { "id": 934402, "author_name": "Manh Lab", "author_url": "", "post_date": "2020-07-18T12:13:38.367000", "content": "

I only reach ~70 today. And i don't use that kernel to reach that position. Don't know why you think that!

", "votes": 0, "replies": [] }, { "id": 934405, "author_name": "", "author_url": "", "post_date": "2020-07-18T12:16:06.227000", "content": "", "votes": 0, "replies": [] }, { "id": 934410, "author_name": "Manh Lab", "author_url": "", "post_date": "2020-07-18T12:18:56.680000", "content": "

@synked

", "votes": 0, "replies": [] } ] }, { "id": 937266, "author_name": "Shiro", "author_url": "", "post_date": "2020-07-20T21:47:23.123000", "content": "

I don't see any issue with a high LB notebook, if there is the method (architecture / algorithm etc).\nThe issue is when the high score is based only on blending to me.

", "votes": 1, "replies": [] }, { "id": 934293, "author_name": "lr", "author_url": "", "post_date": "2020-07-18T11:09:27.567000", "content": "

Agree. Notebooks should be only used for providing some idea or hints and not providing answers.

", "votes": 1, "replies": [] }, { "id": 934172, "author_name": "Nitesh Chaudhry", "author_url": "", "post_date": "2020-07-18T09:26:56.283000", "content": "

Full code to generate Submission.csv should not be provided.\nThe author can change the model parameters so that you still have to work for a good score!

", "votes": 1, "replies": [] }, { "id": 932579, "author_name": "Aman Arora", "author_url": "", "post_date": "2020-07-17T06:48:55.547000", "content": "

I agree. I see the public PB getting crowded with most getting the same score as the public notebook every time someone posts a high scoring LB notebook.

", "votes": 1, "replies": [] }, { "id": 937274, "author_name": "Jacek Poplawski", "author_url": "", "post_date": "2020-07-20T22:13:32.543000", "content": "

This topic appears in every popular Kaggle competition :)

", "votes": 0, "replies": [] }, { "id": 934245, "author_name": "", "author_url": "", "post_date": "2020-07-18T10:36:31.283000", "content": "", "votes": -3, "replies": [] }, { "id": 932600, "author_name": "", "author_url": "", "post_date": "2020-07-17T07:06:43.190000", "content": "", "votes": 2, "replies": [ { "id": 933642, "author_name": "Manh Lab", "author_url": "", "post_date": "2020-07-17T21:09:56.760000", "content": "

Nop! Your notebook is very helpful! I am very intesting w when i read it.

", "votes": -1, "replies": [] }, { "id": 933861, "author_name": "", "author_url": "", "post_date": "2020-07-18T04:53:40.810000", "content": "", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "932522": "I think many people like me, try hard on this competition, and have > 200 ratings. But if anyone public high LB notebooks, the LB is changed. That's terible!! You can give the method, not the SUBMISSION.CSV", "934287": "Kaggle should allow us to downvote kernels as well. This might discourage people of sharing blend of blends and inference kernels with a very high score.", "934241": "Yep definitely agree it's frustrating some times, I have been working a lot to get over the 0.94 baseline and when I did I saw that there's a public notebook that gives you better score by just ensembling a bunch of other submissions. I always remind myself that anyone who uses those notebooks to get higher lb won't get too far, he'll be stuck so it's one more reason to overtake them. I also agree with you the method/ indication should be given not the submission file. Good luck", "937266": "I don't see any issue with a high LB notebook, if there is the method (architecture / algorithm etc).\nThe issue is when the high score is based only on blending to me.", "934293": "Agree. Notebooks should be only used for providing some idea or hints and not providing answers. ", "934172": "Full code to generate `Submission.csv` should not be provided.\nThe author can change the model parameters so that you still have to work for a good score!", "932579": "I agree. I see the public PB getting crowded with most getting the same score as the public notebook every time someone posts a high scoring LB notebook. ", "937274": "This topic appears in every popular Kaggle competition :)", "934245": "", "932600": "" } }