Amazing to see the organizers publishing a high scoring kernel, that will be gard to beat without very good hardware...
\n\nThus pushing kagglers that cannot match the required H/W to just copy/edit and submit to get 0.271. The earlier the better right ?
\n\nObviously this could have been done by any of the competitors but just don't understand the goal of organizers there.
", "messages": [ { "id": 955278, "postDate": "2020-08-02T14:10:45Z", "content": "Amazing to see the organizers publishing a high scoring kernel, that will be gard to beat without very good hardware...
\n\nThus pushing kagglers that cannot match the required H/W to just copy/edit and submit to get 0.271. The earlier the better right ?
\n\nObviously this could have been done by any of the competitors but just don't understand the goal of organizers there.
", "rawMarkdown": "Amazing to see the organizers publishing a high scoring kernel, that will be gard to beat without very good hardware... \n\nThus pushing kagglers that cannot match the required H/W to just copy/edit and submit to get 0.271. The earlier the better right ?\n\nObviously this could have been done by any of the competitors but just don't understand the goal of organizers there.", "votes": 44 }, { "id": 956336, "postDate": "2020-08-03T12:22:10.580Z", "content": "My view Kaggle Points system should not use order of submission (timestamp) to establish points if scores are equal. They should average across those with the score.
\n\n22nd-304th of 362 teams all with same .271 score means 22nd gets 5400pts. Meanwhile, toshi_k in 20th only gets 5817pts, it doesn't seem fair. 304th gets 756pts
\n\nRemoving timestamp should average 22-304 places to around 1200pts. Even then, not particularly fair.
", "rawMarkdown": "My view Kaggle Points system should not use order of submission (timestamp) to establish points if scores are equal. They should average across those with the score.\n\n22nd-304th of 362 teams all with same .271 score means 22nd gets 5400pts. Meanwhile, toshi_k in 20th only gets 5817pts, it doesn't seem fair. 304th gets 756pts\n\nRemoving timestamp should average 22-304 places to around 1200pts. Even then, not particularly fair.", "votes": 5, "replies": [ { "id": 956377, "postDate": "2020-08-03T12:55:58.773Z", "content": "Thx @robga. It's a hard subject really.
\n\nI would maybe go by : if you're simply using someone else's predictions why would you get any points ?
", "rawMarkdown": "Thx @robga. It's a hard subject really. \n\nI would maybe go by : if you're simply using someone else's predictions why would you get any points ?", "votes": 2 }, { "id": 956454, "postDate": "2020-08-03T14:15:46.370Z", "content": "Yes I agree completely.
", "rawMarkdown": "Yes I agree completely. " } ] }, { "id": 955339, "postDate": "2020-08-02T15:07:46.747Z", "content": "Agreed. I was planning to participate but was discouraged because <20 teams have been able to beat the baseline so far, so will probably pass.
", "rawMarkdown": "Agreed. I was planning to participate but was discouraged because <20 teams have been able to beat the baseline so far, so will probably pass.", "votes": 3, "replies": [ { "id": 955389, "postDate": "2020-08-02T15:43:33.843Z", "content": "I worked very hard to beat the baseline. I feel unfair that early-baselines-guys will get the almost same ranking points.
", "rawMarkdown": "I worked very hard to beat the baseline. I feel unfair that early-baselines-guys will get the almost same ranking points.", "votes": 10, "replies": [ { "id": 955522, "postDate": "2020-08-02T17:16:36.893Z", "content": "I feel really bad for you and your hard work @toshik. The baseline is very high indeed !\nFully agree with you and I hope the retrieval challenge does not end the same way !
", "rawMarkdown": "I feel really bad for you and your hard work @toshik. The baseline is very high indeed !\nFully agree with you and I hope the retrieval challenge does not end the same way !" } ] }, { "id": 955399, "postDate": "2020-08-02T15:48:14.857Z", "content": "Yeah, that sucks. Having a high-performing baseline on the leaderboard is fine. Allowing competitors to submit it in a few seconds without having to do any work on their own is not fine.
", "rawMarkdown": "Yeah, that sucks. Having a high-performing baseline on the leaderboard is fine. Allowing competitors to submit it in a few seconds without having to do any work on their own is not fine. ", "votes": 1 }, { "id": 955726, "postDate": "2020-08-02T21:25:28.307Z", "content": "I feel your pain very feel, I also feel demotivated to even start participating from the begining of the competition. Not even worth to start working on the competition until the last 2 weeks or so...
", "rawMarkdown": "I feel your pain very feel, I also feel demotivated to even start participating from the begining of the competition. Not even worth to start working on the competition until the last 2 weeks or so...", "votes": 2 } ] }, { "id": 978558, "postDate": "2020-08-20T08:57:21.053Z", "content": "I agree with that. It's unsportsmanlike.
\nI also submitted it to have a starting point, but I was sure that the organizers would find a way to delete these identical copies of the decision. I trained my model for a long time, but, unfortunately, it did not rise above this decision.
\nThe task is really difficult. And the rules of the competition need to be improved.
", "rawMarkdown": "I agree with that. It's unsportsmanlike.\n\nI also submitted it to have a starting point, but I was sure that the organizers would find a way to delete these identical copies of the decision. I trained my model for a long time, but, unfortunately, it did not rise above this decision. \n\nThe task is really difficult. And the rules of the competition need to be improved.", "votes": 1, "replies": [ { "id": 978613, "postDate": "2020-08-20T09:34:31.417Z", "content": "I thought they would disqualify the dupes and keep the 2nd best submission but that did not happen.
", "rawMarkdown": "I thought they would disqualify the dupes and keep the 2nd best submission but that did not happen.", "votes": 1 }, { "id": 979344, "postDate": "2020-08-20T19:15:05.703Z", "content": "I was sure of that too
", "rawMarkdown": "I was sure of that too" } ] }, { "id": 960099, "postDate": "2020-08-06T06:20:11.547Z", "content": "Nobody has submitted the baseline in Google Landmark Recognition Comp yet here. It is currently 2nd place LB beating 75 teams.
", "rawMarkdown": "Nobody has submitted the baseline in Google Landmark Recognition Comp yet [here][1]. It is currently 2nd place LB beating 75 teams. \n\n[1]: https://www.kaggle.com/camaskew/host-baseline-example/comments?scriptVersionId=40191321", "votes": 1 }, { "id": 959405, "postDate": "2020-08-05T14:58:48.477Z", "content": "Yes, the interesting thing was that in an attempt to show an example on how to make a basic submission, the organizers handed us a very high scoring model. \nMaybe they were expecting much more from us...
\n\nAnyway, what I think should be done is to (if possible) not award points to anyone with the same score as the public notebook, and the rest (who beat the baseline) should get medals and points. After all, these people probably worked a tonne to beat 0.271. My guess is that by the end of the competition, a lot of 0.271s will still get a silver, so it's only fair to those who worked hard to be distinguished.
", "rawMarkdown": "Yes, the interesting thing was that in an attempt to show an example on how to make a basic submission, the organizers handed us a very high scoring model. \nMaybe they were expecting much more from us...\n\nAnyway, what I think should be done is to (if possible) not award points to anyone with the same score as the public notebook, and the rest (who beat the baseline) should get medals and points. After all, these people probably worked a tonne to beat 0.271. My guess is that by the end of the competition, a lot of 0.271s will still get a silver, so it's only fair to those who worked hard to be distinguished.", "votes": 1 }, { "id": 955806, "postDate": "2020-08-03T01:06:25.810Z", "content": "Providing sample is good, but LB shouldn't include baseline submission.
", "rawMarkdown": "Providing sample is good, but LB shouldn't include baseline submission.", "votes": 1 }, { "id": 955375, "postDate": "2020-08-02T15:29:19.277Z", "content": "I was willing to learn and get some incentive while competing and somehow thought the baseline was not a qualified submission... but it's publicly available so I guess all is good.
\n\nI will probably reserve my seat lol !
", "rawMarkdown": "I was willing to learn and get some incentive while competing and somehow thought the baseline was not a qualified submission... but it's publicly available so I guess all is good. \n\nI will probably reserve my seat lol !", "votes": 1 }, { "id": 956162, "postDate": "2020-08-03T09:52:36.673Z", "content": "yeah,that is really sad :(\nBut i'm atleast happy to know what techniques are working :)
", "rawMarkdown": "yeah,that is really sad :(\nBut i'm atleast happy to know what techniques are working :)", "votes": 2 }, { "id": 955931, "postDate": "2020-08-03T05:15:42.740Z", "content": "Seems like they learned from this and decided to not (yet) publish a similar high-scoring kernel for the recognition challenge.
", "rawMarkdown": "Seems like they learned from this and decided to not (yet) publish a similar high-scoring kernel for the recognition challenge.", "votes": 2, "replies": [ { "id": 955994, "postDate": "2020-08-03T06:24:40.920Z", "content": "yes @alexanderliao hopefully they don't do that again.
", "rawMarkdown": "yes @alexanderliao hopefully they don't do that again." } ] }, { "id": 960704, "postDate": "2020-08-06T15:56:53.117Z", "content": "The baseline is already used state-of-the-art methods, it is not importent for the baseline score, but get a method that better than state-of-the-art methods
", "rawMarkdown": "The baseline is already used state-of-the-art methods, it is not importent for the baseline score, but get a method that better than state-of-the-art methods" }, { "id": 955725, "postDate": "2020-08-02T21:24:10.070Z", "rawMarkdown": "", "votes": 3, "isDeleted": true } ], "comments": [ { "id": 956336, "author_name": "robga", "author_url": "", "post_date": "2020-08-03T12:22:10.580000", "content": "My view Kaggle Points system should not use order of submission (timestamp) to establish points if scores are equal. They should average across those with the score.
\n\n22nd-304th of 362 teams all with same .271 score means 22nd gets 5400pts. Meanwhile, toshi_k in 20th only gets 5817pts, it doesn't seem fair. 304th gets 756pts
\n\nRemoving timestamp should average 22-304 places to around 1200pts. Even then, not particularly fair.
", "votes": 5, "replies": [ { "id": 956377, "author_name": "olivier", "author_url": "", "post_date": "2020-08-03T12:55:58.773000", "content": "Thx @robga. It's a hard subject really.
\n\nI would maybe go by : if you're simply using someone else's predictions why would you get any points ?
", "votes": 2, "replies": [] }, { "id": 956454, "author_name": "robga", "author_url": "", "post_date": "2020-08-03T14:15:46.370000", "content": "Yes I agree completely.
", "votes": 0, "replies": [] } ] }, { "id": 955339, "author_name": "Ian Pan", "author_url": "", "post_date": "2020-08-02T15:07:46.747000", "content": "Agreed. I was planning to participate but was discouraged because <20 teams have been able to beat the baseline so far, so will probably pass.
", "votes": 3, "replies": [ { "id": 955389, "author_name": "toshi_k", "author_url": "", "post_date": "2020-08-02T15:43:33.843000", "content": "I worked very hard to beat the baseline. I feel unfair that early-baselines-guys will get the almost same ranking points.
", "votes": 10, "replies": [ { "id": 955522, "author_name": "olivier", "author_url": "", "post_date": "2020-08-02T17:16:36.893000", "content": "I feel really bad for you and your hard work @toshik. The baseline is very high indeed !\nFully agree with you and I hope the retrieval challenge does not end the same way !
", "votes": 0, "replies": [] } ] }, { "id": 955399, "author_name": "Ian Pan", "author_url": "", "post_date": "2020-08-02T15:48:14.857000", "content": "Yeah, that sucks. Having a high-performing baseline on the leaderboard is fine. Allowing competitors to submit it in a few seconds without having to do any work on their own is not fine.
", "votes": 1, "replies": [] }, { "id": 955726, "author_name": "Alin Cijov", "author_url": "", "post_date": "2020-08-02T21:25:28.307000", "content": "I feel your pain very feel, I also feel demotivated to even start participating from the begining of the competition. Not even worth to start working on the competition until the last 2 weeks or so...
", "votes": 2, "replies": [] } ] }, { "id": 978558, "author_name": "Vitalii Mokin", "author_url": "", "post_date": "2020-08-20T08:57:21.053000", "content": "I agree with that. It's unsportsmanlike.
\nI also submitted it to have a starting point, but I was sure that the organizers would find a way to delete these identical copies of the decision. I trained my model for a long time, but, unfortunately, it did not rise above this decision.
\nThe task is really difficult. And the rules of the competition need to be improved.
", "votes": 1, "replies": [ { "id": 978613, "author_name": "olivier", "author_url": "", "post_date": "2020-08-20T09:34:31.417000", "content": "I thought they would disqualify the dupes and keep the 2nd best submission but that did not happen.
", "votes": 1, "replies": [] }, { "id": 979344, "author_name": "Vitalii Mokin", "author_url": "", "post_date": "2020-08-20T19:15:05.703000", "content": "I was sure of that too
", "votes": 0, "replies": [] } ] }, { "id": 960099, "author_name": "Chris Deotte", "author_url": "", "post_date": "2020-08-06T06:20:11.547000", "content": "Nobody has submitted the baseline in Google Landmark Recognition Comp yet here. It is currently 2nd place LB beating 75 teams.
", "votes": 1, "replies": [] }, { "id": 959405, "author_name": "Ayush Goel", "author_url": "", "post_date": "2020-08-05T14:58:48.477000", "content": "Yes, the interesting thing was that in an attempt to show an example on how to make a basic submission, the organizers handed us a very high scoring model. \nMaybe they were expecting much more from us...
\n\nAnyway, what I think should be done is to (if possible) not award points to anyone with the same score as the public notebook, and the rest (who beat the baseline) should get medals and points. After all, these people probably worked a tonne to beat 0.271. My guess is that by the end of the competition, a lot of 0.271s will still get a silver, so it's only fair to those who worked hard to be distinguished.
", "votes": 1, "replies": [] }, { "id": 955806, "author_name": "Johnny Lee", "author_url": "", "post_date": "2020-08-03T01:06:25.810000", "content": "Providing sample is good, but LB shouldn't include baseline submission.
", "votes": 1, "replies": [] }, { "id": 955375, "author_name": "olivier", "author_url": "", "post_date": "2020-08-02T15:29:19.277000", "content": "I was willing to learn and get some incentive while competing and somehow thought the baseline was not a qualified submission... but it's publicly available so I guess all is good.
\n\nI will probably reserve my seat lol !
", "votes": 1, "replies": [] }, { "id": 956162, "author_name": "prakash", "author_url": "", "post_date": "2020-08-03T09:52:36.673000", "content": "yeah,that is really sad :(\nBut i'm atleast happy to know what techniques are working :)
", "votes": 2, "replies": [] }, { "id": 955931, "author_name": "Peiyuan Liao", "author_url": "", "post_date": "2020-08-03T05:15:42.740000", "content": "Seems like they learned from this and decided to not (yet) publish a similar high-scoring kernel for the recognition challenge.
", "votes": 2, "replies": [ { "id": 955994, "author_name": "olivier", "author_url": "", "post_date": "2020-08-03T06:24:40.920000", "content": "yes @alexanderliao hopefully they don't do that again.
", "votes": 0, "replies": [] } ] }, { "id": 960704, "author_name": "", "author_url": "", "post_date": "2020-08-06T15:56:53.117000", "content": "The baseline is already used state-of-the-art methods, it is not importent for the baseline score, but get a method that better than state-of-the-art methods
", "votes": 0, "replies": [] }, { "id": 955725, "author_name": "", "author_url": "", "post_date": "2020-08-02T21:24:10.070000", "content": "", "votes": 3, "replies": [] } ], "raw_markdown_by_id": { "955278": "Amazing to see the organizers publishing a high scoring kernel, that will be gard to beat without very good hardware... \n\nThus pushing kagglers that cannot match the required H/W to just copy/edit and submit to get 0.271. The earlier the better right ?\n\nObviously this could have been done by any of the competitors but just don't understand the goal of organizers there.", "956336": "My view Kaggle Points system should not use order of submission (timestamp) to establish points if scores are equal. They should average across those with the score.\n\n22nd-304th of 362 teams all with same .271 score means 22nd gets 5400pts. Meanwhile, toshi_k in 20th only gets 5817pts, it doesn't seem fair. 304th gets 756pts\n\nRemoving timestamp should average 22-304 places to around 1200pts. Even then, not particularly fair.", "955339": "Agreed. I was planning to participate but was discouraged because <20 teams have been able to beat the baseline so far, so will probably pass.", "978558": "I agree with that. It's unsportsmanlike.\n\nI also submitted it to have a starting point, but I was sure that the organizers would find a way to delete these identical copies of the decision. I trained my model for a long time, but, unfortunately, it did not rise above this decision. \n\nThe task is really difficult. And the rules of the competition need to be improved.", "960099": "Nobody has submitted the baseline in Google Landmark Recognition Comp yet [here][1]. It is currently 2nd place LB beating 75 teams. \n\n[1]: https://www.kaggle.com/camaskew/host-baseline-example/comments?scriptVersionId=40191321", "959405": "Yes, the interesting thing was that in an attempt to show an example on how to make a basic submission, the organizers handed us a very high scoring model. \nMaybe they were expecting much more from us...\n\nAnyway, what I think should be done is to (if possible) not award points to anyone with the same score as the public notebook, and the rest (who beat the baseline) should get medals and points. After all, these people probably worked a tonne to beat 0.271. My guess is that by the end of the competition, a lot of 0.271s will still get a silver, so it's only fair to those who worked hard to be distinguished.", "955806": "Providing sample is good, but LB shouldn't include baseline submission.", "955375": "I was willing to learn and get some incentive while competing and somehow thought the baseline was not a qualified submission... but it's publicly available so I guess all is good. \n\nI will probably reserve my seat lol !", "956162": "yeah,that is really sad :(\nBut i'm atleast happy to know what techniques are working :)", "955931": "Seems like they learned from this and decided to not (yet) publish a similar high-scoring kernel for the recognition challenge.", "960704": "The baseline is already used state-of-the-art methods, it is not importent for the baseline score, but get a method that better than state-of-the-art methods", "955725": "" } }