For the second half of a month, I have invested time to run experiments, debug experiments, and compare results. Whenever I get into trouble, I will be very annoyed, but after a little improvement, I will feel that my contribution is worthwhile, and I am very motivated to continue to do so, even though it does not have any bonuses. But yesterday, when I was about to continue testing the results, I suddenly saw the test data leaks in the discussion, etc. (Thanks to @Hanke Chen for telling me), I was very disappointed at the moment, and even the mood and thoughts of the test were gone, it was ridiculous and regrettable. It’s just that what you’ve been fighting for and trying to do is like a scam. There is no need to go for a little time. Before the end, these words should be the last thing I left for this competition.
", "messages": [ { "id": "499713", "postDate": "03/25/2019 05:39:04", "content": "For the second half of a month, I have invested time to run experiments, debug experiments, and compare results. Whenever I get into trouble, I will be very annoyed, but after a little improvement, I will feel that my contribution is worthwhile, and I am very motivated to continue to do so, even though it does not have any bonuses. But yesterday, when I was about to continue testing the results, I suddenly saw the test data leaks in the discussion, etc. (Thanks to @Hanke Chen for telling me), I was very disappointed at the moment, and even the mood and thoughts of the test were gone, it was ridiculous and regrettable. It’s just that what you’ve been fighting for and trying to do is like a scam. There is no need to go for a little time. Before the end, these words should be the last thing I left for this competition.
", "rawMarkdown": "For the second half of a month, I have invested time to run experiments, debug experiments, and compare results. Whenever I get into trouble, I will be very annoyed, but after a little improvement, I will feel that my contribution is worthwhile, and I am very motivated to continue to do so, even though it does not have any bonuses. But yesterday, when I was about to continue testing the results, I suddenly saw the test data leaks in the discussion, etc. (Thanks to @Hanke Chen for telling me), I was very disappointed at the moment, and even the mood and thoughts of the test were gone, it was ridiculous and regrettable. It’s just that what you’ve been fighting for and trying to do is like a scam. There is no need to go for a little time. Before the end, these words should be the last thing I left for this competition.", "votes": null }, { "id": "499734", "postDate": "03/25/2019 06:08:04", "content": "Hello, I have the same feelings. The task description clearly states the following:
\n\nIn this competition, you must create an algorithm to identify metastatic cancer in small image patches taken from larger digital pathology scans
\n\nSo it's not about finding the right labels in the Internet, but about developing a method.
", "rawMarkdown": "Hello, I have the same feelings. The task description clearly states the following:\n\n**In this competition, you must create an algorithm to identify metastatic cancer in small image patches taken from larger digital pathology scans**\n\nSo it's not about finding the right labels in the Internet, but about developing a method.", "votes": null }, { "id": "499747", "postDate": "03/25/2019 06:17:05", "content": "Me too, I think kaggle should forbid those who use dirty tricks for seveal month even years.
", "rawMarkdown": "Me too, I think kaggle should forbid those who use dirty tricks for seveal month even years.", "votes": null }, { "id": "499750", "postDate": "03/25/2019 06:21:02", "content": "Indeed. A total leak like this rather extreme, its one thing to find and disclose it, choosing to use it for your submissions is another. I asked about the situation here in the kaggle user forum https://www.kaggle.com/general I suppose we will see.
", "rawMarkdown": "Indeed. A total leak like this rather extreme, its one thing to find and disclose it, choosing to use it for your submissions is another. I asked about the situation here in the kaggle user forum https://www.kaggle.com/general I suppose we will see.", "votes": null }, { "id": "499797", "postDate": "03/25/2019 07:53:51", "content": "Thank you！But I feel that the feedback should be small.
", "rawMarkdown": "Thank you！But I feel that the feedback should be small.", "votes": null }, { "id": "499799", "postDate": "03/25/2019 07:55:32", "content": "This doesn't matter, because kaggle should have a similar situation before. But it is still happening now, it is enough to explain the problem.
", "rawMarkdown": "This doesn't matter, because kaggle should have a similar situation before. But it is still happening now, it is enough to explain the problem.", "votes": null }, { "id": "499933", "postDate": "03/25/2019 12:12:04", "content": "Thank you for the real action!
", "rawMarkdown": "Thank you for the real action!", "votes": null }, { "id": "500500", "postDate": "03/26/2019 06:02:07", "content": "Actually it's quite hard for the competition holder to detect cheaters, since one could easily add noises into the 100% correct result, making it seems like an authentic submission. Or just average their results with the 100% correct answer.
", "rawMarkdown": "Actually it's quite hard for the competition holder to detect cheaters, since one could easily add noises into the 100% correct result, making it seems like an authentic submission. Or just average their results with the 100% correct answer.", "votes": null }, { "id": "500533", "postDate": "03/26/2019 07:22:44", "content": "although the leaderboard is a sideshow now, when it’s over I really hope those with strong models will post kernels or links to github or whatever (my full models time out in kernels). I’m far from having the best model but I expended significant effort on learning and improving for this task. I am quite curious about what I could have done differently or why some things work and others don’t, and I look forward to sharing with everyone who wishes.
", "rawMarkdown": "although the leaderboard is a sideshow now, when it’s over I really hope those with strong models will post kernels or links to github or whatever (my full models time out in kernels). I’m far from having the best model but I expended significant effort on learning and improving for this task. I am quite curious about what I could have done differently or why some things work and others don’t, and I look forward to sharing with everyone who wishes.", "votes": null }, { "id": "500542", "postDate": "03/26/2019 07:45:21", "content": "Yes, I am also looking forward to sharing after the competition.
", "rawMarkdown": "Yes, I am also looking forward to sharing after the competition.", "votes": null }, { "id": "500555", "postDate": "03/26/2019 08:18:51", "content": "I know how you feel, but try to think how much you learned from this competition. Maybe in future competitions you can use this knowledge to develop a better model and get improved results.
", "rawMarkdown": "I know how you feel, but try to think how much you learned from this competition. Maybe in future competitions you can use this knowledge to develop a better model and get improved results.", "votes": null }, { "id": "500558", "postDate": "03/26/2019 08:23:41", "content": "Yes, the Pandora's Box is opened.
", "rawMarkdown": "Yes, the Pandora's Box is opened.", "votes": null }, { "id": "500559", "postDate": "03/26/2019 08:25:02", "content": "Me too
", "rawMarkdown": "Me too", "votes": null }, { "id": "501055", "postDate": "03/26/2019 20:34:05", "content": "It's good to remember that this competition is just a playground. Even though I myself get motivated by the leaderboard, having a high ranking shouldn't be a motivation, but instead learning should be. Thus I'm happy that I've read some kernels and learnt many new things from the community.
", "rawMarkdown": "It's good to remember that this competition is just a playground. Even though I myself get motivated by the leaderboard, having a high ranking shouldn't be a motivation, but instead learning should be. Thus I'm happy that I've read some kernels and learnt many new things from the community.", "votes": null } ], "comments": [ { "id": 499734, "author_name": "robotdreams", "author_url": "", "post_date": "03/25/2019 06:08:04", "content": "Hello, I have the same feelings. The task description clearly states the following:
\n\nIn this competition, you must create an algorithm to identify metastatic cancer in small image patches taken from larger digital pathology scans
\n\nSo it's not about finding the right labels in the Internet, but about developing a method.
", "votes": null, "replies": [ { "id": 499750, "author_name": "interneuron", "author_url": "", "post_date": "03/25/2019 06:21:02", "content": "Indeed. A total leak like this rather extreme, its one thing to find and disclose it, choosing to use it for your submissions is another. I asked about the situation here in the kaggle user forum https://www.kaggle.com/general I suppose we will see.
", "votes": null, "replies": [] }, { "id": 499797, "author_name": "tianlei123", "author_url": "", "post_date": "03/25/2019 07:53:51", "content": "Thank you！But I feel that the feedback should be small.
", "votes": null, "replies": [] }, { "id": 499933, "author_name": "kokecacao", "author_url": "", "post_date": "03/25/2019 12:12:04", "content": "Thank you for the real action!
", "votes": null, "replies": [] } ] }, { "id": 499747, "author_name": "zxyu1995", "author_url": "", "post_date": "03/25/2019 06:17:05", "content": "Me too, I think kaggle should forbid those who use dirty tricks for seveal month even years.
", "votes": null, "replies": [ { "id": 499799, "author_name": "tianlei123", "author_url": "", "post_date": "03/25/2019 07:55:32", "content": "This doesn't matter, because kaggle should have a similar situation before. But it is still happening now, it is enough to explain the problem.
", "votes": null, "replies": [] } ] }, { "id": 500500, "author_name": "xzhhzx", "author_url": "", "post_date": "03/26/2019 06:02:07", "content": "Actually it's quite hard for the competition holder to detect cheaters, since one could easily add noises into the 100% correct result, making it seems like an authentic submission. Or just average their results with the 100% correct answer.
", "votes": null, "replies": [ { "id": 500558, "author_name": "robotdreams", "author_url": "", "post_date": "03/26/2019 08:23:41", "content": "Yes, the Pandora's Box is opened.
", "votes": null, "replies": [] } ] }, { "id": 500533, "author_name": "interneuron", "author_url": "", "post_date": "03/26/2019 07:22:44", "content": "although the leaderboard is a sideshow now, when it’s over I really hope those with strong models will post kernels or links to github or whatever (my full models time out in kernels). I’m far from having the best model but I expended significant effort on learning and improving for this task. I am quite curious about what I could have done differently or why some things work and others don’t, and I look forward to sharing with everyone who wishes.
", "votes": null, "replies": [ { "id": 500542, "author_name": "tianlei123", "author_url": "", "post_date": "03/26/2019 07:45:21", "content": "Yes, I am also looking forward to sharing after the competition.
", "votes": null, "replies": [] }, { "id": 500559, "author_name": "robotdreams", "author_url": "", "post_date": "03/26/2019 08:25:02", "content": "Me too
", "votes": null, "replies": [] } ] }, { "id": 500555, "author_name": "afenoy", "author_url": "", "post_date": "03/26/2019 08:18:51", "content": "I know how you feel, but try to think how much you learned from this competition. Maybe in future competitions you can use this knowledge to develop a better model and get improved results.
", "votes": null, "replies": [] }, { "id": 501055, "author_name": "korhanpolat", "author_url": "", "post_date": "03/26/2019 20:34:05", "content": "It's good to remember that this competition is just a playground. Even though I myself get motivated by the leaderboard, having a high ranking shouldn't be a motivation, but instead learning should be. Thus I'm happy that I've read some kernels and learnt many new things from the community.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "499713": "For the second half of a month, I have invested time to run experiments, debug experiments, and compare results. Whenever I get into trouble, I will be very annoyed, but after a little improvement, I will feel that my contribution is worthwhile, and I am very motivated to continue to do so, even though it does not have any bonuses. But yesterday, when I was about to continue testing the results, I suddenly saw the test data leaks in the discussion, etc. (Thanks to @Hanke Chen for telling me), I was very disappointed at the moment, and even the mood and thoughts of the test were gone, it was ridiculous and regrettable. It’s just that what you’ve been fighting for and trying to do is like a scam. There is no need to go for a little time. Before the end, these words should be the last thing I left for this competition.", "499734": "Hello, I have the same feelings. The task description clearly states the following:\n\n**In this competition, you must create an algorithm to identify metastatic cancer in small image patches taken from larger digital pathology scans**\n\nSo it's not about finding the right labels in the Internet, but about developing a method.", "499747": "Me too, I think kaggle should forbid those who use dirty tricks for seveal month even years.", "499750": "Indeed. A total leak like this rather extreme, its one thing to find and disclose it, choosing to use it for your submissions is another. I asked about the situation here in the kaggle user forum https://www.kaggle.com/general I suppose we will see.", "499797": "Thank you！But I feel that the feedback should be small.", "499799": "This doesn't matter, because kaggle should have a similar situation before. But it is still happening now, it is enough to explain the problem.", "499933": "Thank you for the real action!", "500500": "Actually it's quite hard for the competition holder to detect cheaters, since one could easily add noises into the 100% correct result, making it seems like an authentic submission. Or just average their results with the 100% correct answer.", "500533": "although the leaderboard is a sideshow now, when it’s over I really hope those with strong models will post kernels or links to github or whatever (my full models time out in kernels). I’m far from having the best model but I expended significant effort on learning and improving for this task. I am quite curious about what I could have done differently or why some things work and others don’t, and I look forward to sharing with everyone who wishes.", "500542": "Yes, I am also looking forward to sharing after the competition.", "500555": "I know how you feel, but try to think how much you learned from this competition. Maybe in future competitions you can use this knowledge to develop a better model and get improved results.", "500558": "Yes, the Pandora's Box is opened.", "500559": "Me too", "501055": "It's good to remember that this competition is just a playground. Even though I myself get motivated by the leaderboard, having a high ranking shouldn't be a motivation, but instead learning should be. Thus I'm happy that I've read some kernels and learnt many new things from the community." }, "source": "meta" }