![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1620223%2F1d0c244c180274fa5d21f821505fa1de%2F2019-08-22%2011.05.59.png?generation=1566441862077352&alt=media\")
I found it in opendatascience.\nI think it is private sharing.\nis it permitted in kaggle?\n@philculliton, @juliaelliott Please answer this question.\n
What matters most is healthy competitions and community. Many people compete at Kaggle because getting a Kaggle medal is one of the most prestigious honor in the data science community. It differentiates Kaggle from other competition platforms, and indeed, that is why I participate in Kaggle.
\n\nWhat if unfair contestants win over fair contestants? What if Kaggle medals are considered to indicate cheating ability instead of data science and machine learning ability? The outcome is trivial.
\n\nAdmins, please think seriously about dishonest acts. They are about to severely destroy the value of Kaggle.
", "rawMarkdown": "What matters most is healthy competitions and community. Many people compete at Kaggle because getting a Kaggle medal is one of the most prestigious honor in the data science community. It differentiates Kaggle from other competition platforms, and indeed, that is why I participate in Kaggle.\n\nWhat if unfair contestants win over fair contestants? What if Kaggle medals are considered to indicate cheating ability instead of data science and machine learning ability? The outcome is trivial.\n\nAdmins, please think seriously about dishonest acts. They are about to severely destroy the value of Kaggle.", "votes": 41, "replies": [ { "id": 605626, "postDate": "2019-08-22T15:04:03.227Z", "content": "Kaggle= Learn, Compete , Play but not Cheat\nIf you want to share/ask something, do it in the discussion forums. Kagglers will learn/help u :)
", "rawMarkdown": "Kaggle= Learn, Compete , Play but not Cheat\nIf you want to share/ask something, do it in the discussion forums. Kagglers will learn/help u :)", "votes": 1 } ] }, { "id": 605746, "postDate": "2019-08-22T17:33:18.797Z", "content": "The real private sharing is the one you'll never know about. And what you guys are obsessed with is a modern witch-hunting.
", "rawMarkdown": "The real private sharing is the one you'll never know about. And what you guys are obsessed with is a modern witch-hunting.", "votes": 29 }, { "id": 605481, "postDate": "2019-08-22T12:08:54.793Z", "content": "Based on experience, I would say it is permitted from kaggle - BUT (big but), it shouldn't be
", "rawMarkdown": "Based on experience, I would say it is permitted from kaggle - BUT (big but), **it shouldn't be**", "votes": 21 }, { "id": 605780, "postDate": "2019-08-22T18:26:23.270Z", "content": "Kaggle openly promotes learning and sharing across our community, and that extends to countless other media that have spawned from users' love of Kaggle (meetups, social media groups, slack channels, etc). What is prohibited in our rules is private sharing.
\n\nSharing occurs on a spectrum, and for us \"private sharing\" goes beyond non-specific ideas and involves exact models & parameters that would be privileged information for the people who created them. For this reason, we strongly advise any sharing that could be interpreted as private happen on Kaggle's forums or isolated within the team you are working with.
\n\nThe definition of sharing and what ventures into the realm of private will always be a grey area - but we are all professionals with a mutual goal and should exercise ourselves with a base level of civility & respect. We at Kaggle want to host competitions that bring our community together around a common interest - not create an environment that devolves into a thread like this. Let this also be a reminder of Kaggle's community guidelines.
\n\nLet's all focus on real data science work and discussions. Violations of Kaggle's rules are reviewed through our compliance channel. Clear, conclusive evidence should be provided if making a claim. Every claim is read and valid claims are handled accordingly. Note that action on substantiated claims may not be taken until the close of the competition, as part of our standard violators removal process.
", "rawMarkdown": "Kaggle openly promotes learning and sharing across our community, and that extends to countless other media that have spawned from users' love of Kaggle (meetups, social media groups, slack channels, etc). What is prohibited in our rules is private sharing. \n\nSharing occurs on a spectrum, and for us \"private sharing\" goes beyond non-specific ideas and involves exact models & parameters that would be privileged information for the people who created them. For this reason, **we strongly advise any sharing that could be interpreted as private happen on Kaggle's forums or isolated within the team you are working with**.\n\nThe definition of sharing and what ventures into the realm of private will always be a grey area - but we are all professionals with a mutual goal and should exercise ourselves with a base level of civility & respect. We at Kaggle want to host competitions that bring our community together around a common interest - not create an environment that devolves into a thread like this. Let this also be a reminder of [Kaggle's community guidelines](https://www.kaggle.com/general/60914).\n\nLet's all focus on real data science work and discussions. Violations of Kaggle's rules are reviewed through our [compliance channel](https://www.kaggle.com/compliance). Clear, conclusive evidence should be provided if making a claim. Every claim is read and valid claims are handled accordingly. Note that action on substantiated claims may not be taken until the close of the competition, as part of our standard violators removal process.", "votes": 17, "replies": [ { "id": 605786, "postDate": "2019-08-22T18:36:29.633Z", "content": "I think if Kaggle behaves transparent and tells that \"We have investigated X5 case and we didn't find any statistical evidence of private sharing looking at these people's submission times, scores. And the competitions they are successful at the same time are similar to random.\" Then these discussions could finish much earlier. And nobody would talk about it.
", "rawMarkdown": "I think if Kaggle behaves transparent and tells that \"We have investigated X5 case and we didn't find any statistical evidence of private sharing looking at these people's submission times, scores. And the competitions they are successful at the same time are similar to random.\" Then these discussions could finish much earlier. And nobody would talk about it.", "votes": 23 }, { "id": 605791, "postDate": "2019-08-22T18:44:33.527Z", "content": "In this I agree with @divrikwicky
", "rawMarkdown": "In this I agree with @divrikwicky ", "votes": 2 }, { "id": 605940, "postDate": "2019-08-23T01:35:21.500Z", "content": "I think there is a lot of nuance to this that's difficult to capture. There are many scenarios where some small forms of sharing functions as a good way to welcome people into the community, help people along and teach new concepts, but there is also the side of cheating and giving an unfair advantage and stealing other people's well earned medal placements.
\n\nHard to delineate between the two sometimes, but it seems to me that the constant negative public opinion surrounding a few individuals and groups shows it might be necessary to either change some rules or take some actions.
\n\nI think the bottom line is measuring things based on the spirit of the competition and site. Many people seem to feel things that have been going on have been against the spirit of the site.
", "rawMarkdown": "I think there is a lot of nuance to this that's difficult to capture. There are many scenarios where some small forms of sharing functions as a good way to welcome people into the community, help people along and teach new concepts, but there is also the side of cheating and giving an unfair advantage and stealing other people's well earned medal placements. \n\nHard to delineate between the two sometimes, but it seems to me that the constant negative public opinion surrounding a few individuals and groups shows it might be necessary to either change some rules or take some actions. \n\nI think the bottom line is measuring things based on the spirit of the competition and site. Many people seem to feel things that have been going on have been against the spirit of the site. ", "votes": 2 }, { "id": 606002, "postDate": "2019-08-23T04:32:46.970Z", "content": "@ryches \"constant negative public opinion surrounding a few individuals and groups\" statement is much exaggerated. I might not know the whole story, but what I clearly see now in this thread is a few individuals sometimes throwing accusations at open 30000+ people community. Then, there's \"X5\" mentioned, which is the name of the biggest retail company in Russia. \nAs @juliaelliott said in her post, violations of Kaggle's rules must be reviewed through compliance channel, and clear, conclusive evidence should be provided if making each claim. And finally, let's all focus on real data science work and discussions.
", "rawMarkdown": "@ryches \"constant negative public opinion surrounding a few individuals and **groups\"** statement is much exaggerated. I might not know the whole story, but what I clearly see now in this thread is a few individuals sometimes throwing accusations at open 30000+ people community. Then, there's \"X5\" mentioned, which is the name of the biggest retail company in Russia. \nAs @juliaelliott said in her post, violations of Kaggle's rules must be reviewed through compliance channel, and clear, conclusive evidence should be provided if making each claim. And finally, **let's all focus on real data science work and discussions**.", "votes": 1 }, { "id": 606010, "postDate": "2019-08-23T05:00:46.917Z", "content": "I never mentioned Russia or X5 or any 30k+ groups or any particular individuals. I've taken some flak for the meetup group I lead and I'd like it to be resolved and I think the only way to truly do that is to have more clearly defined rules or simply be more permissive of all kinds of sharing.
\n\nI don't know how kaggle has turned into this weird intercontinental battle. I don't care who you are, I don't care what country you're from, I don't care who you work for. I just want kaggle to be an environment where people can learn and be rewarded for their work and the rules to be fairly applied.
", "rawMarkdown": "I never mentioned Russia or X5 or any 30k+ groups or any particular individuals. I've taken some flak for the meetup group I lead and I'd like it to be resolved and I think the only way to truly do that is to have more clearly defined rules or simply be more permissive of all kinds of sharing. \n\nI don't know how kaggle has turned into this weird intercontinental battle. I don't care who you are, I don't care what country you're from, I don't care who you work for. I just want kaggle to be an environment where people can learn and be rewarded for their work and the rules to be fairly applied. ", "votes": 4 }, { "id": 606983, "postDate": "2019-08-24T12:02:00.383Z", "content": "@juliaelliottThanks for your reply.\nI agree with you. I send claim about this topic later.
", "rawMarkdown": "[@juliaelliott](https://www.kaggle.com/juliaelliott)Thanks for your reply.\nI agree with you. I send claim about this topic later.", "votes": 1 } ] }, { "id": 605061, "postDate": "2019-08-22T02:44:43.117Z", "content": "I found it in opendatascience.\nI think it is private sharing.\nis it permitted in kaggle?\n@philculliton, @juliaelliott Please answer this question.\n
What bothers me much more is the fact that you need to train for two weeks on a fat cluster for this competition.
", "rawMarkdown": "What bothers me much more is the fact that you need to train for two weeks on a fat cluster for this competition.", "votes": 12, "replies": [ { "id": 606813, "postDate": "2019-08-24T05:53:31.153Z", "content": "same is true for all image competitions. There is a very high correlation between LB rank and number of high end gpu used.
", "rawMarkdown": "same is true for all image competitions. There is a very high correlation between LB rank and number of high end gpu used.", "votes": 3 }, { "id": 606961, "postDate": "2019-08-24T11:29:05.877Z", "content": "Image and NLP competitions definitely benefit the mostly from having large racks. But there are also stark differences between image competitions, you can usually see it by the number of participants and whether Google is hosting it.
", "rawMarkdown": "Image and NLP competitions definitely benefit the mostly from having large racks. But there are also stark differences between image competitions, you can usually see it by the number of participants and whether Google is hosting it.", "votes": 4 }, { "id": 606970, "postDate": "2019-08-24T11:42:50.117Z", "content": "Indeed, the larger the data, the higher correlation between HW and LB. I loved TGS competition because dataset was small enough to be dealt with if you only had one gpu.
", "rawMarkdown": "Indeed, the larger the data, the higher correlation between HW and LB. I loved TGS competition because dataset was small enough to be dealt with if you only had one gpu.", "votes": 1 }, { "id": 611135, "postDate": "2019-08-29T06:00:41.437Z", "content": "Recommend Steel competition which limits the kernel running time to less than 1 hour
", "rawMarkdown": "Recommend Steel competition which limits the kernel running time to less than 1 hour" } ] }, { "id": 605502, "postDate": "2019-08-22T12:38:13.750Z", "content": "He is Private Sharing Grandmaster.
", "rawMarkdown": "He is Private Sharing Grandmaster.", "votes": 18 }, { "id": 605728, "postDate": "2019-08-22T16:57:08.687Z", "content": "Some new crazy private sharing from ods.ai:
\n\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F727004%2F37f0a2a20a976888f3df8943be2e0e9f%2F2019-08-22_19-56-21.jpg?generation=1566493025393414&alt=media\")
Could you please write how does this mean? I cannot copy & paste and translate it...
", "rawMarkdown": "Could you please write how does this mean? I cannot copy & paste and translate it..." }, { "id": 611168, "postDate": "2019-08-29T06:28:59.367Z", "content": "It means: \"Why tune LightGBM? It seems that XGBoost gives a better score\" :)
", "rawMarkdown": "It means: \"Why tune LightGBM? It seems that XGBoost gives a better score\" :)", "votes": 2 } ] }, { "id": 605559, "postDate": "2019-08-22T13:51:57.663Z", "content": "All Kagglers should write ideas, code, discussion of ongoing competitions in Kaggle platform, if you want to share those other people.\nwe cannot trace information of competiton. because we cannot check all communication tools.\nThat situation cause lack of fairness for competitors.
", "rawMarkdown": "All Kagglers should write ideas, code, discussion of ongoing competitions in Kaggle platform, if you want to share those other people.\nwe cannot trace information of competiton. because we cannot check all communication tools.\nThat situation cause lack of fairness for competitors. ", "votes": 9 }, { "id": 605672, "postDate": "2019-08-22T15:58:55.863Z", "content": "I believe that this insight doesn't pump ts up on leaderboard. This is not a \"solution\" imho. All we learn from this ps is \"use dgx\". Totally changed game, right? \nIn my eyes it is not about private sharing, it is all about kicking mvp's from lb. And it would rob us from oportunity to see really best solution in the end of competition.
", "rawMarkdown": "I believe that this insight doesn't pump ts up on leaderboard. This is not a \"solution\" imho. All we learn from this ps is \"use dgx\". Totally changed game, right? \nIn my eyes it is not about private sharing, it is all about kicking mvp's from lb. And it would rob us from oportunity to see really best solution in the end of competition. ", "votes": 11 }, { "id": 605711, "postDate": "2019-08-22T16:44:30.687Z", "content": "lol
", "rawMarkdown": "lol", "votes": 8 }, { "id": 605537, "postDate": "2019-08-22T13:17:08.420Z", "content": "He should write his ideas about ONGOING competition in Kaggle discussions.\nIf he want to share ideas in opendatascience, write ideas here then share link.
", "rawMarkdown": "He should write his ideas about ONGOING competition in Kaggle discussions.\nIf he want to share ideas in opendatascience, write ideas here then share link.", "votes": 8 }, { "id": 605589, "postDate": "2019-08-22T14:15:42.293Z", "content": "「Я придумал. Давайте сделаем канал что-то типа #none_specific_ml_general_tricks\nНу и там все будут писать в стиле:\n\"Я тут работаю над задачей многоклассовой инстанс сегменации. и htc_dcn_x101_ibn_inshala работает охуено и по ищущению накидывает японцем за щеку\". Или \"я тут работаю над задачей сегментации легких. И у меня \". А им буду в тредиках отвечать \"о, збс, я кажется тоже работаю над такой задачей, а у меня densenet заходит\".」
\n\nI think it is so toxic.\nWhy he is kaggle grandmaster?\nWhy Kaggle overlook odsai and him?
\n\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1620223%2F13fe82ab957fd369f77d93b2373fc8be%2FScreenshot%20from%202019-08-22%2023-12-03.png?generation=1566483160950062&alt=media\")
Are you our big brother now? It's called free speech, many people actually died for this.
", "rawMarkdown": "Are you our big brother now? It's called free speech, many people actually died for this.", "votes": -10 }, { "id": 605661, "postDate": "2019-08-22T15:43:50.540Z", "content": "Comrade @raisefamous is ready to die for private sharing. As long as Kaggle has such (random) heroes, we will always feel safe and secure.
", "rawMarkdown": "Comrade @raisefamous is ready to die for private sharing. As long as Kaggle has such (random) heroes, we will always feel safe and secure.", "votes": -10 }, { "id": 605694, "postDate": "2019-08-22T16:14:46.600Z", "content": "> I'm right. Argue with me = you are a biased fanatic. I win, bye-bye.
\n\nNice logic 👍
", "rawMarkdown": "> I'm right. Argue with me = you are a biased fanatic. I win, bye-bye. \n\nNice logic 👍", "votes": 1 }, { "id": 605696, "postDate": "2019-08-22T16:17:57.177Z", "content": "Calm down Ahmed, private sharing is not something I care about here. My point is about \"toxicity\", if you want to call someone a bad name on a platform where it is permitted you are free to do so.
", "rawMarkdown": "Calm down Ahmed, private sharing is not something I care about here. My point is about \"toxicity\", if you want to call someone a bad name on a platform where it is permitted you are free to do so.", "votes": 3 }, { "id": 605713, "postDate": "2019-08-22T16:45:15.617Z", "content": "\n\n\nRandom Hero wrote:
\n \nAre you our big brother now? It's called free speech, many people actually died for this.
\n
Free speech does not mean you can say whatever you want, especially those offensive words, Comrade @raisefamous.
", "rawMarkdown": "> **Random Hero wrote:**\n> \n> Are you our big brother now? It's called free speech, many people actually died for this.\n\n\nFree speech does not mean you can say whatever you want, especially those offensive words, Comrade @raisefamous. \n", "votes": 1 }, { "id": 605717, "postDate": "2019-08-22T16:47:14.900Z", "content": "it actually does
", "rawMarkdown": "it actually does", "votes": 2 }, { "id": 605723, "postDate": "2019-08-22T16:53:58.953Z", "content": "I didn't know that @phalanx has the same power as Big Brother. \nGenerally, people who feel guilty tend to label people as “Big Brother” , anyway, if you can't read the clear bad faith in the above speech , your ideal home is in Oceania at 1984.
", "rawMarkdown": "I didn't know that @phalanx has the same power as Big Brother. \nGenerally, people who feel guilty tend to label people as “Big Brother” , anyway, if you can't read the clear bad faith in the above speech , your ideal home is in Oceania at 1984.", "votes": 4 }, { "id": 605732, "postDate": "2019-08-22T17:06:45.100Z", "rawMarkdown": "", "votes": -4, "isDeleted": true }, { "id": 605763, "postDate": "2019-08-22T18:07:56.283Z", "rawMarkdown": "", "votes": 4, "isDeleted": true }, { "id": 605784, "postDate": "2019-08-22T18:31:22.763Z", "rawMarkdown": "", "votes": -1, "isDeleted": true }, { "id": 605794, "postDate": "2019-08-22T18:47:23Z", "content": ".
", "rawMarkdown": ".", "votes": -3 }, { "id": 605800, "postDate": "2019-08-22T18:51:03.810Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 605801, "postDate": "2019-08-22T18:53:46.230Z", "rawMarkdown": "", "votes": 2, "isDeleted": true }, { "id": 605806, "postDate": "2019-08-22T19:02:40.600Z", "rawMarkdown": "", "votes": -3, "isDeleted": true }, { "id": 605808, "postDate": "2019-08-22T19:11:44.337Z", "rawMarkdown": "", "votes": 4, "isDeleted": true }, { "id": 605813, "postDate": "2019-08-22T19:18:57.190Z", "rawMarkdown": "", "votes": -1, "isDeleted": true }, { "id": 605864, "postDate": "2019-08-22T21:00:28.120Z", "rawMarkdown": "", "votes": -1, "isDeleted": true }, { "id": 605872, "postDate": "2019-08-22T21:22:36.963Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ] }, { "id": 605573, "postDate": "2019-08-22T14:02:29.260Z", "content": "n01z3「 текущий результат на inc seg это одна модель или ансамбль? Это модель только на 1 уровень иерархии или все?」\nzfturbo「Ансамбль и все 300 классов」\nThis is also private sharing.\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1620223%2Ff8e11a24da13e202f9ece784e719e2de%2FScreenshot%20from%202019-08-22%2022-59-35.png?generation=1566482399768149&alt=media\")
I'd like to guard ZFTurbo here. The fact that a top model is an ensemble is not news for Kaggle. It's so in 99.95% of cases. As well as the fact that it makes use of all data.
", "rawMarkdown": "I'd like to guard ZFTurbo here. The fact that a top model is an ensemble is not news for Kaggle. It's so in 99.95% of cases. As well as the fact that it makes use of all data.", "votes": 19 }, { "id": 606066, "postDate": "2019-08-23T06:36:03.340Z", "content": "ZFTurbo is true top.
", "rawMarkdown": "ZFTurbo is true top. ", "votes": 5 } ] }, { "id": 605106, "postDate": "2019-08-22T04:34:56.203Z", "content": "hm... let me think... looks more like an envy
", "rawMarkdown": "hm... let me think... looks more like an envy", "votes": -1, "replies": [ { "id": 605114, "postDate": "2019-08-22T04:45:08.687Z", "content": "not envy.\nI want to know wheter it is private sharing or not.\nIf not, I can share my solution about current competition in kaggler-ja.
", "rawMarkdown": "not envy.\nI want to know wheter it is private sharing or not.\nIf not, I can share my solution about current competition in kaggler-ja.", "votes": 10 }, { "id": 605116, "postDate": "2019-08-22T04:46:43.767Z", "content": "It's not about being envy or something....It's more about competing fairly within the rules of the competitions!! If everybody start sharing their secrets in some random slack channels, then people in that channel will have an unfair advantage compared to other competitors!! So letz compete hard but within the spirit of the competition
", "rawMarkdown": "It's not about being envy or something....It's more about competing fairly within the rules of the competitions!! If everybody start sharing their secrets in some random slack channels, then people in that channel will have an unfair advantage compared to other competitors!! So letz compete hard but within the spirit of the competition", "votes": 8 }, { "id": 605398, "postDate": "2019-08-22T10:51:16.140Z", "content": "I can hardly understand some people(at least 5) who downvote this comment by @bibek777 .\nDo you say you want an unfair competition?
", "rawMarkdown": "I can hardly understand some people(at least 5) who downvote this comment by @bibek777 .\nDo you say you want an unfair competition?", "votes": 7 }, { "id": 605412, "postDate": "2019-08-22T11:07:04.177Z", "content": "I say this particular example is not breaking current Kaggle rules. Moreover, sharing ideas (not code or data!) in an open community is within the spirit of Kaggle competitions and open science in general. I think, that's why the downwotes to bibek's comment.
", "rawMarkdown": "I say this particular example is not breaking current Kaggle rules. Moreover, sharing ideas (not code or data!) in an open community *is* within the spirit of Kaggle competitions and open science in general. I think, that's why the downwotes to bibek's comment.", "votes": 9 }, { "id": 605413, "postDate": "2019-08-22T11:07:41.243Z", "content": "Thank u soo much @cashfeg !!! was feeling little bit down after seeing so many downvotes :(
", "rawMarkdown": "Thank u soo much @cashfeg !!! was feeling little bit down after seeing so many downvotes :(" }, { "id": 605442, "postDate": "2019-08-22T11:31:49.803Z", "content": "Sharing ideas here in Kaggle discussions is clearly the best for open data science and Kaggle. Isn't it?
", "rawMarkdown": "Sharing ideas here in Kaggle discussions is clearly the best for open data science and Kaggle. Isn't it?", "votes": 1 }, { "id": 605452, "postDate": "2019-08-22T11:38:02.950Z", "content": "Even though it might be beneficial to Kaggle as a commercial organization, attempting to force all discussions into one platform is not the best for open science.
", "rawMarkdown": "Even though it might be beneficial to Kaggle as a commercial organization, attempting to **force** all discussions into one platform is not the best for open science. ", "votes": -2 }, { "id": 605488, "postDate": "2019-08-22T12:19:01.767Z", "content": "Is that more important than to keep Kaggle competition fair?\nThis competition is ONGOING. \nKaggle is the organizer of this competition, neither an unrelated commercial organization, nor an unrelated platform.
", "rawMarkdown": "Is that more important than to keep Kaggle competition fair?\nThis competition is ONGOING. \nKaggle is the organizer of this competition, neither an unrelated commercial organization, nor an unrelated platform. ", "votes": 3 }, { "id": 605494, "postDate": "2019-08-22T12:24:47.363Z", "content": "You are mistranslating what I'm saying. Once again, in my opinion it is not breaking the Kaggle rules and is in within the spirit. Therefore it does not make this ongoing competition unfair.
", "rawMarkdown": "You are mistranslating what I'm saying. Once again, in my opinion it is not breaking the Kaggle rules and is in within the spirit. Therefore it does not make this ongoing competition unfair. ", "votes": 4 }, { "id": 605501, "postDate": "2019-08-22T12:35:54.883Z", "content": "I think it does make this competition unfair. \nTrue this sentence is not code or data, so it might be not breaking Kaggle rules, but it's almost equivalent as code. \nSince some people say twisted logics here and there, I do want Kaggle to set strict rules without gray zone.
", "rawMarkdown": "I think it does make this competition unfair. \nTrue this sentence is not code or data, so it might be not breaking Kaggle rules, but it's almost equivalent as code. \nSince some people say twisted logics here and there, I do want Kaggle to set strict rules without gray zone. ", "votes": -2 }, { "id": 605534, "postDate": "2019-08-22T13:15:00.800Z", "content": "\n\n\nideas (not code or data!)
\n
When sharing top solutions after competition, code and data are generally not shared.\nI think it's strange that there is no problem if code and data are not shared.
", "rawMarkdown": "> ideas (not code or data!)\n\nWhen sharing top solutions after competition, code and data are generally not shared.\nI think it's strange that there is no problem if code and data are not shared.", "votes": 4 } ] }, { "id": 605739, "postDate": "2019-08-22T17:12:00.210Z", "content": "Well, if you have a DGX :)
", "rawMarkdown": "Well, if you have a DGX :)", "votes": 4 }, { "id": 605544, "postDate": "2019-08-22T13:24:27.863Z", "content": "I read all this thread and I cannot find a translation of what is quoted from ods.ai. This is really weird. Maybe there is no real info shared here after all...
", "rawMarkdown": "I read all this thread and I cannot find a translation of what is quoted from ods.ai. This is really weird. Maybe there is no real info shared here after all...\n\n", "votes": 4, "replies": [ { "id": 605554, "postDate": "2019-08-22T13:40:20.600Z", "content": "Here is a translation:
\n\nMy current result is htc_dcn_x101, TTA_minor. It trained only on the first level of hierarchy and it took ~2 weeks on DGX
Here's transcribed text from image. \nМой текущий результат зто htc_dcn_x101, TTA_minor.\nОбучался только на 1й уровень иерархии в течении~ 2 недели на DGX.
", "rawMarkdown": "Here's transcribed text from image. \nМой текущий результат зто htc_dcn_x101, TTA_minor.\nОбучался только на 1й уровень иерархии в течении~ 2 недели на DGX.", "votes": 1 }, { "id": 606800, "postDate": "2019-08-24T05:24:42.867Z", "content": "Andrew, thanks!
", "rawMarkdown": "Andrew, thanks!" } ] }, { "id": 605079, "postDate": "2019-08-22T03:18:55.687Z", "content": "Some people never learn !!!
", "rawMarkdown": "Some people never learn !!!", "votes": 1, "replies": [ { "id": 605210, "postDate": "2019-08-22T06:59:27.750Z", "content": "No punishment = no learning\nIt is not their fault, it is Kaggle's fault.
", "rawMarkdown": "No punishment = no learning\nIt is not their fault, it is Kaggle's fault.", "votes": 21 }, { "id": 605499, "postDate": "2019-08-22T12:32:09.343Z", "content": "yeah, they can't backprop correctly without the correct target.
", "rawMarkdown": "yeah, they can't backprop correctly without the correct target.", "votes": 8 }, { "id": 605621, "postDate": "2019-08-22T14:54:50.107Z", "content": "Looks like his learning rate is completely 0.
", "rawMarkdown": "Looks like his learning rate is completely 0.", "votes": 9 } ] }, { "id": 605666, "postDate": "2019-08-22T15:54:51.757Z", "content": "Yearly Hot Kaggle Word: \"X5\"
", "rawMarkdown": "Yearly Hot Kaggle Word: \"X5\"", "votes": 2, "replies": [ { "id": 605674, "postDate": "2019-08-22T16:00:35.690Z", "rawMarkdown": "", "votes": -9 } ] }, { "id": 605179, "postDate": "2019-08-22T06:25:13.547Z", "content": "@phalanx if it is private sharing, how in fact did you get to this info? 🤔
", "rawMarkdown": "@phalanx *if* it is private sharing, how in fact did *you* get to this info? 🤔 ", "votes": -7, "replies": [ { "id": 605191, "postDate": "2019-08-22T06:48:38.560Z", "content": "If you want to say \"slack is public\", it's outdated discussion.\nRefer to https://www.kaggle.com/c/instant-gratification/discussion/94894#549365
", "rawMarkdown": "If you want to say \"slack is public\", it's outdated discussion.\nRefer to https://www.kaggle.com/c/instant-gratification/discussion/94894#549365", "votes": 17 } ] }, { "id": 605373, "postDate": "2019-08-22T10:04:57.197Z", "content": "It's a new easy way to became a discussion master.
", "rawMarkdown": "It's a new easy way to became a discussion master.", "votes": -6 }, { "id": 607179, "postDate": "2019-08-24T18:16:08.980Z", "content": "I think they have to work with the sample submission data, by submitting that data you directly get the result and i think most of the people do that because i used to do it in my early time in kaggle but then i realize that by doing that what you are getting with it so i took a time around a year to learn as much as i can.
", "rawMarkdown": "I think they have to work with the sample submission data, by submitting that data you directly get the result and i think most of the people do that because i used to do it in my early time in kaggle but then i realize that by doing that what you are getting with it so i took a time around a year to learn as much as i can.", "votes": -1 }, { "id": 623642, "postDate": "2019-09-11T07:06:27.170Z", "content": "Hello Guys, I am looking for teammates.
", "rawMarkdown": "Hello Guys, I am looking for teammates.\n" }, { "id": 606887, "postDate": "2019-08-24T08:59:18.110Z", "rawMarkdown": "", "isDeleted": true }, { "id": 606199, "postDate": "2019-08-23T10:06:01.240Z", "rawMarkdown": "", "isDeleted": true }, { "id": 605776, "postDate": "2019-08-22T18:21:41.820Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 605617, "author_name": "iwiwi", "author_url": "", "post_date": "2019-08-22T14:50:51.540000", "content": "What matters most is healthy competitions and community. Many people compete at Kaggle because getting a Kaggle medal is one of the most prestigious honor in the data science community. It differentiates Kaggle from other competition platforms, and indeed, that is why I participate in Kaggle.
\n\nWhat if unfair contestants win over fair contestants? What if Kaggle medals are considered to indicate cheating ability instead of data science and machine learning ability? The outcome is trivial.
\n\nAdmins, please think seriously about dishonest acts. They are about to severely destroy the value of Kaggle.
", "votes": 41, "replies": [ { "id": 605626, "author_name": "Bibek", "author_url": "", "post_date": "2019-08-22T15:04:03.227000", "content": "Kaggle= Learn, Compete , Play but not Cheat\nIf you want to share/ask something, do it in the discussion forums. Kagglers will learn/help u :)
", "votes": 1, "replies": [] } ] }, { "id": 605746, "author_name": "Kostya Atarik", "author_url": "", "post_date": "2019-08-22T17:33:18.797000", "content": "The real private sharing is the one you'll never know about. And what you guys are obsessed with is a modern witch-hunting.
", "votes": 29, "replies": [] }, { "id": 605481, "author_name": "Μαριος Μιχαηλιδης KazAnova", "author_url": "", "post_date": "2019-08-22T12:08:54.793000", "content": "Based on experience, I would say it is permitted from kaggle - BUT (big but), it shouldn't be
", "votes": 21, "replies": [] }, { "id": 605780, "author_name": "Julia Elliott", "author_url": "", "post_date": "2019-08-22T18:26:23.270000", "content": "Kaggle openly promotes learning and sharing across our community, and that extends to countless other media that have spawned from users' love of Kaggle (meetups, social media groups, slack channels, etc). What is prohibited in our rules is private sharing.
\n\nSharing occurs on a spectrum, and for us \"private sharing\" goes beyond non-specific ideas and involves exact models & parameters that would be privileged information for the people who created them. For this reason, we strongly advise any sharing that could be interpreted as private happen on Kaggle's forums or isolated within the team you are working with.
\n\nThe definition of sharing and what ventures into the realm of private will always be a grey area - but we are all professionals with a mutual goal and should exercise ourselves with a base level of civility & respect. We at Kaggle want to host competitions that bring our community together around a common interest - not create an environment that devolves into a thread like this. Let this also be a reminder of Kaggle's community guidelines.
\n\nLet's all focus on real data science work and discussions. Violations of Kaggle's rules are reviewed through our compliance channel. Clear, conclusive evidence should be provided if making a claim. Every claim is read and valid claims are handled accordingly. Note that action on substantiated claims may not be taken until the close of the competition, as part of our standard violators removal process.
", "votes": 17, "replies": [ { "id": 605786, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2019-08-22T18:36:29.633000", "content": "I think if Kaggle behaves transparent and tells that \"We have investigated X5 case and we didn't find any statistical evidence of private sharing looking at these people's submission times, scores. And the competitions they are successful at the same time are similar to random.\" Then these discussions could finish much earlier. And nobody would talk about it.
", "votes": 23, "replies": [] }, { "id": 605791, "author_name": "Andrey Lukyanenko", "author_url": "", "post_date": "2019-08-22T18:44:33.527000", "content": "In this I agree with @divrikwicky
", "votes": 2, "replies": [] }, { "id": 605940, "author_name": "ryches", "author_url": "", "post_date": "2019-08-23T01:35:21.500000", "content": "I think there is a lot of nuance to this that's difficult to capture. There are many scenarios where some small forms of sharing functions as a good way to welcome people into the community, help people along and teach new concepts, but there is also the side of cheating and giving an unfair advantage and stealing other people's well earned medal placements.
\n\nHard to delineate between the two sometimes, but it seems to me that the constant negative public opinion surrounding a few individuals and groups shows it might be necessary to either change some rules or take some actions.
\n\nI think the bottom line is measuring things based on the spirit of the competition and site. Many people seem to feel things that have been going on have been against the spirit of the site.
", "votes": 2, "replies": [] }, { "id": 606002, "author_name": "Kostya Kravchenko", "author_url": "", "post_date": "2019-08-23T04:32:46.970000", "content": "@ryches \"constant negative public opinion surrounding a few individuals and groups\" statement is much exaggerated. I might not know the whole story, but what I clearly see now in this thread is a few individuals sometimes throwing accusations at open 30000+ people community. Then, there's \"X5\" mentioned, which is the name of the biggest retail company in Russia. \nAs @juliaelliott said in her post, violations of Kaggle's rules must be reviewed through compliance channel, and clear, conclusive evidence should be provided if making each claim. And finally, let's all focus on real data science work and discussions.
", "votes": 1, "replies": [] }, { "id": 606010, "author_name": "ryches", "author_url": "", "post_date": "2019-08-23T05:00:46.917000", "content": "I never mentioned Russia or X5 or any 30k+ groups or any particular individuals. I've taken some flak for the meetup group I lead and I'd like it to be resolved and I think the only way to truly do that is to have more clearly defined rules or simply be more permissive of all kinds of sharing.
\n\nI don't know how kaggle has turned into this weird intercontinental battle. I don't care who you are, I don't care what country you're from, I don't care who you work for. I just want kaggle to be an environment where people can learn and be rewarded for their work and the rules to be fairly applied.
", "votes": 4, "replies": [] }, { "id": 606983, "author_name": "phalanx", "author_url": "", "post_date": "2019-08-24T12:02:00.383000", "content": "@juliaelliottThanks for your reply.\nI agree with you. I send claim about this topic later.
", "votes": 1, "replies": [] } ] }, { "id": 606528, "author_name": "Psi", "author_url": "", "post_date": "2019-08-23T17:55:19.570000", "content": "What bothers me much more is the fact that you need to train for two weeks on a fat cluster for this competition.
", "votes": 12, "replies": [ { "id": 606813, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-24T05:53:31.153000", "content": "same is true for all image competitions. There is a very high correlation between LB rank and number of high end gpu used.
", "votes": 3, "replies": [] }, { "id": 606961, "author_name": "Psi", "author_url": "", "post_date": "2019-08-24T11:29:05.877000", "content": "Image and NLP competitions definitely benefit the mostly from having large racks. But there are also stark differences between image competitions, you can usually see it by the number of participants and whether Google is hosting it.
", "votes": 4, "replies": [] }, { "id": 606970, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-24T11:42:50.117000", "content": "Indeed, the larger the data, the higher correlation between HW and LB. I loved TGS competition because dataset was small enough to be dealt with if you only had one gpu.
", "votes": 1, "replies": [] }, { "id": 611135, "author_name": "Yiheng Wang", "author_url": "", "post_date": "2019-08-29T06:00:41.437000", "content": "Recommend Steel competition which limits the kernel running time to less than 1 hour
", "votes": 0, "replies": [] } ] }, { "id": 605502, "author_name": "toshi_k", "author_url": "", "post_date": "2019-08-22T12:38:13.750000", "content": "He is Private Sharing Grandmaster.
", "votes": 18, "replies": [] }, { "id": 605728, "author_name": "Andrey Lukyanenko", "author_url": "", "post_date": "2019-08-22T16:57:08.687000", "content": "Some new crazy private sharing from ods.ai:
\n\n
Could you please write how does this mean? I cannot copy & paste and translate it...
", "votes": 0, "replies": [] }, { "id": 611168, "author_name": "Ilia Zaitsev", "author_url": "", "post_date": "2019-08-29T06:28:59.367000", "content": "It means: \"Why tune LightGBM? It seems that XGBoost gives a better score\" :)
", "votes": 2, "replies": [] } ] }, { "id": 605559, "author_name": "tereka", "author_url": "", "post_date": "2019-08-22T13:51:57.663000", "content": "All Kagglers should write ideas, code, discussion of ongoing competitions in Kaggle platform, if you want to share those other people.\nwe cannot trace information of competiton. because we cannot check all communication tools.\nThat situation cause lack of fairness for competitors.
", "votes": 9, "replies": [] }, { "id": 605672, "author_name": "seed42", "author_url": "", "post_date": "2019-08-22T15:58:55.863000", "content": "I believe that this insight doesn't pump ts up on leaderboard. This is not a \"solution\" imho. All we learn from this ps is \"use dgx\". Totally changed game, right? \nIn my eyes it is not about private sharing, it is all about kicking mvp's from lb. And it would rob us from oportunity to see really best solution in the end of competition.
", "votes": 11, "replies": [] }, { "id": 605711, "author_name": "mephistopheies", "author_url": "", "post_date": "2019-08-22T16:44:30.687000", "content": "lol
", "votes": 8, "replies": [] }, { "id": 605537, "author_name": "kaerururu", "author_url": "", "post_date": "2019-08-22T13:17:08.420000", "content": "He should write his ideas about ONGOING competition in Kaggle discussions.\nIf he want to share ideas in opendatascience, write ideas here then share link.
", "votes": 8, "replies": [] }, { "id": 605589, "author_name": "phalanx", "author_url": "", "post_date": "2019-08-22T14:15:42.293000", "content": "「Я придумал. Давайте сделаем канал что-то типа #none_specific_ml_general_tricks\nНу и там все будут писать в стиле:\n\"Я тут работаю над задачей многоклассовой инстанс сегменации. и htc_dcn_x101_ibn_inshala работает охуено и по ищущению накидывает японцем за щеку\". Или \"я тут работаю над задачей сегментации легких. И у меня \". А им буду в тредиках отвечать \"о, збс, я кажется тоже работаю над такой задачей, а у меня densenet заходит\".」
\n\nI think it is so toxic.\nWhy he is kaggle grandmaster?\nWhy Kaggle overlook odsai and him?
\n\n
Are you our big brother now? It's called free speech, many people actually died for this.
", "votes": -10, "replies": [] }, { "id": 605661, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2019-08-22T15:43:50.540000", "content": "Comrade @raisefamous is ready to die for private sharing. As long as Kaggle has such (random) heroes, we will always feel safe and secure.
", "votes": -10, "replies": [] }, { "id": 605694, "author_name": "Qwer Ty", "author_url": "", "post_date": "2019-08-22T16:14:46.600000", "content": "> I'm right. Argue with me = you are a biased fanatic. I win, bye-bye.
\n\nNice logic 👍
", "votes": 1, "replies": [] }, { "id": 605696, "author_name": "Random Hero", "author_url": "", "post_date": "2019-08-22T16:17:57.177000", "content": "Calm down Ahmed, private sharing is not something I care about here. My point is about \"toxicity\", if you want to call someone a bad name on a platform where it is permitted you are free to do so.
", "votes": 3, "replies": [] }, { "id": 605713, "author_name": "Xuan Cao", "author_url": "", "post_date": "2019-08-22T16:45:15.617000", "content": "\n\n\nRandom Hero wrote:
\n \nAre you our big brother now? It's called free speech, many people actually died for this.
\n
Free speech does not mean you can say whatever you want, especially those offensive words, Comrade @raisefamous.
", "votes": 1, "replies": [] }, { "id": 605717, "author_name": "Random Hero", "author_url": "", "post_date": "2019-08-22T16:47:14.900000", "content": "it actually does
", "votes": 2, "replies": [] }, { "id": 605723, "author_name": "ykaneko", "author_url": "", "post_date": "2019-08-22T16:53:58.953000", "content": "I didn't know that @phalanx has the same power as Big Brother. \nGenerally, people who feel guilty tend to label people as “Big Brother” , anyway, if you can't read the clear bad faith in the above speech , your ideal home is in Oceania at 1984.
", "votes": 4, "replies": [] }, { "id": 605732, "author_name": "", "author_url": "", "post_date": "2019-08-22T17:06:45.100000", "content": "", "votes": -4, "replies": [] }, { "id": 605763, "author_name": "", "author_url": "", "post_date": "2019-08-22T18:07:56.283000", "content": "", "votes": 4, "replies": [] }, { "id": 605784, "author_name": "", "author_url": "", "post_date": "2019-08-22T18:31:22.763000", "content": "", "votes": -1, "replies": [] }, { "id": 605794, "author_name": "Random Hero", "author_url": "", "post_date": "2019-08-22T18:47:23", "content": ".
", "votes": -3, "replies": [] }, { "id": 605800, "author_name": "", "author_url": "", "post_date": "2019-08-22T18:51:03.810000", "content": "", "votes": 1, "replies": [] }, { "id": 605801, "author_name": "", "author_url": "", "post_date": "2019-08-22T18:53:46.230000", "content": "", "votes": 2, "replies": [] }, { "id": 605806, "author_name": "", "author_url": "", "post_date": "2019-08-22T19:02:40.600000", "content": "", "votes": -3, "replies": [] }, { "id": 605808, "author_name": "", "author_url": "", "post_date": "2019-08-22T19:11:44.337000", "content": "", "votes": 4, "replies": [] }, { "id": 605813, "author_name": "", "author_url": "", "post_date": "2019-08-22T19:18:57.190000", "content": "", "votes": -1, "replies": [] }, { "id": 605864, "author_name": "", "author_url": "", "post_date": "2019-08-22T21:00:28.120000", "content": "", "votes": -1, "replies": [] }, { "id": 605872, "author_name": "", "author_url": "", "post_date": "2019-08-22T21:22:36.963000", "content": "", "votes": -1, "replies": [] } ] }, { "id": 605573, "author_name": "phalanx", "author_url": "", "post_date": "2019-08-22T14:02:29.260000", "content": "n01z3「 текущий результат на inc seg это одна модель или ансамбль? Это модель только на 1 уровень иерархии или все?」\nzfturbo「Ансамбль и все 300 классов」\nThis is also private sharing.\n
I'd like to guard ZFTurbo here. The fact that a top model is an ensemble is not news for Kaggle. It's so in 99.95% of cases. As well as the fact that it makes use of all data.
", "votes": 19, "replies": [] }, { "id": 606066, "author_name": "Sergey Bryansky", "author_url": "", "post_date": "2019-08-23T06:36:03.340000", "content": "ZFTurbo is true top.
", "votes": 5, "replies": [] } ] }, { "id": 605106, "author_name": "mephistopheies", "author_url": "", "post_date": "2019-08-22T04:34:56.203000", "content": "hm... let me think... looks more like an envy
", "votes": -1, "replies": [ { "id": 605114, "author_name": "phalanx", "author_url": "", "post_date": "2019-08-22T04:45:08.687000", "content": "not envy.\nI want to know wheter it is private sharing or not.\nIf not, I can share my solution about current competition in kaggler-ja.
", "votes": 10, "replies": [] }, { "id": 605116, "author_name": "Bibek", "author_url": "", "post_date": "2019-08-22T04:46:43.767000", "content": "It's not about being envy or something....It's more about competing fairly within the rules of the competitions!! If everybody start sharing their secrets in some random slack channels, then people in that channel will have an unfair advantage compared to other competitors!! So letz compete hard but within the spirit of the competition
", "votes": 8, "replies": [] }, { "id": 605398, "author_name": "Kei Harada", "author_url": "", "post_date": "2019-08-22T10:51:16.140000", "content": "I can hardly understand some people(at least 5) who downvote this comment by @bibek777 .\nDo you say you want an unfair competition?
", "votes": 7, "replies": [] }, { "id": 605412, "author_name": "Kostya Kravchenko", "author_url": "", "post_date": "2019-08-22T11:07:04.177000", "content": "I say this particular example is not breaking current Kaggle rules. Moreover, sharing ideas (not code or data!) in an open community is within the spirit of Kaggle competitions and open science in general. I think, that's why the downwotes to bibek's comment.
", "votes": 9, "replies": [] }, { "id": 605413, "author_name": "Bibek", "author_url": "", "post_date": "2019-08-22T11:07:41.243000", "content": "Thank u soo much @cashfeg !!! was feeling little bit down after seeing so many downvotes :(
", "votes": 0, "replies": [] }, { "id": 605442, "author_name": "Kei Harada", "author_url": "", "post_date": "2019-08-22T11:31:49.803000", "content": "Sharing ideas here in Kaggle discussions is clearly the best for open data science and Kaggle. Isn't it?
", "votes": 1, "replies": [] }, { "id": 605452, "author_name": "Kostya Kravchenko", "author_url": "", "post_date": "2019-08-22T11:38:02.950000", "content": "Even though it might be beneficial to Kaggle as a commercial organization, attempting to force all discussions into one platform is not the best for open science.
", "votes": -2, "replies": [] }, { "id": 605488, "author_name": "Kei Harada", "author_url": "", "post_date": "2019-08-22T12:19:01.767000", "content": "Is that more important than to keep Kaggle competition fair?\nThis competition is ONGOING. \nKaggle is the organizer of this competition, neither an unrelated commercial organization, nor an unrelated platform.
", "votes": 3, "replies": [] }, { "id": 605494, "author_name": "Kostya Kravchenko", "author_url": "", "post_date": "2019-08-22T12:24:47.363000", "content": "You are mistranslating what I'm saying. Once again, in my opinion it is not breaking the Kaggle rules and is in within the spirit. Therefore it does not make this ongoing competition unfair.
", "votes": 4, "replies": [] }, { "id": 605501, "author_name": "Kei Harada", "author_url": "", "post_date": "2019-08-22T12:35:54.883000", "content": "I think it does make this competition unfair. \nTrue this sentence is not code or data, so it might be not breaking Kaggle rules, but it's almost equivalent as code. \nSince some people say twisted logics here and there, I do want Kaggle to set strict rules without gray zone.
", "votes": -2, "replies": [] }, { "id": 605534, "author_name": "sasakama", "author_url": "", "post_date": "2019-08-22T13:15:00.800000", "content": "\n\n\nideas (not code or data!)
\n
When sharing top solutions after competition, code and data are generally not shared.\nI think it's strange that there is no problem if code and data are not shared.
", "votes": 4, "replies": [] } ] }, { "id": 605739, "author_name": "Ilia Zaitsev", "author_url": "", "post_date": "2019-08-22T17:12:00.210000", "content": "Well, if you have a DGX :)
", "votes": 4, "replies": [] }, { "id": 605544, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-22T13:24:27.863000", "content": "I read all this thread and I cannot find a translation of what is quoted from ods.ai. This is really weird. Maybe there is no real info shared here after all...
", "votes": 4, "replies": [ { "id": 605554, "author_name": "Andrey Lukyanenko", "author_url": "", "post_date": "2019-08-22T13:40:20.600000", "content": "Here is a translation:
\n\nMy current result is htc_dcn_x101, TTA_minor. It trained only on the first level of hierarchy and it took ~2 weeks on DGX
Here's transcribed text from image. \nМой текущий результат зто htc_dcn_x101, TTA_minor.\nОбучался только на 1й уровень иерархии в течении~ 2 недели на DGX.
", "votes": 1, "replies": [] }, { "id": 606800, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-24T05:24:42.867000", "content": "Andrew, thanks!
", "votes": 0, "replies": [] } ] }, { "id": 605079, "author_name": "Bibek", "author_url": "", "post_date": "2019-08-22T03:18:55.687000", "content": "Some people never learn !!!
", "votes": 1, "replies": [ { "id": 605210, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2019-08-22T06:59:27.750000", "content": "No punishment = no learning\nIt is not their fault, it is Kaggle's fault.
", "votes": 21, "replies": [] }, { "id": 605499, "author_name": "mamas", "author_url": "", "post_date": "2019-08-22T12:32:09.343000", "content": "yeah, they can't backprop correctly without the correct target.
", "votes": 8, "replies": [] }, { "id": 605621, "author_name": "Miyatti", "author_url": "", "post_date": "2019-08-22T14:54:50.107000", "content": "Looks like his learning rate is completely 0.
", "votes": 9, "replies": [] } ] }, { "id": 605666, "author_name": "Yiheng Wang", "author_url": "", "post_date": "2019-08-22T15:54:51.757000", "content": "Yearly Hot Kaggle Word: \"X5\"
", "votes": 2, "replies": [ { "id": 605674, "author_name": "Bibek", "author_url": "", "post_date": "2019-08-22T16:00:35.690000", "content": "", "votes": -9, "replies": [] } ] }, { "id": 605179, "author_name": "Kostya Kravchenko", "author_url": "", "post_date": "2019-08-22T06:25:13.547000", "content": "@phalanx if it is private sharing, how in fact did you get to this info? 🤔
", "votes": -7, "replies": [ { "id": 605191, "author_name": "ONODERA", "author_url": "", "post_date": "2019-08-22T06:48:38.560000", "content": "If you want to say \"slack is public\", it's outdated discussion.\nRefer to https://www.kaggle.com/c/instant-gratification/discussion/94894#549365
", "votes": 17, "replies": [] } ] }, { "id": 605373, "author_name": "IggiSv9t", "author_url": "", "post_date": "2019-08-22T10:04:57.197000", "content": "It's a new easy way to became a discussion master.
", "votes": -6, "replies": [] }, { "id": 607179, "author_name": "Ravi Sharma", "author_url": "", "post_date": "2019-08-24T18:16:08.980000", "content": "I think they have to work with the sample submission data, by submitting that data you directly get the result and i think most of the people do that because i used to do it in my early time in kaggle but then i realize that by doing that what you are getting with it so i took a time around a year to learn as much as i can.
", "votes": -1, "replies": [] }, { "id": 623642, "author_name": "Harshit", "author_url": "", "post_date": "2019-09-11T07:06:27.170000", "content": "Hello Guys, I am looking for teammates.
", "votes": 0, "replies": [] }, { "id": 606887, "author_name": "", "author_url": "", "post_date": "2019-08-24T08:59:18.110000", "content": "", "votes": 0, "replies": [] }, { "id": 606199, "author_name": "", "author_url": "", "post_date": "2019-08-23T10:06:01.240000", "content": "", "votes": 0, "replies": [] }, { "id": 605776, "author_name": "", "author_url": "", "post_date": "2019-08-22T18:21:41.820000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "605617": "What matters most is healthy competitions and community. Many people compete at Kaggle because getting a Kaggle medal is one of the most prestigious honor in the data science community. It differentiates Kaggle from other competition platforms, and indeed, that is why I participate in Kaggle.\n\nWhat if unfair contestants win over fair contestants? What if Kaggle medals are considered to indicate cheating ability instead of data science and machine learning ability? The outcome is trivial.\n\nAdmins, please think seriously about dishonest acts. They are about to severely destroy the value of Kaggle.", "605746": "The real private sharing is the one you'll never know about. And what you guys are obsessed with is a modern witch-hunting.", "605481": "Based on experience, I would say it is permitted from kaggle - BUT (big but), **it shouldn't be**", "605780": "Kaggle openly promotes learning and sharing across our community, and that extends to countless other media that have spawned from users' love of Kaggle (meetups, social media groups, slack channels, etc). What is prohibited in our rules is private sharing. \n\nSharing occurs on a spectrum, and for us \"private sharing\" goes beyond non-specific ideas and involves exact models & parameters that would be privileged information for the people who created them. For this reason, **we strongly advise any sharing that could be interpreted as private happen on Kaggle's forums or isolated within the team you are working with**.\n\nThe definition of sharing and what ventures into the realm of private will always be a grey area - but we are all professionals with a mutual goal and should exercise ourselves with a base level of civility & respect. We at Kaggle want to host competitions that bring our community together around a common interest - not create an environment that devolves into a thread like this. Let this also be a reminder of [Kaggle's community guidelines](https://www.kaggle.com/general/60914).\n\nLet's all focus on real data science work and discussions. Violations of Kaggle's rules are reviewed through our [compliance channel](https://www.kaggle.com/compliance). Clear, conclusive evidence should be provided if making a claim. Every claim is read and valid claims are handled accordingly. Note that action on substantiated claims may not be taken until the close of the competition, as part of our standard violators removal process.", "605061": "I found it in opendatascience.\nI think it is private sharing.\nis it permitted in kaggle?\n[@philculliton](https://www.kaggle.com/philculliton), [@juliaelliott](https://www.kaggle.com/juliaelliott) Please answer this question.\n\n", "606528": "What bothers me much more is the fact that you need to train for two weeks on a fat cluster for this competition.", "605502": "He is Private Sharing Grandmaster.", "605728": "Some new crazy private sharing from ods.ai:\n\n\n", "605559": "All Kagglers should write ideas, code, discussion of ongoing competitions in Kaggle platform, if you want to share those other people.\nwe cannot trace information of competiton. because we cannot check all communication tools.\nThat situation cause lack of fairness for competitors. ", "605672": "I believe that this insight doesn't pump ts up on leaderboard. This is not a \"solution\" imho. All we learn from this ps is \"use dgx\". Totally changed game, right? \nIn my eyes it is not about private sharing, it is all about kicking mvp's from lb. And it would rob us from oportunity to see really best solution in the end of competition. ", "605711": "lol", "605537": "He should write his ideas about ONGOING competition in Kaggle discussions.\nIf he want to share ideas in opendatascience, write ideas here then share link.", "605589": "「Я придумал. Давайте сделаем канал что-то типа #none_specific_ml_general_tricks\nНу и там все будут писать в стиле:\n\"Я тут работаю над задачей многоклассовой инстанс сегменации. и htc_dcn_x101_ibn_inshala работает охуено и по ищущению накидывает японцем за щеку\". Или \"я тут работаю над задачей сегментации легких. И у меня \". А им буду в тредиках отвечать \"о, збс, я кажется тоже работаю над такой задачей, а у меня densenet заходит\".」\n\nI think it is so toxic.\nWhy he is kaggle grandmaster?\nWhy Kaggle overlook odsai and him?\n\n\n\n", "605573": "n01z3「 текущий результат на inc seg это одна модель или ансамбль? Это модель только на 1 уровень иерархии или все?」\nzfturbo「Ансамбль и все 300 классов」\nThis is also private sharing.\n\n", "605106": "hm... let me think... looks more like an envy", "605739": "Well, if you have a DGX :)", "605544": "I read all this thread and I cannot find a translation of what is quoted from ods.ai. This is really weird. Maybe there is no real info shared here after all...\n\n", "605079": "Some people never learn !!!", "605666": "Yearly Hot Kaggle Word: \"X5\"", "605179": "@phalanx *if* it is private sharing, how in fact did *you* get to this info? 🤔 ", "605373": "It's a new easy way to became a discussion master.", "607179": "I think they have to work with the sample submission data, by submitting that data you directly get the result and i think most of the people do that because i used to do it in my early time in kaggle but then i realize that by doing that what you are getting with it so i took a time around a year to learn as much as i can.", "623642": "Hello Guys, I am looking for teammates.\n", "606887": "", "606199": "", "605776": "" } }