![\"kaggle_mem\"](\"https://i.ibb.co/zGzM1Dc/photo-2021-02-04-21-42-22.jpg\")
Hi Kaggle community,
\nA few days ago some respectful kaggler @aerdem4 accused me and my teammates in rules violation and cheating. Before writing this post, we asked the Kaggle administration to clarify this incident, but there was no clear answer from their side, so I will try to explain how our team sees this situation.
\nIt is obvious that we were going to merge anyway. If you will look at the competition history of @yakuben, @vladimirsydor, @abazdyrev, @frednavruzov, you will see there were a lot of common/in-class competitions where they were together in some combinations. Moreover, @snakayama texted to @frednavruzov 2 weeks ago and asked to merge. @frednavruzov politely refused that offer and mentioned in his response that he already has teammates and even listed the names above. Unfortunately we can’t attach the screenshot because it was sent via Kaggle, so the response is only available on @snakayama email and at Kaggle’s data storage.
\nSo yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. So we haven’t violated the rules. The reason not merging at the beginning is because everyone does his own research. Also it helps to make sure that everyone makes his impact in competition. I’m sure everyone has faced at least once such a situation when you start to work with someone in a team, but in result do all the hard work alone. Also, we haven’t got any advantage over other teams from the submissions number side. We have LESS team submissions then even some solo competitors.
\nObviously, for many team merging processes some kind of shady \"private sharing\" may take place. As was mentioned in the previous discussion people usually ask some basic questions before they merge (e.g. is your score based on SED or non-SED), to know whether they would benefit from a merge and ensemble. In fact, such kind of private sharing gives you almost no benefit if all people that shared will merge after this. It's a completely different matter when people after this kind of things don't merge to a team. But, why then respectful grandmaster @aerdem4 decided to show his deductive skills BEFORE the merge deadline. If he had a goal to punish cheaters he would post it after the merge deadline and the guilt of the corresponding people would be obvious. But in this case it looks more like upvotes collection and hype around him. We could bring similar charges against him. We see a lot of people at the top from Nvidia. Also, we see that @aerdem4, @christofhenkel and @cpmpml (who scored 0.931 from the 1st submit!!!) were teammates before, they all are from Nvidia and, we are sure, it may be some situations when they were submitting submissions at the same time with different, but close scores. But we are not raising hype around it. Because it sounds ridiculous and we trust their integrity.
\nDear Kaggle community, please wait for rock solid evidence before you blame someone publicly. If you really feel that someone is cheating, you should rather use appropriate form https://www.kaggle.com/compliance then starting toxic threads.
\n
P.S
\nif you like this post, please downvote it :-)
Jumping in here to reiterate that Kaggle's official stance on this matter is that sharing among participants should only happen if they are joined as a team or if it is done so publicly for all participants to enjoy. To be clear, this means that if you share with individuals whom you are not teamed up with, you are in violation of this rule, even if you later team. We are detecting and acting on private sharing offenders as part of our competition close process/clean-up.
\nWe also encourage compliance matters such as this be raised directly with us. We review all reports and take action where evidence is conclusive.
", "rawMarkdown": "Jumping in here to reiterate that Kaggle's official stance on this matter is that sharing among participants should only happen if they are joined as a team or if it is done so publicly for all participants to enjoy. To be clear, this means that if you share with individuals whom you are not teamed up with, you are in violation of this rule, *even if* you later team. We are detecting and acting on private sharing offenders as part of our competition close process/clean-up.\n\nWe also encourage compliance matters such as this be [raised directly with us](https://www.kaggle.com/compliance). We review all reports and take action where evidence is conclusive.", "votes": 31, "replies": [ { "id": 1193182, "postDate": "2021-02-09T13:54:48.930Z", "content": "@juliaelliott I think Kaggle needs more explicit rules.
", "rawMarkdown": "@juliaelliott I think Kaggle needs more explicit rules. ", "votes": 5 } ] }, { "id": 1186526, "postDate": "2021-02-04T21:52:04.927Z", "content": "\n\n@cpmpml (who scored 0.931 from the 1st submit!!!)
\n
I am quite proud of that actually. Thanks for highlighting it.
", "rawMarkdown": "> @cpmpml (who scored 0.931 from the 1st submit!!!)\n\nI am quite proud of that actually. Thanks for highlighting it.", "votes": 22 }, { "id": 1186530, "postDate": "2021-02-04T21:55:07.240Z", "content": "Admin says; You must merge within the Kaggle platform before sharing any private competition information to be considered in adherence to competition rules.
\nhttps://www.kaggle.com/c/rfcx-species-audio-detection/discussion/197744
Contraposition is;
\nIf you don't care about breaking the competition rules, you don't need to merge within the Kaggle platform before sharing any private competition information.
Oh why downvote?? I just repeated what admin said and showed contraposition for easier understanding.
", "rawMarkdown": "Oh why downvote?? I just repeated what admin said and showed contraposition for easier understanding.", "votes": 7 }, { "id": 1186569, "postDate": "2021-02-04T22:39:51.923Z", "content": "@onodera look what the actual rules of the competition say regarding this situation:
\n\n\nNo private sharing outside teams
\n
\n Privately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.Team Mergers
\n
\n Team mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.
We didn't share anything outside our team, so we didn't violate the No private sharing outside teams rules block. Since we've managed to merge we didn't not violate the Team Mergers rules block.
\nThere is no rule or requirement to read all kaggle discussions and to consider everything that admins say (in discussion, but not in the rule page) as a rock solid rule. To be honest, We didn't even read the mentioned thread before your reply, because actually we had an agreement to team up in this particular competition.
\nBTW, if kaggle use mentioned quote as a ban-rule, then I assume that many teams should be banned, but in fact there is no accurate way to prove that there was some kind of \"private sharing\" before the merge, if there is no submissions from different accounts using the same IP or identical submissions from different users. So my personal point is that this kind of selective attacks using mentioned quote (just a discussion topic and not the rule) seems like a toxic witch-hunting
", "rawMarkdown": "@onodera look what the actual [rules](https://www.kaggle.com/c/rfcx-species-audio-detection/rules) of the competition say regarding this situation:\n> **No private sharing outside teams**\nPrivately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.\n\n> **Team Mergers**\nTeam mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.\n\nWe didn't share anything outside our team, so we didn't violate the **No private sharing outside teams** rules block. Since we've managed to merge we didn't not violate the **Team Mergers** rules block.\n\nThere is no rule or requirement to read all kaggle discussions and to consider everything that admins say (in discussion, but not in the rule page) as a rock solid rule. To be honest, We didn't even read the mentioned [thread](https://www.kaggle.com/c/rfcx-species-audio-detection/discussion/197744) before your reply, because actually we had an agreement to team up in this particular competition.\n\nBTW, if kaggle use mentioned quote as a ban-rule, then I assume that many teams should be banned, but in fact there is no accurate way to prove that there was some kind of \"private sharing\" before the merge, if there is no submissions from different accounts using the same IP or identical submissions from different users. So my personal point is that this kind of selective attacks using mentioned quote (just a discussion topic and not the rule) seems like a toxic witch-hunting", "votes": 5 }, { "id": 1186688, "postDate": "2021-02-05T02:02:44.173Z", "content": "\n\nWe didn't share anything outside our team, so we didn't violate the No private sharing outside teams rules block. Since we've managed to merge we didn't not violate the Team Mergers rules block.
\n
It depends on how Kaggle admin sees the definition of a \"Team\". If that definition is \"A group of people who have decided to create a merged team on kaggle platform before merger deadline\" then it's ok, but personally I don't think so.
", "rawMarkdown": "> We didn't share anything outside our team, so we didn't violate the No private sharing outside teams rules block. Since we've managed to merge we didn't not violate the Team Mergers rules block.\n\nIt depends on how Kaggle admin sees the definition of a \"Team\". If that definition is \"A group of people who have decided to create a merged team on kaggle platform before merger deadline\" then it's ok, but personally I don't think so.", "votes": 16 } ] }, { "id": 1187793, "postDate": "2021-02-05T18:06:40.490Z", "content": "![\"\"](\"https://media.tenor.com/images/c1707fa67f9c7d5d6a0bb1dae682a1af/tenor.gif\")
\n\nSo yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning.
\nSo we haven’t violated the rules.
\n
Why do you think that? Now the question is what the rules really are, I am not sure myself. You are right that it has been tolerated before.
", "rawMarkdown": "> So yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. \n\n> So we haven’t violated the rules.\n\nWhy do you think that? Now the question is what the rules really are, I am not sure myself. You are right that it has been tolerated before.", "votes": 17 }, { "id": 1186545, "postDate": "2021-02-04T22:08:59.627Z", "content": "Just out of curiousity I checked the official rules. (Did not read through the small letter sections…)
\n\n\nNo private sharing outside teams
\n
\n Privately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.
I have a feeling about the spirit of the rules and I try to avoid to share event the slightest hint like a paranoid.
\nSome clarification from @kaggle would be useful especially for the future.
This was answered by @juliaelliott here: https://www.kaggle.com/c/rfcx-species-audio-detection/discussion/197744
\nIt reads:
\n\n\nPlease note: You must merge within the Kaggle platform before sharing any private competition information to be considered in adherence to competition rules.
\n
It would be much better if this text was part of the rules. Not everyone reads all forum topics.
", "rawMarkdown": "This was answered by @juliaelliott here: https://www.kaggle.com/c/rfcx-species-audio-detection/discussion/197744\n\nIt reads:\n\n> Please note: You must merge within the Kaggle platform before sharing any private competition information to be considered in adherence to competition rules.\n\nIt would be much better if this text was part of the rules. Not everyone reads all forum topics.", "votes": 6 } ] }, { "id": 1186446, "postDate": "2021-02-04T20:30:59.553Z", "content": "Hi Kaggle community,
\nA few days ago some respectful kaggler @aerdem4 accused me and my teammates in rules violation and cheating. Before writing this post, we asked the Kaggle administration to clarify this incident, but there was no clear answer from their side, so I will try to explain how our team sees this situation.
\nIt is obvious that we were going to merge anyway. If you will look at the competition history of @yakuben, @vladimirsydor, @abazdyrev, @frednavruzov, you will see there were a lot of common/in-class competitions where they were together in some combinations. Moreover, @snakayama texted to @frednavruzov 2 weeks ago and asked to merge. @frednavruzov politely refused that offer and mentioned in his response that he already has teammates and even listed the names above. Unfortunately we can’t attach the screenshot because it was sent via Kaggle, so the response is only available on @snakayama email and at Kaggle’s data storage.
\nSo yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. So we haven’t violated the rules. The reason not merging at the beginning is because everyone does his own research. Also it helps to make sure that everyone makes his impact in competition. I’m sure everyone has faced at least once such a situation when you start to work with someone in a team, but in result do all the hard work alone. Also, we haven’t got any advantage over other teams from the submissions number side. We have LESS team submissions then even some solo competitors.
\nObviously, for many team merging processes some kind of shady \"private sharing\" may take place. As was mentioned in the previous discussion people usually ask some basic questions before they merge (e.g. is your score based on SED or non-SED), to know whether they would benefit from a merge and ensemble. In fact, such kind of private sharing gives you almost no benefit if all people that shared will merge after this. It's a completely different matter when people after this kind of things don't merge to a team. But, why then respectful grandmaster @aerdem4 decided to show his deductive skills BEFORE the merge deadline. If he had a goal to punish cheaters he would post it after the merge deadline and the guilt of the corresponding people would be obvious. But in this case it looks more like upvotes collection and hype around him. We could bring similar charges against him. We see a lot of people at the top from Nvidia. Also, we see that @aerdem4, @christofhenkel and @cpmpml (who scored 0.931 from the 1st submit!!!) were teammates before, they all are from Nvidia and, we are sure, it may be some situations when they were submitting submissions at the same time with different, but close scores. But we are not raising hype around it. Because it sounds ridiculous and we trust their integrity.
\nDear Kaggle community, please wait for rock solid evidence before you blame someone publicly. If you really feel that someone is cheating, you should rather use appropriate form https://www.kaggle.com/compliance then starting toxic threads.
\n
P.S
\nif you like this post, please downvote it :-)
… Did you just make a thread to admit that you were sharing information before merging?
", "rawMarkdown": "... Did you just make a thread to admit that you were sharing information before merging?", "votes": 12, "replies": [ { "id": 1188452, "postDate": "2021-02-06T08:58:32.527Z", "content": "We just clarified our position on this issue
", "rawMarkdown": "We just clarified our position on this issue", "votes": 3 } ] }, { "id": 1186787, "postDate": "2021-02-05T03:28:13.607Z", "content": "I think you're right, and people should stop making these witchhunting threads. Calling \"cheaters\" out in discussion threads is just trying to start a fight and publicly hurts the legitimacy of the website. If you think someone is cheating, fill out a form at https://www.kaggle.com/compliance and read this post https://www.kaggle.com/general/125466
", "rawMarkdown": "I think you're right, and people should stop making these witchhunting threads. Calling \"cheaters\" out in discussion threads is just trying to start a fight and publicly hurts the legitimacy of the website. If you think someone is cheating, fill out a form at https://www.kaggle.com/compliance and read this post https://www.kaggle.com/general/125466", "votes": 11, "replies": [ { "id": 1186866, "postDate": "2021-02-05T05:02:52.033Z", "rawMarkdown": "", "isDeleted": true }, { "id": 1187030, "postDate": "2021-02-05T07:15:22.863Z", "content": "So you think it is better if they spend one month in this competition and get removed? Because Kaggle only removes teams at the end. Then both their efforts are wasted and people below them got demotivated. Would you be more motivated if you were 30th or 20th? I obviously don't try to start a fight, otherwise I would be fighting. Please read the original post carefully. I am more concerned about rights of the people who don't cheat.
", "rawMarkdown": "So you think it is better if they spend one month in this competition and get removed? Because Kaggle only removes teams at the end. Then both their efforts are wasted and people below them got demotivated. Would you be more motivated if you were 30th or 20th? I obviously don't try to start a fight, otherwise I would be fighting. Please read the original post carefully. I am more concerned about rights of the people who don't cheat.", "votes": 1 } ] }, { "id": 1186829, "postDate": "2021-02-05T04:28:09.497Z", "content": "My simple question is, why didn't you merge within the Kaggle platform from the beginning?
\nSorry, I cannot fully understand the reason you described above.
\nBy early team-merging, we can share any information freely, and also decide what to share or what not to share in order to make great impact in the end (in terms of ensemble, heterogeneity, diversity, etc).
We, as people, who are friends in real life, decided to work on this competition together (from the very beginning). But, taking into account the probability that some of us may not have enough time to work on this task, we decided to wait until each of us contributes to the score and merge closer to the merge deadline, when it is clear who wants (and who is actually able) to participate and to bring value. If we would merge at the beginning we would probably 1) lose spare team places for nothing and 2) acting as an escalator, let someone stay idle but still get a medal / high score, which is not quite fair, honestly, even for friends.
\nАnd since some of us only started doing something substantial a few days ago (to stop considering themselves freeloaders), we didn’t team up - there was a lot of time left until the merge deadline.
", "rawMarkdown": "We, as people, who are friends in real life, decided to work on this competition together (from the very beginning). But, taking into account the probability that some of us may not have enough time to work on this task, we decided to wait until each of us contributes to the score and merge closer to the merge deadline, when it is clear who wants (and who is actually able) to participate and to bring value. If we would merge at the beginning we would probably 1) lose spare team places for nothing and 2) acting as an escalator, let someone stay idle but still get a medal / high score, which is not quite fair, honestly, even for friends.\n\nАnd since some of us only started doing something substantial a few days ago (to stop considering themselves freeloaders), we didn’t team up - there was a lot of time left until the merge deadline.", "votes": 2 }, { "id": 1188466, "postDate": "2021-02-06T09:09:00.913Z", "content": "@yakuben you could do this without sharing code and ideas. Everyone could work independent and then merge.
\nAnd I recommend you to be careful with such agreements. Even though they are your friends, some of them could discover an additional trick and get tempted by solo gold and could continue solo despite you agreed to merge and shared information already.
", "rawMarkdown": "@yakuben you could do this without sharing code and ideas. Everyone could work independent and then merge.\n\nAnd I recommend you to be careful with such agreements. Even though they are your friends, some of them could discover an additional trick and get tempted by solo gold and could continue solo despite you agreed to merge and shared information already.", "votes": 6 }, { "id": 1188478, "postDate": "2021-02-06T09:27:03.170Z", "content": "Thanks for your sincere reply to my question.
\nNow I understand your situation.
\nIn that case, after promising to team up together, not sharing any information each other about the ongoing competition before merging seems to be the best way we can take, which is my interpretation though.
@aerdem4
\nUntil recently we do this without sharing code and ideas.
\nBut when we consider ourselves a team, it gets hard not to share something.
\nI think that from our answers it becomes clear that we did not think that we were breaking the rules or received any advantages over other participants. Therefore, we did not think a lot about this situation and just waited for everyone to join.
About the situation you described above.
\n1) I am 100% confident in my teammates
\n2) If this situation happened, it would be cheating. And our team would clearly not have tolerated this (at least I can speak for myself).
It is your responsibility to read the rules. Unfortunately, thinking that you were not breaking the rules is not enough. You should have researched the rules or at least think \"so why do other people team up early?\". If your definition of team was right, everyone would team up in the last day.
\nAnyway, I don't want anyone's efforts to be wasted. So you being removed from LB and you being above other people who obeyed the rules both make me sad. Instead of having such defensive (also offensive) post shared by Dima, an apology post could be more meaningful.
\nAnd I believe you will learn to not be 100% confident with people over time. Life teaches it:)
", "rawMarkdown": "It is your responsibility to read the rules. Unfortunately, thinking that you were not breaking the rules is not enough. You should have researched the rules or at least think \"so why do other people team up early?\". If your definition of team was right, everyone would team up in the last day.\n\nAnyway, I don't want anyone's efforts to be wasted. So you being removed from LB and you being above other people who obeyed the rules both make me sad. Instead of having such defensive (also offensive) post shared by Dima, an apology post could be more meaningful.\n\nAnd I believe you will learn to not be 100% confident with people over time. Life teaches it:)", "votes": 5 }, { "id": 1188676, "postDate": "2021-02-06T12:51:04.027Z", "content": "1) Regarding whether we broke the rules or not.
\nI believe that we did not break the rules. I agree that what is written in the rules can be interpreted in different ways. I suspect that what is written in the discussion more a recommendation than a rule for two reasons:
In any case, it is up to the administration to decide, regardless of what the participants write here.
\n2) About \"defensive\" part of Dima post
\nIn his post, he wrote the truth about this situation from our side. He did not write something in the spirit of \"this is an accident\" or some other kind of excuses. He wrote everything as it is. And clarify our position that everything was fair play.
3) About \"offensive\" part of Dima post
\nRegardless of what the administration decides on our situation. I believe that creating posts with public accusations without serious evidence (especially before the merge deadline) (especially with evidence like same country/same university/teammates before ) is not ok. And this applies not only to kaggle, but also to social networks.
\n\nBut when we consider ourselves a team
\n
then you make it official and merge on Kaggle. ;)
", "rawMarkdown": "> But when we consider ourselves a team\n\nthen you make it official and merge on Kaggle. ;)", "votes": 9 }, { "id": 1188721, "postDate": "2021-02-06T13:16:54.940Z", "content": "\n\nRegardless of what the administration decides on our situation. I believe that creating posts with public accusations without serious evidence (especially before the merge deadline) (especially with evidence like same country/same university/teammates before ) is not ok. And this applies not only to kaggle, but also to social networks.
\n
I accused 2 people for private sharing and they confessed that. So I was right. Otherwise, I would apologize.
\n\n\nBut in this case it looks more like upvotes collection and hype around him.
\n
Your teammate accused me on having childish intentions. I don't care about discussion upvotes or me getting popular. I was trying to make Kaggle more fair even before I was a GM. Repeating \"respectful grandmaster\" on whole text and GM comparison between past and now have no point. It was only for taunting me.
\n\n\nWe could bring similar charges against him. We see a lot of people at the top from Nvidia. Also, we see that @aerdem4, @christofhenkel and @cpmpml (who scored 0.931 from the 1st submit!!!) were teammates before, they all are from Nvidia and, we are sure, it may be some situations when they were submitting submissions at the same time with different, but close scores.
\n
Please do so when we are at the same score and submit at the same time. I would be happy to share my solution with Kaggle. Attacking us won't make you guys right. You admit your guilt, immediately merged teams after the post and your score did not even improve after the merge since your solutions were probably almost identical.
\nI tried to be constructive in my comments but I will stop trying and focus on the competition. Kaggle will decide if it was a private sharing or not.
", "rawMarkdown": "> Regardless of what the administration decides on our situation. I believe that creating posts with public accusations without serious evidence (especially before the merge deadline) (especially with evidence like same country/same university/teammates before ) is not ok. And this applies not only to kaggle, but also to social networks.\n\nI accused 2 people for private sharing and they confessed that. So I was right. Otherwise, I would apologize. \n\n> But in this case it looks more like upvotes collection and hype around him.\n\nYour teammate accused me on having childish intentions. I don't care about discussion upvotes or me getting popular. I was trying to make Kaggle more fair even before I was a GM. Repeating \"respectful grandmaster\" on whole text and GM comparison between past and now have no point. It was only for taunting me.\n\n> We could bring similar charges against him. We see a lot of people at the top from Nvidia. Also, we see that @aerdem4, @christofhenkel and @cpmpml (who scored 0.931 from the 1st submit!!!) were teammates before, they all are from Nvidia and, we are sure, it may be some situations when they were submitting submissions at the same time with different, but close scores.\n\nPlease do so when we are at the same score and submit at the same time. I would be happy to share my solution with Kaggle. Attacking us won't make you guys right. You admit your guilt, immediately merged teams after the post and your score did not even improve after the merge since your solutions were probably almost identical.\n\nI tried to be constructive in my comments but I will stop trying and focus on the competition. Kaggle will decide if it was a private sharing or not.", "votes": 1 }, { "id": 1188820, "postDate": "2021-02-06T14:37:19.723Z", "content": "@yakuben All your explanation is okay, but as soon as you share ideas / concepts / codes you should merge on Kaggle. You even admit that you considered yourselves as a team from that point on.
\nI think it is pretty obvious that you get a clear advantage by waiting as long as possible to merge, namely additional submissions. Additional submissions can be more and more important the closer you get to the end of the competition and the higher you potentially are on the LB, so don't start with \"we did not exceed the overall limit otherwise we could not merge\". Everyone who has worked in serious teams knows that deciding on submissions per day can lead to many discussions and you have to put things on the side. So overall, it should be clear to you guys that what you did was shady and definitely on the edge of the spectrum of rules, regardless if it has been accepted in the past. So an apology post would have been far better for the situation.
", "rawMarkdown": "@yakuben All your explanation is okay, but as soon as you share ideas / concepts / codes you should merge on Kaggle. You even admit that you considered yourselves as a team from that point on. \n\nI think it is pretty obvious that you get a clear advantage by waiting as long as possible to merge, namely additional submissions. Additional submissions can be more and more important the closer you get to the end of the competition and the higher you potentially are on the LB, so don't start with \"we did not exceed the overall limit otherwise we could not merge\". Everyone who has worked in serious teams knows that deciding on submissions per day can lead to many discussions and you have to put things on the side. So overall, it should be clear to you guys that what you did was shady and definitely on the edge of the spectrum of rules, regardless if it has been accepted in the past. So an apology post would have been far better for the situation.", "votes": 16 }, { "id": 1189062, "postDate": "2021-02-06T17:33:05.020Z", "content": "I concur on the number of subs. All teams in top 10 currently have used at least 25 more subs than me, and some hundreds more.
\nI'd welcome a daily allowance of 10 subs now ;)
\nMaybe I should team with my wife account without teaming officially till one week end? And even that way I won't catch up in number of subs.
\nNB. The trick of teaming with one's spouse/partner account has been used by some Kaggle GM. I didn't pick it at random.
", "rawMarkdown": "I concur on the number of subs. All teams in top 10 currently have used at least 25 more subs than me, and some hundreds more. \n\nI'd welcome a daily allowance of 10 subs now ;) \n\nMaybe I should team with my wife account without teaming officially till one week end? And even that way I won't catch up in number of subs.\n\nNB. The trick of teaming with one's spouse/partner account has been used by some Kaggle GM. I didn't pick it at random.", "votes": 1 }, { "id": 1189063, "postDate": "2021-02-06T17:33:40.417Z", "content": "But I prefer to focus on finding what I missed so far compared to those who passed me. ;)
", "rawMarkdown": "But I prefer to focus on finding what I missed so far compared to those who passed me. ;)", "votes": 2 }, { "id": 1189105, "postDate": "2021-02-06T18:01:59.170Z", "content": "\n\nI concur on the number of subs. All teams in top 10 currently have used at least 25 more subs than me, and some hundreds more.
\n
Actually I got 0.970 from around 30 subs. With other 100 I tried differrent models and approaches but all of them hit the wall at this score or below.
\nSo, unfortunately, the number of subs cannot be directly converted to the public score, otherwise I would score 0.980+ by now😀
Regarding the topic: from Kaggle perspective obviously team is a clear concept as even one person is a team with one team member. Five solo competitors means five different teams until they merge.
", "rawMarkdown": "> I concur on the number of subs. All teams in top 10 currently have used at least 25 more subs than me, and some hundreds more.\n\nActually I got 0.970 from around 30 subs. With other 100 I tried differrent models and approaches but all of them hit the wall at this score or below. \nSo, unfortunately, the number of subs cannot be directly converted to the public score, otherwise I would score 0.980+ by now😀\n\nRegarding the topic: from Kaggle perspective obviously team is a clear concept as even one person is a team with one team member. Five solo competitors means five *different* teams until they merge. \n", "votes": 6 }, { "id": 1189168, "postDate": "2021-02-06T19:05:55.817Z", "content": "Good to know, because you had 67 subs when I joined. I assumed you need all of them to get to 970.
", "rawMarkdown": "Good to know, because you had 67 subs when I joined. I assumed you need all of them to get to 970." }, { "id": 1192058, "postDate": "2021-02-08T22:00:03.220Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 1190105, "postDate": "2021-02-07T13:42:06.880Z", "content": "Sometimes social battles in Kaggle competitions are more impressive than the competitions itself.
\nReady for great melodrama 😎
", "rawMarkdown": "Sometimes social battles in Kaggle competitions are more impressive than the competitions itself.\n\nReady for great melodrama 😎", "votes": 8 }, { "id": 1209946, "postDate": "2021-02-19T05:54:14.070Z", "content": "This didn't age well now, did it?
", "rawMarkdown": "This didn't age well now, did it?", "votes": 4 }, { "id": 1186460, "postDate": "2021-02-04T20:39:49.753Z", "content": "Well, you just started another toxic thread.
", "rawMarkdown": "Well, you just started another toxic thread.", "votes": 3 }, { "id": 1186520, "postDate": "2021-02-04T21:45:16.463Z", "content": "Didn't read after
\n\n", "rawMarkdown": "Didn't read after \n> So yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. So we haven’t violated the rules.", "votes": 2 }, { "id": 1188193, "postDate": "2021-02-06T03:51:21.360Z", "content": "So yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. So we haven’t violated the rules.
\n
Rule is much simpler than competition itself. You cheated for sure. If I were you I would had started a apologize thread instead.
", "rawMarkdown": "Rule is much simpler than competition itself. You cheated for sure. If I were you I would had started a apologize thread instead. ", "votes": 3, "replies": [ { "id": 1188678, "postDate": "2021-02-06T12:52:09.503Z", "content": "We don't think so
", "rawMarkdown": "We don't think so", "votes": -2 }, { "id": 1188730, "postDate": "2021-02-06T13:23:57.740Z", "content": "\n\nSo yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. So we haven’t violated the rules.
\n
That means you share from the early beginning, while you merge days ago.
\nInfact kagglers really don't care whether you cheat, we just want to keep kaggle fair still.
\nIf we want, we all could think up more excuses than you:)) to prove we really don't private sharing and merge at teammerge deadline. While we mostly prefer a simpler way, team once we choose our teammate. All things about how teammates perform along the competion or some teammates just disappearing is our choose, we shall be responsible for every decision we make.
Ps: \"be responsible\" is the second important thing you shall learn.
\nSry if something toxic above, I'm not good at express myself in English.
\nOh look that I just find an excuse for my impoliteness.
Jumping in here to reiterate that Kaggle's official stance on this matter is that sharing among participants should only happen if they are joined as a team or if it is done so publicly for all participants to enjoy. To be clear, this means that if you share with individuals whom you are not teamed up with, you are in violation of this rule, even if you later team. We are detecting and acting on private sharing offenders as part of our competition close process/clean-up.
\nWe also encourage compliance matters such as this be raised directly with us. We review all reports and take action where evidence is conclusive.
", "votes": 31, "replies": [ { "id": 1193182, "author_name": "Mukharbek Organokov", "author_url": "", "post_date": "2021-02-09T13:54:48.930000", "content": "@juliaelliott I think Kaggle needs more explicit rules.
", "votes": 5, "replies": [] } ] }, { "id": 1186526, "author_name": "CPMP", "author_url": "", "post_date": "2021-02-04T21:52:04.927000", "content": "\n\n@cpmpml (who scored 0.931 from the 1st submit!!!)
\n
I am quite proud of that actually. Thanks for highlighting it.
", "votes": 22, "replies": [] }, { "id": 1186530, "author_name": "ONODERA", "author_url": "", "post_date": "2021-02-04T21:55:07.240000", "content": "Admin says; You must merge within the Kaggle platform before sharing any private competition information to be considered in adherence to competition rules.
\nhttps://www.kaggle.com/c/rfcx-species-audio-detection/discussion/197744
Contraposition is;
\nIf you don't care about breaking the competition rules, you don't need to merge within the Kaggle platform before sharing any private competition information.
Oh why downvote?? I just repeated what admin said and showed contraposition for easier understanding.
", "votes": 7, "replies": [] }, { "id": 1186569, "author_name": "Anton Bazdyrev", "author_url": "", "post_date": "2021-02-04T22:39:51.923000", "content": "@onodera look what the actual rules of the competition say regarding this situation:
\n\n\nNo private sharing outside teams
\n
\n Privately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.Team Mergers
\n
\n Team mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.
We didn't share anything outside our team, so we didn't violate the No private sharing outside teams rules block. Since we've managed to merge we didn't not violate the Team Mergers rules block.
\nThere is no rule or requirement to read all kaggle discussions and to consider everything that admins say (in discussion, but not in the rule page) as a rock solid rule. To be honest, We didn't even read the mentioned thread before your reply, because actually we had an agreement to team up in this particular competition.
\nBTW, if kaggle use mentioned quote as a ban-rule, then I assume that many teams should be banned, but in fact there is no accurate way to prove that there was some kind of \"private sharing\" before the merge, if there is no submissions from different accounts using the same IP or identical submissions from different users. So my personal point is that this kind of selective attacks using mentioned quote (just a discussion topic and not the rule) seems like a toxic witch-hunting
", "votes": 5, "replies": [] }, { "id": 1186688, "author_name": "Hidehisa Arai", "author_url": "", "post_date": "2021-02-05T02:02:44.173000", "content": "\n\nWe didn't share anything outside our team, so we didn't violate the No private sharing outside teams rules block. Since we've managed to merge we didn't not violate the Team Mergers rules block.
\n
It depends on how Kaggle admin sees the definition of a \"Team\". If that definition is \"A group of people who have decided to create a merged team on kaggle platform before merger deadline\" then it's ok, but personally I don't think so.
", "votes": 16, "replies": [] } ] }, { "id": 1187793, "author_name": "Theo Viel", "author_url": "", "post_date": "2021-02-05T18:06:40.490000", "content": "
\n\nSo yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning.
\nSo we haven’t violated the rules.
\n
Why do you think that? Now the question is what the rules really are, I am not sure myself. You are right that it has been tolerated before.
", "votes": 17, "replies": [] }, { "id": 1186545, "author_name": "beluga", "author_url": "", "post_date": "2021-02-04T22:08:59.627000", "content": "Just out of curiousity I checked the official rules. (Did not read through the small letter sections…)
\n\n\nNo private sharing outside teams
\n
\n Privately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.
I have a feeling about the spirit of the rules and I try to avoid to share event the slightest hint like a paranoid.
\nSome clarification from @kaggle would be useful especially for the future.
This was answered by @juliaelliott here: https://www.kaggle.com/c/rfcx-species-audio-detection/discussion/197744
\nIt reads:
\n\n\nPlease note: You must merge within the Kaggle platform before sharing any private competition information to be considered in adherence to competition rules.
\n
It would be much better if this text was part of the rules. Not everyone reads all forum topics.
", "votes": 6, "replies": [] } ] }, { "id": 1186603, "author_name": "RNA", "author_url": "", "post_date": "2021-02-04T23:26:25.003000", "content": "… Did you just make a thread to admit that you were sharing information before merging?
", "votes": 12, "replies": [ { "id": 1188452, "author_name": "Oleksii", "author_url": "", "post_date": "2021-02-06T08:58:32.527000", "content": "We just clarified our position on this issue
", "votes": 3, "replies": [] } ] }, { "id": 1186787, "author_name": "CoreyJamesLevinson", "author_url": "", "post_date": "2021-02-05T03:28:13.607000", "content": "I think you're right, and people should stop making these witchhunting threads. Calling \"cheaters\" out in discussion threads is just trying to start a fight and publicly hurts the legitimacy of the website. If you think someone is cheating, fill out a form at https://www.kaggle.com/compliance and read this post https://www.kaggle.com/general/125466
", "votes": 11, "replies": [ { "id": 1186866, "author_name": "", "author_url": "", "post_date": "2021-02-05T05:02:52.033000", "content": "", "votes": 0, "replies": [] }, { "id": 1187030, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2021-02-05T07:15:22.863000", "content": "So you think it is better if they spend one month in this competition and get removed? Because Kaggle only removes teams at the end. Then both their efforts are wasted and people below them got demotivated. Would you be more motivated if you were 30th or 20th? I obviously don't try to start a fight, otherwise I would be fighting. Please read the original post carefully. I am more concerned about rights of the people who don't cheat.
", "votes": 1, "replies": [] } ] }, { "id": 1186829, "author_name": "cool_rabbit", "author_url": "", "post_date": "2021-02-05T04:28:09.497000", "content": "My simple question is, why didn't you merge within the Kaggle platform from the beginning?
\nSorry, I cannot fully understand the reason you described above.
\nBy early team-merging, we can share any information freely, and also decide what to share or what not to share in order to make great impact in the end (in terms of ensemble, heterogeneity, diversity, etc).
We, as people, who are friends in real life, decided to work on this competition together (from the very beginning). But, taking into account the probability that some of us may not have enough time to work on this task, we decided to wait until each of us contributes to the score and merge closer to the merge deadline, when it is clear who wants (and who is actually able) to participate and to bring value. If we would merge at the beginning we would probably 1) lose spare team places for nothing and 2) acting as an escalator, let someone stay idle but still get a medal / high score, which is not quite fair, honestly, even for friends.
\nАnd since some of us only started doing something substantial a few days ago (to stop considering themselves freeloaders), we didn’t team up - there was a lot of time left until the merge deadline.
", "votes": 2, "replies": [] }, { "id": 1188466, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2021-02-06T09:09:00.913000", "content": "@yakuben you could do this without sharing code and ideas. Everyone could work independent and then merge.
\nAnd I recommend you to be careful with such agreements. Even though they are your friends, some of them could discover an additional trick and get tempted by solo gold and could continue solo despite you agreed to merge and shared information already.
", "votes": 6, "replies": [] }, { "id": 1188478, "author_name": "cool_rabbit", "author_url": "", "post_date": "2021-02-06T09:27:03.170000", "content": "Thanks for your sincere reply to my question.
\nNow I understand your situation.
\nIn that case, after promising to team up together, not sharing any information each other about the ongoing competition before merging seems to be the best way we can take, which is my interpretation though.
@aerdem4
\nUntil recently we do this without sharing code and ideas.
\nBut when we consider ourselves a team, it gets hard not to share something.
\nI think that from our answers it becomes clear that we did not think that we were breaking the rules or received any advantages over other participants. Therefore, we did not think a lot about this situation and just waited for everyone to join.
About the situation you described above.
\n1) I am 100% confident in my teammates
\n2) If this situation happened, it would be cheating. And our team would clearly not have tolerated this (at least I can speak for myself).
It is your responsibility to read the rules. Unfortunately, thinking that you were not breaking the rules is not enough. You should have researched the rules or at least think \"so why do other people team up early?\". If your definition of team was right, everyone would team up in the last day.
\nAnyway, I don't want anyone's efforts to be wasted. So you being removed from LB and you being above other people who obeyed the rules both make me sad. Instead of having such defensive (also offensive) post shared by Dima, an apology post could be more meaningful.
\nAnd I believe you will learn to not be 100% confident with people over time. Life teaches it:)
", "votes": 5, "replies": [] }, { "id": 1188676, "author_name": "Oleksii", "author_url": "", "post_date": "2021-02-06T12:51:04.027000", "content": "1) Regarding whether we broke the rules or not.
\nI believe that we did not break the rules. I agree that what is written in the rules can be interpreted in different ways. I suspect that what is written in the discussion more a recommendation than a rule for two reasons:
In any case, it is up to the administration to decide, regardless of what the participants write here.
\n2) About \"defensive\" part of Dima post
\nIn his post, he wrote the truth about this situation from our side. He did not write something in the spirit of \"this is an accident\" or some other kind of excuses. He wrote everything as it is. And clarify our position that everything was fair play.
3) About \"offensive\" part of Dima post
\nRegardless of what the administration decides on our situation. I believe that creating posts with public accusations without serious evidence (especially before the merge deadline) (especially with evidence like same country/same university/teammates before ) is not ok. And this applies not only to kaggle, but also to social networks.
\n\nBut when we consider ourselves a team
\n
then you make it official and merge on Kaggle. ;)
", "votes": 9, "replies": [] }, { "id": 1188721, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2021-02-06T13:16:54.940000", "content": "\n\nRegardless of what the administration decides on our situation. I believe that creating posts with public accusations without serious evidence (especially before the merge deadline) (especially with evidence like same country/same university/teammates before ) is not ok. And this applies not only to kaggle, but also to social networks.
\n
I accused 2 people for private sharing and they confessed that. So I was right. Otherwise, I would apologize.
\n\n\nBut in this case it looks more like upvotes collection and hype around him.
\n
Your teammate accused me on having childish intentions. I don't care about discussion upvotes or me getting popular. I was trying to make Kaggle more fair even before I was a GM. Repeating \"respectful grandmaster\" on whole text and GM comparison between past and now have no point. It was only for taunting me.
\n\n\nWe could bring similar charges against him. We see a lot of people at the top from Nvidia. Also, we see that @aerdem4, @christofhenkel and @cpmpml (who scored 0.931 from the 1st submit!!!) were teammates before, they all are from Nvidia and, we are sure, it may be some situations when they were submitting submissions at the same time with different, but close scores.
\n
Please do so when we are at the same score and submit at the same time. I would be happy to share my solution with Kaggle. Attacking us won't make you guys right. You admit your guilt, immediately merged teams after the post and your score did not even improve after the merge since your solutions were probably almost identical.
\nI tried to be constructive in my comments but I will stop trying and focus on the competition. Kaggle will decide if it was a private sharing or not.
", "votes": 1, "replies": [] }, { "id": 1188820, "author_name": "Psi", "author_url": "", "post_date": "2021-02-06T14:37:19.723000", "content": "@yakuben All your explanation is okay, but as soon as you share ideas / concepts / codes you should merge on Kaggle. You even admit that you considered yourselves as a team from that point on.
\nI think it is pretty obvious that you get a clear advantage by waiting as long as possible to merge, namely additional submissions. Additional submissions can be more and more important the closer you get to the end of the competition and the higher you potentially are on the LB, so don't start with \"we did not exceed the overall limit otherwise we could not merge\". Everyone who has worked in serious teams knows that deciding on submissions per day can lead to many discussions and you have to put things on the side. So overall, it should be clear to you guys that what you did was shady and definitely on the edge of the spectrum of rules, regardless if it has been accepted in the past. So an apology post would have been far better for the situation.
", "votes": 16, "replies": [] }, { "id": 1189062, "author_name": "CPMP", "author_url": "", "post_date": "2021-02-06T17:33:05.020000", "content": "I concur on the number of subs. All teams in top 10 currently have used at least 25 more subs than me, and some hundreds more.
\nI'd welcome a daily allowance of 10 subs now ;)
\nMaybe I should team with my wife account without teaming officially till one week end? And even that way I won't catch up in number of subs.
\nNB. The trick of teaming with one's spouse/partner account has been used by some Kaggle GM. I didn't pick it at random.
", "votes": 1, "replies": [] }, { "id": 1189063, "author_name": "CPMP", "author_url": "", "post_date": "2021-02-06T17:33:40.417000", "content": "But I prefer to focus on finding what I missed so far compared to those who passed me. ;)
", "votes": 2, "replies": [] }, { "id": 1189105, "author_name": "Selim Seferbekov", "author_url": "", "post_date": "2021-02-06T18:01:59.170000", "content": "\n\nI concur on the number of subs. All teams in top 10 currently have used at least 25 more subs than me, and some hundreds more.
\n
Actually I got 0.970 from around 30 subs. With other 100 I tried differrent models and approaches but all of them hit the wall at this score or below.
\nSo, unfortunately, the number of subs cannot be directly converted to the public score, otherwise I would score 0.980+ by now😀
Regarding the topic: from Kaggle perspective obviously team is a clear concept as even one person is a team with one team member. Five solo competitors means five different teams until they merge.
", "votes": 6, "replies": [] }, { "id": 1189168, "author_name": "CPMP", "author_url": "", "post_date": "2021-02-06T19:05:55.817000", "content": "Good to know, because you had 67 subs when I joined. I assumed you need all of them to get to 970.
", "votes": 0, "replies": [] }, { "id": 1192058, "author_name": "", "author_url": "", "post_date": "2021-02-08T22:00:03.220000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 1190105, "author_name": "Mukharbek Organokov", "author_url": "", "post_date": "2021-02-07T13:42:06.880000", "content": "Sometimes social battles in Kaggle competitions are more impressive than the competitions itself.
\nReady for great melodrama 😎
", "votes": 8, "replies": [] }, { "id": 1209946, "author_name": "Gilles Vandewiele", "author_url": "", "post_date": "2021-02-19T05:54:14.070000", "content": "This didn't age well now, did it?
", "votes": 4, "replies": [] }, { "id": 1186460, "author_name": "Pascal Pfeiffer", "author_url": "", "post_date": "2021-02-04T20:39:49.753000", "content": "Well, you just started another toxic thread.
", "votes": 3, "replies": [] }, { "id": 1186520, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2021-02-04T21:45:16.463000", "content": "Didn't read after
\n\n", "votes": 2, "replies": [] }, { "id": 1188193, "author_name": "MaChaogong", "author_url": "", "post_date": "2021-02-06T03:51:21.360000", "content": "So yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. So we haven’t violated the rules.
\n
Rule is much simpler than competition itself. You cheated for sure. If I were you I would had started a apologize thread instead.
", "votes": 3, "replies": [ { "id": 1188678, "author_name": "DimaSakovich", "author_url": "", "post_date": "2021-02-06T12:52:09.503000", "content": "We don't think so
", "votes": -2, "replies": [] }, { "id": 1188730, "author_name": "MaChaogong", "author_url": "", "post_date": "2021-02-06T13:23:57.740000", "content": "\n\nSo yes, we shared some information inside our team, approaches and ideas. Because we were a team from the early beginning. So we haven’t violated the rules.
\n
That means you share from the early beginning, while you merge days ago.
\nInfact kagglers really don't care whether you cheat, we just want to keep kaggle fair still.
\nIf we want, we all could think up more excuses than you:)) to prove we really don't private sharing and merge at teammerge deadline. While we mostly prefer a simpler way, team once we choose our teammate. All things about how teammates perform along the competion or some teammates just disappearing is our choose, we shall be responsible for every decision we make.
Ps: \"be responsible\" is the second important thing you shall learn.
\nSry if something toxic above, I'm not good at express myself in English.
\nOh look that I just find an excuse for my impoliteness.