Really, what is that? What is the point of it?!?
", "messages": [ { "id": "1528702", "postDate": "09/29/2021 22:34:02", "content": "Really, what is that? What is the point of it?!?
", "rawMarkdown": "Really, what is that? What is the point of it?!?", "votes": null }, { "id": "1528704", "postDate": "09/29/2021 22:37:04", "content": "And now we have dropped 10 place immediately, thanks!!!!!!
", "rawMarkdown": "And now we have dropped 10 place immediately, thanks!!!!!!", "votes": null }, { "id": "1528705", "postDate": "09/29/2021 22:39:01", "content": "@snnclsr - I think I just saw that also - it looks like it was removed now though. Still that is not cool at all!
", "rawMarkdown": "snnclsr - I think I just saw that also - it looks like it was removed now though. Still that is not cool at all!", "votes": null }, { "id": "1528708", "postDate": "09/29/2021 22:41:36", "content": "Yea, they removed it which was the correct action. It was most probably an accident, but will cause a lot pain for the rest of teams
", "rawMarkdown": "Yea, they removed it which was the correct action. It was most probably an accident, but will cause a lot pain for the rest of teams", "votes": null }, { "id": "1528714", "postDate": "09/29/2021 22:49:09", "content": "Can Kaggle do something with this disaster? Because with that pub notebook accidentally released, we were forced to submit our blend with it to remain on roughly the proper place.
\nNow that it was being deleted, other people cant even do that.
\nFreezing silver/bronze and below scores on the time of that notebook release, perhaps?
\n@maggiemd ? (i do not know who to tag properly)
", "rawMarkdown": "Can Kaggle do something with this disaster? Because with that pub notebook accidentally released, we were forced to submit our blend with it to remain on roughly the proper place.\n\nNow that it was being deleted, other people cant even do that.\n\nFreezing silver/bronze and below scores on the time of that notebook release, perhaps?\n\n@maggiemd ? (i do not know who to tag properly)", "votes": null }, { "id": "1528719", "postDate": "09/29/2021 22:52:41", "content": "Kaggle needs to do the right thing and revert those submissions in my opinion. I didn't participate here but I could only imagine how upsetting this is for the participants
", "rawMarkdown": "Kaggle needs to do the right thing and revert those submissions in my opinion. I didn't participate here but I could only imagine how upsetting this is for the participants", "votes": null }, { "id": "1528724", "postDate": "09/29/2021 23:04:07", "content": "They are nuts. First in discussions, then releasing code.
", "rawMarkdown": "They are nuts. First in discussions, then releasing code.", "votes": null }, { "id": "1528725", "postDate": "09/29/2021 23:05:54", "content": "@fffrrt @Snnclsr @hengck23 - Can Kaggle recover the leaderboard just before that share? I think the orange provisional banner may be up on the LB for a while after this mess.
", "rawMarkdown": "fffrrt @Snnclsr @hengck23 - Can Kaggle recover the leaderboard just before that share? I think the orange provisional banner may be up on the LB for a while after this mess.", "votes": null }, { "id": "1528731", "postDate": "09/29/2021 23:18:43", "content": "I don't know that information, unfortunately. We should wait for some reaction from the Kaggle team maybe? I believe it is strongly discouraged but not against the rules.
", "rawMarkdown": "I don't know that information, unfortunately. We should wait for some reaction from the Kaggle team maybe? I believe it is strongly discouraged but not against the rules.", "votes": null }, { "id": "1528737", "postDate": "09/29/2021 23:29:38", "content": "Aah, that looks scary. I hope something would be done regarding this.
", "rawMarkdown": "Aah, that looks scary. I hope something would be done regarding this.", "votes": null }, { "id": "1528739", "postDate": "09/29/2021 23:31:06", "content": "Perhaps in the future Kaggle could disable creating public notebooks that have competition data in the last week of the competition?
\nThis way accidental shares would be prevented.
", "rawMarkdown": "Perhaps in the future Kaggle could disable creating public notebooks that have competition data in the last week of the competition?\n\nThis way accidental shares would be prevented.", "votes": null }, { "id": "1528743", "postDate": "09/29/2021 23:40:03", "content": "It would be even better if they disabled sharing a notebook if it has the potential to affect x% of the leaderboard. I'd say x is at least 90. Executing that would probably be no easy task though
", "rawMarkdown": "It would be even better if they disabled sharing a notebook if it has the potential to affect x% of the leaderboard. I'd say x is at least 90. Executing that would probably be no easy task though", "votes": null }, { "id": "1528744", "postDate": "09/29/2021 23:42:41", "content": "Just to be clear @cpmpml I don't think author is referring to our team. @fffrrt . We did not share any code.
", "rawMarkdown": "Just to be clear @cpmpml I don't think author is referring to our team. @fffrrt . We did not share any code.", "votes": null }, { "id": "1528757", "postDate": "09/30/2021 00:16:03", "content": "It's insane, we dropped about 20 rank on LB due to this in last few hours. Kaggle must enforce non-sharing towards the end rather than just \"recommending\"
", "rawMarkdown": "It's insane, we dropped about 20 rank on LB due to this in last few hours. Kaggle must enforce non-sharing towards the end rather than just \"recommending\"", "votes": null }, { "id": "1528758", "postDate": "09/30/2021 00:17:07", "content": "I 💯 agree
", "rawMarkdown": "I 💯 agree", "votes": null }, { "id": "1528784", "postDate": "09/30/2021 00:50:20", "content": "Fair enough, I didn't see what had been shared in notebooks.
\nBut this proves the point I made in the forum: late sharing can be destructive.
", "rawMarkdown": "Fair enough, I didn't see what had been shared in notebooks.\n\nBut this proves the point I made in the forum: late sharing can be destructive.", "votes": null }, { "id": "1528800", "postDate": "09/30/2021 01:09:42", "content": "Late sharing is indeed destructive. I don't think any of our team members did that. We were inactive in the discussion forums.
\nThe only thing that was shared was that 1D CNN + 2D CNN gives a boost in ensemble, which was mentioned previously as well and can be found in the comment section here
\nStill the way the topic was framed,top 3x solution, led many to believe that it was actually our team that shared a high scoring Public Notebook in the final hours. Even the members of our team got confused. We do not even have kaggle notebooks for this competition and neither did any one of share towards the end.
@benihime91 I think what your team did in discussions is absolutely idiotic, I'm sorry. A lot of things was mentioned on a lot of discussions, but that coming from top 3 team is not the same thing. It surely does not change anything for you guys, but can (and did) move lower LB a lot. There is no reason to share that type of information with 10+h left. It's just bad that you dont care about it and dont see a problem with that. (I understand that mentioned notebook share was not by you, what was that btw can we get a link or something?)
", "rawMarkdown": "benihime91 I think what your team did in discussions is absolutely idiotic, I'm sorry. A lot of things was mentioned on a lot of discussions, but that coming from top 3 team is not the same thing. It surely does not change anything for you guys, but can (and did) move lower LB a lot. There is no reason to share that type of information with 10+h left. It's just bad that you dont care about it and dont see a problem with that. (I understand that mentioned notebook share was not by you, what was that btw can we get a link or something?)", "votes": null }, { "id": "1528863", "postDate": "09/30/2021 02:23:27", "content": "@bakeryproducts for context this is what 1 of our team mates shared
\n1D + 2D ensemble --> BoOm\nNo arch details, no training details. There's literally nothing here and those who were exploring/knew about 1D CNNs (most were) already knew about this. And did I mention that this is literally pasted everywhere in the comment section of this discussion.
\nResnet34 --> boom 💥\nThis was another comment. Here we weren't even talking about our models. We were referring to the resnet34 scores of the 1st place team which they posted about in the discussion.
\nand finally this comment by me was just an inside joke.
\n@pheadrus My brain 🧠 ---> boom 💥\nStill if it did move lower LB a lot, let me apologise on behalf of my team, and in future I personally will refrain from sharing anything and just stick to competitions only.
@benihime91 Oh, thank you for not sharing the code before deadline, that was so great of you! This is a very serious discussion about results of work for a lot of people, your ironic apology is not accepted as you still think that you did nothing wrong.
", "rawMarkdown": "benihime91 Oh, thank you for not sharing the code before deadline, that was so great of you! This is a very serious discussion about results of work for a lot of people, your ironic apology is not accepted as you still think that you did nothing wrong.", "votes": null }, { "id": "1528883", "postDate": "09/30/2021 02:43:10", "content": "@bakeryproducts i did not mean to sound ironic sorry again for that as well.
\nIt's just that I wanted to put forward our side of the story as well. We just shared 1 ensembling tip (that worked for us) and nothing about the models themselves. If you did not have the appropriate models, it's not like someone can just whip up a very good and diverse 1d cnn/2 d cnn, ensemble their predictions and just jump up.
\nAlso if I am not wrong, the lower half of the LB was shifted due the release of the public notebook, rather than our comment (again correct me if I am wrong! )
", "rawMarkdown": "bakeryproducts i did not mean to sound ironic sorry again for that as well. \n\nIt's just that I wanted to put forward our side of the story as well. We just shared 1 ensembling tip (that worked for us) and nothing about the models themselves. If you did not have the appropriate models, it's not like someone can just whip up a very good and diverse 1d cnn/2 d cnn, ensemble their predictions and just jump up.\n\nAlso if I am not wrong, the lower half of the LB was shifted due the release of the public notebook, rather than our comment (again correct me if I am wrong! )", "votes": null }, { "id": "1528894", "postDate": "09/30/2021 02:57:10", "content": "IMO archs doesnt matter that much, I choose random networks for competitions and its working, ensembling is a great deal. What is you are probably also missing that there is most important part where you choose your final submission, based on top team hint it can change a lot. And it can be not about jump up from your hint, it can be jumping very much down because of implementation details (and choosing that submit) of same idea. Late sharing is bad universally i think. There are no sides in this story, It's just me personally have a problem with that. Will read the solution anyway.
", "rawMarkdown": "IMO archs doesnt matter that much, I choose random networks for competitions and its working, ensembling is a great deal. What is you are probably also missing that there is most important part where you choose your final submission, based on top team hint it can change a lot. And it can be not about jump up from your hint, it can be jumping very much down because of implementation details (and choosing that submit) of same idea. Late sharing is bad universally i think. There are no sides in this story, It's just me personally have a problem with that. Will read the solution anyway.", "votes": null }, { "id": "1528922", "postDate": "09/30/2021 03:21:51", "content": "I think there is a bug in Kaggle. I can share my experience here. I forked my personal notebook and submitted it a few hrs before competition. It is a private notebook as can be seen here
\nHowever, to my surprise I see that someone has upvoted it and also made comments on it. How did someone find my private kernel ?? Looks like there was a period of time when some private notebooks became public and then became private again. That seems to be the only explanation. For example, the high scoring notebook that is being referred in this discussion is nowhere to be found and I doubt the team that shared it, did it on purpose.
\nPlease note that I am not a new user to Kaggle and would never share a kernel in the last stages of the competition.
\nThankfully my noteobook is not a highscoring notebook and I have been working on this competition only from personal learning from perspective. But I can very easily see this happening with others too.
Is it something worth investigating Kaggle team ?
", "rawMarkdown": "I think there is a bug in Kaggle. I can share my experience here. I forked my personal notebook and submitted it a few hrs before competition. It is a private notebook as can be seen [here](https://pasteboard.co/lt1pSJ7thPtH.png)\n\nHowever, to my surprise I see that someone has upvoted it and also made comments on it. How did someone find my private kernel ?? Looks like there was a period of time when **some private notebooks became public and then became private again. That seems to be the only explanation**. For example, the high scoring notebook that is being referred in this discussion is nowhere to be found and I doubt the team that shared it, did it on purpose. \n\nPlease note that I am not a new user to Kaggle and would never share a kernel in the last stages of the competition.\nThankfully my noteobook is not a highscoring notebook and I have been working on this competition only from personal learning from perspective. But I can very easily see this happening with others too. \n\nIs it something worth investigating Kaggle team ?", "votes": null }, { "id": "1529047", "postDate": "09/30/2021 05:43:14", "content": "Dear @watzisname
\n\n\n\"some private notebooks became public\"
\n
I can concur; that also happened to me last night, around 22:00 UTC (in a different competition).
\nAll the best,
\ncarl
sympathize with you 😒
", "rawMarkdown": "sympathize with you 😒", "votes": null }, { "id": "1529103", "postDate": "09/30/2021 06:51:00", "content": "I also noticed the same! One of my notebooks that contained a csv with oof and preds literally went public by its own!!! (I was preparing it to easily share a few results with my teammates). Fortunately I noticed it right away (5min or so late) and remove it. But trust me this was not an error by my part!
\nNote: it was not the notebook the author is referring. It was just a stupid model that didn’t help at the end with cv of 8660 but this is scary anyway
", "rawMarkdown": "I also noticed the same! One of my notebooks that contained a csv with oof and preds literally went public by its own!!! (I was preparing it to easily share a few results with my teammates). Fortunately I noticed it right away (5min or so late) and remove it. But trust me **this was not an error by my part!**\n\nNote: it was not the notebook the author is referring. It was just a stupid model that didn’t help at the end with cv of 8660 but this is scary anyway", "votes": null }, { "id": "1529131", "postDate": "09/30/2021 07:18:38", "content": "5 hours remaining: 26th rank I can sleep in peace now
\nSomeone decided to spoil whole LB while I'm sleeping
\nWakes up in the morningL 43 rank with (-1 public -> private difference) ☹️
Seemed to be a bug. I have experienced this yesterday evening as well in another competition. I recognized it when somebody gave me an upvote to a kernel that was just meant to be a playing around for myself.
", "rawMarkdown": "Seemed to be a bug. I have experienced this yesterday evening as well in another competition. I recognized it when somebody gave me an upvote to a kernel that was just meant to be a playing around for myself.", "votes": null }, { "id": "1529359", "postDate": "09/30/2021 10:20:14", "content": "Yes, there was a crazy bug yesterday, horrible that that happened, but not the fault of the team. What was the public score of that NB?
", "rawMarkdown": "Yes, there was a crazy bug yesterday, horrible that that happened, but not the fault of the team. What was the public score of that NB?", "votes": null }, { "id": "1529360", "postDate": "09/30/2021 10:22:11", "content": "\n\nWhat was the public score of that NB?
\n
@philippsinger Scores were 0.8815 (mid-silver) for the one people are most worried about, 0.8788 (just below bronze, but potentially shaking up lower medal zone) for @watzisname's one.
", "rawMarkdown": "> What was the public score of that NB?\n\n@philippsinger Scores were 0.8815 (mid-silver) for the one people are most worried about, 0.8788 (just below bronze, but potentially shaking up lower medal zone) for @watzisname's one.", "votes": null }, { "id": "1529363", "postDate": "09/30/2021 10:27:24", "content": "Something similar happened in the Molecular Properties comp (accidental not a bug), and the Kaggle team handled it well (IMO). I wonder if something similar can be done in retrospect?
\nKaggle's response for reference: https://www.kaggle.com/c/champs-scalar-coupling/discussion/105237
", "rawMarkdown": "Something similar happened in the Molecular Properties comp (accidental not a bug), and the Kaggle team handled it well (IMO). I wonder if something similar can be done in retrospect?\n\nKaggle's response for reference: https://www.kaggle.com/c/champs-scalar-coupling/discussion/105237", "votes": null }, { "id": "1529365", "postDate": "09/30/2021 10:29:36", "content": "Wow 0.8815 is very high… do we know what the private score of that team is now?
\nIt will be very tricky to disable those submissions as people potentially blended with it. @inversion
", "rawMarkdown": "Wow 0.8815 is very high... do we know what the private score of that team is now?\n\nIt will be very tricky to disable those submissions as people potentially blended with it. @inversion", "votes": null }, { "id": "1529369", "postDate": "09/30/2021 10:33:26", "content": "I think linear blends should be easy to find. @inversion and the team are smart people and probably have an arsenal of tools for this type of thing
", "rawMarkdown": "I think linear blends should be easy to find. @inversion and the team are smart people and probably have an arsenal of tools for this type of thing", "votes": null }, { "id": "1529496", "postDate": "09/30/2021 12:57:14", "content": "Notebook sharing should be reviewed as a whole. In theory is a great learning resource. In practice it is just easy medal farming.
", "rawMarkdown": "Notebook sharing should be reviewed as a whole. In theory is a great learning resource. In practice it is just easy medal farming.", "votes": null }, { "id": "1530211", "postDate": "10/01/2021 03:03:06", "content": "Completely insane ! Worst time for a kaggle bug. We lost ~15 places :(
", "rawMarkdown": "Completely insane ! Worst time for a kaggle bug. We lost ~15 places :(", "votes": null }, { "id": "1530409", "postDate": "10/01/2021 06:37:28", "content": "https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/275580
\nResponse from kaggle team: Yesterday's Notebook Sharing Bug and Impact to G2Net
I put the link of kaggle official reply here. So do not blame the team as they didn't do it on purpose.
\n(maybe the author of the post can update the information too)
Hey All,
\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.
\nThank you all,
\nRegards,
\nChris
And now we have dropped 10 place immediately, thanks!!!!!!
", "votes": null, "replies": [] }, { "id": 1528705, "author_name": "rdizzl3", "author_url": "", "post_date": "09/29/2021 22:39:01", "content": "@snnclsr - I think I just saw that also - it looks like it was removed now though. Still that is not cool at all!
", "votes": null, "replies": [ { "id": 1528708, "author_name": "snnclsr", "author_url": "", "post_date": "09/29/2021 22:41:36", "content": "Yea, they removed it which was the correct action. It was most probably an accident, but will cause a lot pain for the rest of teams
", "votes": null, "replies": [] } ] }, { "id": 1528714, "author_name": "fffrrt", "author_url": "", "post_date": "09/29/2021 22:49:09", "content": "Can Kaggle do something with this disaster? Because with that pub notebook accidentally released, we were forced to submit our blend with it to remain on roughly the proper place.
\nNow that it was being deleted, other people cant even do that.
\nFreezing silver/bronze and below scores on the time of that notebook release, perhaps?
\n@maggiemd ? (i do not know who to tag properly)
", "votes": null, "replies": [ { "id": 1528719, "author_name": "rdizzl3", "author_url": "", "post_date": "09/29/2021 22:52:41", "content": "Kaggle needs to do the right thing and revert those submissions in my opinion. I didn't participate here but I could only imagine how upsetting this is for the participants
", "votes": null, "replies": [] } ] }, { "id": 1528724, "author_name": "cpmpml", "author_url": "", "post_date": "09/29/2021 23:04:07", "content": "They are nuts. First in discussions, then releasing code.
", "votes": null, "replies": [ { "id": 1528744, "author_name": "benihime91", "author_url": "", "post_date": "09/29/2021 23:42:41", "content": "Just to be clear @cpmpml I don't think author is referring to our team. @fffrrt . We did not share any code.
", "votes": null, "replies": [] }, { "id": 1528784, "author_name": "cpmpml", "author_url": "", "post_date": "09/30/2021 00:50:20", "content": "Fair enough, I didn't see what had been shared in notebooks.
\nBut this proves the point I made in the forum: late sharing can be destructive.
", "votes": null, "replies": [] }, { "id": 1528800, "author_name": "benihime91", "author_url": "", "post_date": "09/30/2021 01:09:42", "content": "Late sharing is indeed destructive. I don't think any of our team members did that. We were inactive in the discussion forums.
\nThe only thing that was shared was that 1D CNN + 2D CNN gives a boost in ensemble, which was mentioned previously as well and can be found in the comment section here
\nStill the way the topic was framed,top 3x solution, led many to believe that it was actually our team that shared a high scoring Public Notebook in the final hours. Even the members of our team got confused. We do not even have kaggle notebooks for this competition and neither did any one of share towards the end.
@benihime91 I think what your team did in discussions is absolutely idiotic, I'm sorry. A lot of things was mentioned on a lot of discussions, but that coming from top 3 team is not the same thing. It surely does not change anything for you guys, but can (and did) move lower LB a lot. There is no reason to share that type of information with 10+h left. It's just bad that you dont care about it and dont see a problem with that. (I understand that mentioned notebook share was not by you, what was that btw can we get a link or something?)
", "votes": null, "replies": [] }, { "id": 1528863, "author_name": "benihime91", "author_url": "", "post_date": "09/30/2021 02:23:27", "content": "@bakeryproducts for context this is what 1 of our team mates shared
\n1D + 2D ensemble --> BoOm\nNo arch details, no training details. There's literally nothing here and those who were exploring/knew about 1D CNNs (most were) already knew about this. And did I mention that this is literally pasted everywhere in the comment section of this discussion.
\nResnet34 --> boom 💥\nThis was another comment. Here we weren't even talking about our models. We were referring to the resnet34 scores of the 1st place team which they posted about in the discussion.
\nand finally this comment by me was just an inside joke.
\n@pheadrus My brain 🧠 ---> boom 💥\nStill if it did move lower LB a lot, let me apologise on behalf of my team, and in future I personally will refrain from sharing anything and just stick to competitions only.
@benihime91 Oh, thank you for not sharing the code before deadline, that was so great of you! This is a very serious discussion about results of work for a lot of people, your ironic apology is not accepted as you still think that you did nothing wrong.
", "votes": null, "replies": [] }, { "id": 1528883, "author_name": "benihime91", "author_url": "", "post_date": "09/30/2021 02:43:10", "content": "@bakeryproducts i did not mean to sound ironic sorry again for that as well.
\nIt's just that I wanted to put forward our side of the story as well. We just shared 1 ensembling tip (that worked for us) and nothing about the models themselves. If you did not have the appropriate models, it's not like someone can just whip up a very good and diverse 1d cnn/2 d cnn, ensemble their predictions and just jump up.
\nAlso if I am not wrong, the lower half of the LB was shifted due the release of the public notebook, rather than our comment (again correct me if I am wrong! )
", "votes": null, "replies": [] }, { "id": 1528894, "author_name": "bakeryproducts", "author_url": "", "post_date": "09/30/2021 02:57:10", "content": "IMO archs doesnt matter that much, I choose random networks for competitions and its working, ensembling is a great deal. What is you are probably also missing that there is most important part where you choose your final submission, based on top team hint it can change a lot. And it can be not about jump up from your hint, it can be jumping very much down because of implementation details (and choosing that submit) of same idea. Late sharing is bad universally i think. There are no sides in this story, It's just me personally have a problem with that. Will read the solution anyway.
", "votes": null, "replies": [] } ] }, { "id": 1528725, "author_name": "jbomitchell", "author_url": "", "post_date": "09/29/2021 23:05:54", "content": "@fffrrt @Snnclsr @hengck23 - Can Kaggle recover the leaderboard just before that share? I think the orange provisional banner may be up on the LB for a while after this mess.
", "votes": null, "replies": [ { "id": 1528731, "author_name": "snnclsr", "author_url": "", "post_date": "09/29/2021 23:18:43", "content": "I don't know that information, unfortunately. We should wait for some reaction from the Kaggle team maybe? I believe it is strongly discouraged but not against the rules.
", "votes": null, "replies": [] } ] }, { "id": 1528737, "author_name": "nischaydnk", "author_url": "", "post_date": "09/29/2021 23:29:38", "content": "Aah, that looks scary. I hope something would be done regarding this.
", "votes": null, "replies": [] }, { "id": 1528739, "author_name": "fffrrt", "author_url": "", "post_date": "09/29/2021 23:31:06", "content": "Perhaps in the future Kaggle could disable creating public notebooks that have competition data in the last week of the competition?
\nThis way accidental shares would be prevented.
", "votes": null, "replies": [ { "id": 1528743, "author_name": "rdizzl3", "author_url": "", "post_date": "09/29/2021 23:40:03", "content": "It would be even better if they disabled sharing a notebook if it has the potential to affect x% of the leaderboard. I'd say x is at least 90. Executing that would probably be no easy task though
", "votes": null, "replies": [] } ] }, { "id": 1528757, "author_name": "moizsaifee", "author_url": "", "post_date": "09/30/2021 00:16:03", "content": "It's insane, we dropped about 20 rank on LB due to this in last few hours. Kaggle must enforce non-sharing towards the end rather than just \"recommending\"
", "votes": null, "replies": [ { "id": 1528758, "author_name": "rdizzl3", "author_url": "", "post_date": "09/30/2021 00:17:07", "content": "I 💯 agree
", "votes": null, "replies": [] }, { "id": 1529082, "author_name": "hedgehed", "author_url": "", "post_date": "09/30/2021 06:31:35", "content": "sympathize with you 😒
", "votes": null, "replies": [] } ] }, { "id": 1528922, "author_name": "watzisname", "author_url": "", "post_date": "09/30/2021 03:21:51", "content": "I think there is a bug in Kaggle. I can share my experience here. I forked my personal notebook and submitted it a few hrs before competition. It is a private notebook as can be seen here
\nHowever, to my surprise I see that someone has upvoted it and also made comments on it. How did someone find my private kernel ?? Looks like there was a period of time when some private notebooks became public and then became private again. That seems to be the only explanation. For example, the high scoring notebook that is being referred in this discussion is nowhere to be found and I doubt the team that shared it, did it on purpose.
\nPlease note that I am not a new user to Kaggle and would never share a kernel in the last stages of the competition.
\nThankfully my noteobook is not a highscoring notebook and I have been working on this competition only from personal learning from perspective. But I can very easily see this happening with others too.
Is it something worth investigating Kaggle team ?
", "votes": null, "replies": [ { "id": 1529047, "author_name": "carlmcbrideellis", "author_url": "", "post_date": "09/30/2021 05:43:14", "content": "Dear @watzisname
\n\n\n\"some private notebooks became public\"
\n
I can concur; that also happened to me last night, around 22:00 UTC (in a different competition).
\nAll the best,
\ncarl
I also noticed the same! One of my notebooks that contained a csv with oof and preds literally went public by its own!!! (I was preparing it to easily share a few results with my teammates). Fortunately I noticed it right away (5min or so late) and remove it. But trust me this was not an error by my part!
\nNote: it was not the notebook the author is referring. It was just a stupid model that didn’t help at the end with cv of 8660 but this is scary anyway
", "votes": null, "replies": [] }, { "id": 1529137, "author_name": "allunia", "author_url": "", "post_date": "09/30/2021 07:20:57", "content": "Seemed to be a bug. I have experienced this yesterday evening as well in another competition. I recognized it when somebody gave me an upvote to a kernel that was just meant to be a playing around for myself.
", "votes": null, "replies": [] }, { "id": 1529359, "author_name": "philippsinger", "author_url": "", "post_date": "09/30/2021 10:20:14", "content": "Yes, there was a crazy bug yesterday, horrible that that happened, but not the fault of the team. What was the public score of that NB?
", "votes": null, "replies": [ { "id": 1529360, "author_name": "jbomitchell", "author_url": "", "post_date": "09/30/2021 10:22:11", "content": "\n\nWhat was the public score of that NB?
\n
@philippsinger Scores were 0.8815 (mid-silver) for the one people are most worried about, 0.8788 (just below bronze, but potentially shaking up lower medal zone) for @watzisname's one.
", "votes": null, "replies": [] } ] }, { "id": 1529363, "author_name": "anjum48", "author_url": "", "post_date": "09/30/2021 10:27:24", "content": "Something similar happened in the Molecular Properties comp (accidental not a bug), and the Kaggle team handled it well (IMO). I wonder if something similar can be done in retrospect?
\nKaggle's response for reference: https://www.kaggle.com/c/champs-scalar-coupling/discussion/105237
", "votes": null, "replies": [] }, { "id": 1529365, "author_name": "philippsinger", "author_url": "", "post_date": "09/30/2021 10:29:36", "content": "Wow 0.8815 is very high… do we know what the private score of that team is now?
\nIt will be very tricky to disable those submissions as people potentially blended with it. @inversion
", "votes": null, "replies": [] }, { "id": 1529369, "author_name": "anjum48", "author_url": "", "post_date": "09/30/2021 10:33:26", "content": "I think linear blends should be easy to find. @inversion and the team are smart people and probably have an arsenal of tools for this type of thing
", "votes": null, "replies": [] } ] }, { "id": 1529131, "author_name": "muhammad4hmed", "author_url": "", "post_date": "09/30/2021 07:18:38", "content": "5 hours remaining: 26th rank I can sleep in peace now
\nSomeone decided to spoil whole LB while I'm sleeping
\nWakes up in the morningL 43 rank with (-1 public -> private difference) ☹️
Notebook sharing should be reviewed as a whole. In theory is a great learning resource. In practice it is just easy medal farming.
", "votes": null, "replies": [] }, { "id": 1530211, "author_name": "zarif98sjs", "author_url": "", "post_date": "10/01/2021 03:03:06", "content": "Completely insane ! Worst time for a kaggle bug. We lost ~15 places :(
", "votes": null, "replies": [] }, { "id": 1530409, "author_name": "hengck23", "author_url": "", "post_date": "10/01/2021 06:37:28", "content": "https://www.kaggle.com/c/g2net-gravitational-wave-detection/discussion/275580
\nResponse from kaggle team: Yesterday's Notebook Sharing Bug and Impact to G2Net
I put the link of kaggle official reply here. So do not blame the team as they didn't do it on purpose.
\n(maybe the author of the post can update the information too)
Hey All,
\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.
\nThank you all,
\nRegards,
\nChris