![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1182060%2Fbf4f94ca741e896a10affc6af1ba1194%2FScreenshot%20from%202021-01-20%2008-17-40.png?generation=1611141552251993&alt=media\")
I was browsing through the notebooks section and I quickly find almost 20 notebooks that are essentially exact copies of some of my work posted here, people doing that are novices or contributors, but still, I find it disrespectful, to have your work copied and posted as someones else, after investing many hours into creating something helpful for the community, at some level, this always happens in competitions but here it was more than I expected.
\nFor those who have reported notebooks for plagiarism, does anything happens at a reasonable time?
", "messages": [ { "id": 1159692, "postDate": "2021-01-19T12:02:17.830Z", "content": "I was browsing through the notebooks section and I quickly find almost 20 notebooks that are essentially exact copies of some of my work posted here, people doing that are novices or contributors, but still, I find it disrespectful, to have your work copied and posted as someones else, after investing many hours into creating something helpful for the community, at some level, this always happens in competitions but here it was more than I expected.
\nFor those who have reported notebooks for plagiarism, does anything happens at a reasonable time?
", "rawMarkdown": "I was browsing through the notebooks section and I quickly find almost 20 notebooks that are essentially exact copies of some of my work posted here, people doing that are novices or contributors, but still, I find it disrespectful, to have your work copied and posted as someones else, after investing many hours into creating something helpful for the community, at some level, this always happens in competitions but here it was more than I expected.\n\nFor those who have reported notebooks for plagiarism, does anything happens at a reasonable time?", "votes": 97 }, { "id": 1160160, "postDate": "2021-01-19T17:45:08.300Z", "content": "You're doing great work, and I think they'll be hard to see in other competitions after all. So please continue your great work. @dimitreoliveira
", "rawMarkdown": "You're doing great work, and I think they'll be hard to see in other competitions after all. So please continue your great work. @dimitreoliveira ", "votes": 12 }, { "id": 1177950, "postDate": "2021-01-30T15:30:18.557Z", "content": "I have a solution for you : start coding in Julia or R.
\nMore seriously yes, I guess it is a pain. Citing properly work is the base of scientific demarch, especially for people that want to go deeper into details. I actually commented one of your kernels recently, which was quite usefull for me, and you even comment back to answer my question.
\nI guess we could flag also flag plagiarism in comment, even it is seems a bit mean to me (but maybe necessary). I did not even know that there were a system to report plagiarism.
Sub-question : is the commentary sections of this notebook with plagiarism are filled with \"I upvoted nananan please checked my kernel ?\". I fear that people hunting for up-vote tend to up-vote each other and persist.
", "rawMarkdown": "I have a solution for you : start coding in Julia or R.\n\nMore seriously yes, I guess it is a pain. Citing properly work is the base of scientific demarch, especially for people that want to go deeper into details. I actually commented one of your kernels recently, which was quite usefull for me, and you even comment back to answer my question.\nI guess we could flag also flag plagiarism in comment, even it is seems a bit mean to me (but maybe necessary). I did not even know that there were a system to report plagiarism. \n\nSub-question : is the commentary sections of this notebook with plagiarism are filled with \"I upvoted nananan please checked my kernel ?\". I fear that people hunting for up-vote tend to up-vote each other and persist. ", "votes": 9, "replies": [ { "id": 1184229, "postDate": "2021-02-03T12:58:37.273Z", "content": "\n\nI fear that people hunting for up-vote tend to up-vote each other and persist.
\n
For sure. I noticed that they spam newly publish notebooks everywhere.
", "rawMarkdown": "> I fear that people hunting for up-vote tend to up-vote each other and persist.\n\nFor sure. I noticed that they spam newly publish notebooks everywhere.", "votes": 1 } ] }, { "id": 1175096, "postDate": "2021-01-28T22:29:28.847Z", "content": "Rampant plagiarism is simply because people think that they could faster and more easily reach to master and grandmaster title ranks in such a way (discussion and data are also other ways to do so and in such a sense they also have got their own distorsions). Also people think that if they get to that grade they can boast it out and that will bring great advantages to them and their career.
\nSincerely, I do hire people and I do look at what they do on Kaggle, but I absolutely do not consider rankings gained on discussion, notebooks and data unless I see something really unique and original. I do not know if there are around other hiring managers that have a different opinion than mine, but mine is that having good grades in discussion/notebooks/data is generally (there are exceptions, though) really not much predictive of one's ability in data science.
\nDiscussion/notebooks/data were intended to help Kagglers and support them in their learning and competiting and they are much useful without any doubt. Anyway, until you will have grades similar to competitions attached to discussion/notebooks/data, you will always experience a host of plagiars and free riders that you cannot avoid and you cannot mitigate in any way.
", "rawMarkdown": "Rampant plagiarism is simply because people think that they could faster and more easily reach to master and grandmaster title ranks in such a way (discussion and data are also other ways to do so and in such a sense they also have got their own distorsions). Also people think that if they get to that grade they can boast it out and that will bring great advantages to them and their career.\n\nSincerely, I do hire people and I do look at what they do on Kaggle, but I absolutely do not consider rankings gained on discussion, notebooks and data unless I see something really unique and original. I do not know if there are around other hiring managers that have a different opinion than mine, but mine is that having good grades in discussion/notebooks/data is generally (there are exceptions, though) really not much predictive of one's ability in data science. \n\nDiscussion/notebooks/data were intended to help Kagglers and support them in their learning and competiting and they are much useful without any doubt. Anyway, until you will have grades similar to competitions attached to discussion/notebooks/data, you will always experience a host of plagiars and free riders that you cannot avoid and you cannot mitigate in any way. ", "votes": 7, "replies": [ { "id": 1175169, "postDate": "2021-01-29T01:14:09.643Z", "content": "I agree @lucamassaron , I have been interviewed a few times and my Kaggle titles had no impact so far, what mattered was the experienced I gained here, competing or creating content, sometimes event competition medals have little impact, maybe people got lucky with a random blend.
", "rawMarkdown": "I agree @lucamassaron , I have been interviewed a few times and my Kaggle titles had no impact so far, what mattered was the experienced I gained here, competing or creating content, sometimes event competition medals have little impact, maybe people got lucky with a random blend.", "votes": 4 }, { "id": 1197857, "postDate": "2021-02-12T13:41:43.180Z", "content": "100% agree with you @lucamassaron . Even in my case, when I am hiring for my team I do not place much importance on titles gained in Data, Discussions and Notebooks. In terms of hiring, I place importance on Competitions medals - when the candidate can explain the solution and subsequent downstream business value of their solution.
", "rawMarkdown": "100% agree with you @lucamassaron . Even in my case, when I am hiring for my team I do not place much importance on titles gained in Data, Discussions and Notebooks. In terms of hiring, I place importance on Competitions medals - when the candidate can explain the solution and subsequent downstream business value of their solution. ", "votes": 1 } ] }, { "id": 1170298, "postDate": "2021-01-26T05:49:50.160Z", "content": "This is a genuinely worrying trend to me. Whenever I look at the notebooks for competitions I see increasing numbers of ripoffs and it is becoming harder to find the original source. Very often the ripoffs have important notes removed but rise to the top because they beat the source work by .0001 on the lb.
", "rawMarkdown": "This is a genuinely worrying trend to me. Whenever I look at the notebooks for competitions I see increasing numbers of ripoffs and it is becoming harder to find the original source. Very often the ripoffs have important notes removed but rise to the top because they beat the source work by .0001 on the lb. ", "votes": 5 }, { "id": 1176692, "postDate": "2021-01-29T18:20:47.423Z", "content": "It is a shame to see your notebooks being plagiarized, I have seen most of your notebooks(you may even recognize me), and it's very cruel for those who have copied your beautiful original notebooks that have helped multitudes of people(including me).
\nDo realize, they won't be the ones improving, they won't have the fun experience making their own notebooks, hence they won't even go far in their career without simply crediting someone else for their work.
\nUltimately, I believe that the competition for kaggle medals has gone so far to the extent people are cheating for them. The reason why I(and many other alike), are using kaggle is a means to gain experience, to learn. As long as we learn something new, we give ourselves a medal. Sorry to hear people are taking your credit away.
", "rawMarkdown": "It is a shame to see your notebooks being plagiarized, I have seen most of your notebooks(you may even recognize me), and it's very cruel for those who have copied your beautiful original notebooks that have helped multitudes of people(including me).\n\nDo realize, they won't be the ones improving, they won't have the fun experience making their own notebooks, hence they won't even go far in their career without simply crediting someone else for their work. \n\nUltimately, I believe that the competition for kaggle medals has gone so far to the extent people are cheating for them. The reason why I(and many other alike), are using kaggle is a means to gain experience, to learn. As long as we learn something new, we give ourselves a medal. Sorry to hear people are taking your credit away. ", "votes": 3, "replies": [ { "id": 1176903, "postDate": "2021-01-29T21:55:07.667Z", "content": "Thanks for the kind words @andyjianzhou ,
\nI also believe learning should be the main drive, that was the reason I joined Kaggle and is the reason I stay :)
You are absolutely right. When i browse the top competitors in any ongoing competitions
\nAlmost all of them have exact same score.
\nHow is that even possible.
\nThis can not be coincidence definitely.
\nKaggle team should pay more attention to code plagarism.
That's actually quite normal for many people to have the same score, but not the same CV
", "rawMarkdown": "That's actually quite normal for many people to have the same score, but not the same CV", "votes": 1 } ] }, { "id": 1160093, "postDate": "2021-01-19T17:06:52.843Z", "content": "I feel you. That's why I like to only share helpful concepts and ideas that work, but not the full implementation. It's a shame, your notebook with TPU was really high quality too.
", "rawMarkdown": "I feel you. That's why I like to only share helpful concepts and ideas that work, but not the full implementation. It's a shame, your notebook with TPU was really high quality too.", "votes": 3, "replies": [ { "id": 1161209, "postDate": "2021-01-20T12:12:59.593Z", "content": "I have used your notebook as a reference and mentioned that I have referred it. Will it be okay to do that.
\nhttps://www.kaggle.com/sajidhussain3/beginners-take-on-cassava-kfold-augment-keras
Please go ahead! Feel free to fork notebooks I make public :)
", "rawMarkdown": "Please go ahead! Feel free to fork notebooks I make public :)", "votes": 1 } ] }, { "id": 1159903, "postDate": "2021-01-19T14:34:11.653Z", "content": "As someone who is new to kaggle and was listening to lots of kaggle-related podcasts, one of the consensus was really how prevalent the problem of plagiarism is in this platform and how most GMs really think it won't go away easily.
\nThat said, I made sure to develop the habit of placing credits even if i only have a notebook in private.
\nFor what it's worth, your notebooks and your contributions in the discussion are really helpful especially to competition novices like me. Let me know how can I also help report plagiarisms.
", "rawMarkdown": "As someone who is new to kaggle and was listening to lots of kaggle-related podcasts, one of the consensus was really how prevalent the problem of plagiarism is in this platform and how most GMs really think it won't go away easily.\n\nThat said, I made sure to develop the habit of placing credits even if i only have a notebook in private.\n\nFor what it's worth, your notebooks and your contributions in the discussion are really helpful especially to competition novices like me. Let me know how can I also help report plagiarisms.", "votes": 3, "replies": [ { "id": 1160009, "postDate": "2021-01-19T15:56:30.407Z", "content": "Yes @jasondolorso , I also agree that plagiarism is a tricky problem to solve, especially because it is so easy to copy code here. But having said that the fork feature is also awesome forking code from experienced people and experimented here at Kaggle was one of the main things that enabled me to learn very fast.
Forking public kernels and adding relevant content on top of it, then publish it, is one of the things the makes Kaggle awesome, and makes progress so fast here, my issue is with people re-publish the same code with no relevant modifications, and sometimes they seem to try to claim ownership of the whole thing.
", "rawMarkdown": "Yes @jasondolorso , I also agree that plagiarism is a tricky problem to solve, especially because it is so easy to copy code here. But having said that the `fork` feature is also awesome forking code from experienced people and experimented here at Kaggle was one of the main things that enabled me to learn very fast.\n\nForking public kernels and adding relevant content on top of it, then publish it, is one of the things the makes Kaggle awesome, and makes progress so fast here, my issue is with people re-publish the same code with no relevant modifications, and sometimes they seem to try to claim ownership of the whole thing.", "votes": 2 }, { "id": 1160037, "postDate": "2021-01-19T16:12:31.190Z", "content": "I agree. Republishing without *fork *means they really made the effort to either download the code locally and re-uploading, or copying the code cell by cell.
\nThat got me thinking.. With the thousands of data science talents here in Kaggle, especially those with NLP expertise, it would be a great project to have a model that analyzes plagiarism based on code similarity — of course, given some constraints such as not forking or giving credits.
", "rawMarkdown": "I agree. Republishing without *fork *means they really made the effort to either download the code locally and re-uploading, or copying the code cell by cell.\n\nThat got me thinking.. With the thousands of data science talents here in Kaggle, especially those with NLP expertise, it would be a great project to have a model that analyzes plagiarism based on code similarity — of course, given some constraints such as not forking or giving credits.", "votes": 2 }, { "id": 1160066, "postDate": "2021-01-19T16:41:12.650Z", "content": "Maybe that could be an idea for a playground competition.
", "rawMarkdown": "Maybe that could be an idea for a playground competition.", "votes": 1 }, { "id": 1160073, "postDate": "2021-01-19T16:47:15.820Z", "content": "Jason
\nPretty sure that the analysis that you mention already exists in Kaggle. Another of the more trouble some \"cheats\" is for folks to work as an informal team and share code between them, but without ever becoming a formal Kaggle team. A disadvantage of forming a kaggle team is the loss of daily submissions. A team of 8 only gets 5 daily submissions, while an informal cheating team of 8 would get 40. Looking for duplicated code, not published in a public kernel is one of the methods to control this cheat.
\nTo the best of my knowledge there is no rule that prohibits an exact copy of a shared kernel from being used for the final submissions and leader board ranking. Sharing a high scoring kernel near the end of a competition is considered by many as a form of cheating - your shared code will get tons of forks and upvotes - increasing your ranking in one of the 4 pillars.
\n\nA disadvantage of forming a kaggle team is the loss of daily submissions. A team of 8 only gets 5 daily submissions, while an informal cheating team of 8 would get 40.
\n
I agree with this. Some competitions, in particular, can be easily exploited by scanning through the testing data.
\nCheaters will always find a way, and we can only do so much to mitigate it. But I believe that you would know a legit great data scientist (or Kaggle Master/GM) when you see how they dispose themselves and through their works. I doubt that those who are gaming the system would really rise up to their levels in terms of popularity and skill.
\nI am also relatively new to the platform though, so take my opinion with a grain of salt.
", "rawMarkdown": "> A disadvantage of forming a kaggle team is the loss of daily submissions. A team of 8 only gets 5 daily submissions, while an informal cheating team of 8 would get 40.\n\nI agree with this. Some competitions, in particular, can be easily exploited by scanning through the testing data. \n\nCheaters will always find a way, and we can only do so much to mitigate it. But I believe that you would know a legit great data scientist (or Kaggle Master/GM) when you see how they dispose themselves and through their works. I doubt that those who are *gaming* the system would really rise up to their levels in terms of popularity and skill.\n\nI am also relatively new to the platform though, so take my opinion with a grain of salt." }, { "id": 1160977, "postDate": "2021-01-20T08:46:45.477Z", "content": "\n\nCheaters will always find a way>
\n
I don't cheat. I don't publish kernels unless they are mostly my original work. I do fork and use shared kernels on my local PC and for submission - it's the greatest thing about Kaggle that I can learn from some pretty sharp folks who share their code.
\nIn the years I have been active on the site, Kaggle has implemented lots of things to prevent \"cheaters\". IMO many of the changes make it extremely difficult for novice folks to participate. One only needs to look at all the posts in the last year, of various competitions, for folks trying to solve \"submission not found\". Pretty sure that this single attempt to \"prevent cheating\" has wasted thousands of hours of novice's time and driven a pretty decent percentage of them away from the site in frustration.
\nSo my general rule about cheater discussions on Kaggle - folks - just STFU.
", "rawMarkdown": "> Cheaters will always find a way> \n\nI don't cheat. I don't publish kernels unless they are mostly my original work. I do fork and use shared kernels on my local PC and for submission - it's the greatest thing about Kaggle that I can learn from some pretty sharp folks who share their code.\n\nIn the years I have been active on the site, Kaggle has implemented lots of things to prevent \"cheaters\". IMO many of the changes make it extremely difficult for novice folks to participate. One only needs to look at all the posts in the last year, of various competitions, for folks trying to solve \"submission not found\". Pretty sure that this single attempt to \"prevent cheating\" has wasted thousands of hours of novice's time and driven a pretty decent percentage of them away from the site in frustration. \n\nSo my general rule about cheater discussions on Kaggle - folks - just STFU. \n\n", "votes": -3 }, { "id": 1161117, "postDate": "2021-01-20T10:42:16.510Z", "content": "@pcjimmmy I don't understand why you are being so aggressive towards the topic being discussed here. Also I don't think Kaggle made any changes that prevent novices from learning, please provide some links if I am wrong. All @dimitreoliveira said was that people keep copying his notebook and publishing them as their own work without crediting the original notebook, which is frustrating because people don't even have the decency to be grateful to the original author. This behavior is also against Kaggle community guidelines.
\n\n\nfor folks trying to solve \"submission not found\". Pretty sure that this single attempt to \"prevent cheating\" has wasted thousands of hours of novice's time
\n
How is the \"submission not found\" error related to anti-cheating? This issue is faced by only a handful of participants and the community helps them solve the issue when they post it on a discussion.
", "rawMarkdown": "@pcjimmmy I don't understand why you are being so aggressive towards the topic being discussed here. Also I don't think Kaggle made any changes that prevent novices from learning, please provide some links if I am wrong. All @dimitreoliveira said was that people keep copying his notebook and publishing them as their own work without crediting the original notebook, which is frustrating because people don't even have the decency to be grateful to the original author. This behavior is also against Kaggle community [guidelines](https://www.kaggle.com/community-guidelines).\n\n> for folks trying to solve \"submission not found\". Pretty sure that this single attempt to \"prevent cheating\" has wasted thousands of hours of novice's time \n\nHow is the \"submission not found\" error related to anti-cheating? This issue is faced by only a handful of participants and the community helps them solve the issue when they post it on a discussion.", "votes": 3 }, { "id": 1161596, "postDate": "2021-01-20T16:30:17.497Z", "content": "Yovin
\nSorry if I appear aggressive. It is frustration - not aggression that was intended.
\nWhen you and I joined Kaggle we could generate a submission file on our local PC and get a LB score within a minute or two. I could process a wide range of ideas, generate the code, make a submission and get a score in rapid fashion. I almost always was using my daily allowance of submissions. I was a complete newbie at R and that system let me have a very rapid learning curve.
\nCheaters spent their time probing the private test set and other actions that took unfair use of the direct submission and test data availability.
\nThe current submission process, as practiced in this competition, is the apparent direct response by Kaggle to prevent those forms of cheating. Now to validate an idea I need to save/submit. Wait a period of time while the public test set is run. Come back and submit - wait up to 9 hours for the code to run, come back and finally get a score. For the last year or more for competitions using this cheat prevention method I almost NEVER use my daily quota of submissions - the process to turn over a single idea takes a full day. When things error out - only a very limited and almost useless bit of information is provided to help debug. I never had a submission not found error back in the days when I could make a direct submission. I never recall a post in discussion asking for help for those type of errors.
\nKaggle knows the failure rate for this type of submission error. They should know how many new folks are hitting that initial roadblock. They can probably estimate the rate at which folks stop being active on the site and how many submission failures they had before they stopped. I spent 50+ years processing data to solve problems - most of the time that needed to be done in my head, I think I got pretty good at making general guesses that would accurate enough for me to solve the issue. My guess is that it's a lot more than a \"handful\". A quick search of this completion with \"submission error\" gets me two handfuls. Not sure I want to try a search for all competitions :}
\nI am not sure the \"community helps\" as much as they can - I know you are very active as a helper.
\nKaggle has done lots of good things over my years here to address the specific issue of the poster. You can report kernels, when a kernel is forked there is an automated citation added, etc.
\nMy frustration is that every time a new form of cheating is invented or a new complaint gains traction what follows is ALWAYS pain for me.
", "rawMarkdown": "Yovin\n\nSorry if I appear aggressive. It is frustration - not aggression that was intended.\n\nWhen you and I joined Kaggle we could generate a submission file on our local PC and get a LB score within a minute or two. I could process a wide range of ideas, generate the code, make a submission and get a score in rapid fashion. I almost always was using my daily allowance of submissions. I was a complete newbie at R and that system let me have a very rapid learning curve.\n\nCheaters spent their time probing the private test set and other actions that took unfair use of the direct submission and test data availability.\n\nThe current submission process, as practiced in this competition, is the apparent direct response by Kaggle to prevent those forms of cheating. Now to validate an idea I need to save/submit. Wait a period of time while the public test set is run. Come back and submit - wait up to 9 hours for the code to run, come back and finally get a score. For the last year or more for competitions using this cheat prevention method I almost NEVER use my daily quota of submissions - the process to turn over a single idea takes a full day. When things error out - only a very limited and almost useless bit of information is provided to help debug. I never had a submission not found error back in the days when I could make a direct submission. I never recall a post in discussion asking for help for those type of errors.\n\nKaggle knows the failure rate for this type of submission error. They should know how many new folks are hitting that initial roadblock. They can probably estimate the rate at which folks stop being active on the site and how many submission failures they had before they stopped. I spent 50+ years processing data to solve problems - most of the time that needed to be done in my head, I think I got pretty good at making general guesses that would accurate enough for me to solve the issue. My guess is that it's a lot more than a \"handful\". A quick search of this completion with \"submission error\" gets me two handfuls. Not sure I want to try a search for all competitions :}\n\nI am not sure the \"community helps\" as much as they can - I know you are very active as a helper.\n\nKaggle has done lots of good things over my years here to address the specific issue of the poster. You can report kernels, when a kernel is forked there is an automated citation added, etc.\n\nMy frustration is that every time a new form of cheating is invented or a new complaint gains traction what follows is ALWAYS pain for me.\n\n\n\n\n\n\n\n" }, { "id": 1161637, "postDate": "2021-01-20T16:59:35.267Z", "content": "@pcjimmmy The main reason for having a public/private re-run of the submission notebook is to make sure the solution presented to the host of the competition is capable of running efficiently. When people were allowed to submit the CSV file, each participant's code will require different computing resources, which may not be ideal for the competition host since some participants may have extremely powerful local workstations for training and inference.
\nBy setting a rule to submit your solution through the Kaggle kernel with the same amount of resources available to everyone, it produces a more real-world deployable model. This is why you see different submission run times for some competitions because the host expects our solution to work within the specified time and computing resources available. While this also reduces cheating, it is not the main reason for this change.
", "rawMarkdown": "@pcjimmmy The main reason for having a public/private re-run of the submission notebook is to make sure the solution presented to the host of the competition is capable of running efficiently. When people were allowed to submit the CSV file, each participant's code will require different computing resources, which may not be ideal for the competition host since some participants may have extremely powerful local workstations for training and inference.\n\nBy setting a rule to submit your solution through the Kaggle kernel with the same amount of resources available to everyone, it produces a more real-world deployable model. This is why you see different submission run times for some competitions because the host expects our solution to work within the specified time and computing resources available. While this also reduces cheating, it is not the main reason for this change.", "votes": 1 }, { "id": 1161706, "postDate": "2021-01-20T17:42:19.620Z", "content": "Going to follow my own advice and STFU - but two last comments. Apparently one form of whining can get you a gold metal while my form gets me down votes :) Besides - I have drifted too far from the original posters topic.
\nWhen I directly submitted a submission file you are correct in that I could have generated that file using a Supercomputer that I happened to have access to. But pretty sure that the bottom 1000 folks who clearly have submission file issues in this competition are not using super computers to generate the model that than fails. With the current GPU and TPU quota's I not sure how competition hosts are benefiting from us all having to play with the same limited toys. A real-world deployable model is hard to reach on 30 hours a week with a full day turn around for each iteration.
\nSo I don't see it - The current submission process does NOTHING to prevent me from using a super computer to build a model. It just adds one cumbersome step to the submission process.
\nTake a look at the leader board - when I just looked there are 11 folks who's best score is clearly a submission error. There several hundred above that who's best score is likely also a type of submission error. There are more than a handful of names on that list that I am willing to bet will not be seen as active next month. (my standard bet is a Big Ben sandwich at Parnalli's in downtown Pittsburgh - GO STEELERS)
", "rawMarkdown": "Going to follow my own advice and STFU - but two last comments. Apparently one form of whining can get you a gold metal while my form gets me down votes :) Besides - I have drifted too far from the original posters topic.\n\nWhen I directly submitted a submission file you are correct in that I could have generated that file using a Supercomputer that I happened to have access to. But pretty sure that the bottom 1000 folks who clearly have submission file issues in this competition are not using super computers to generate the model that than fails. With the current GPU and TPU quota's I not sure how competition hosts are benefiting from us all having to play with the same limited toys. A real-world deployable model is hard to reach on 30 hours a week with a full day turn around for each iteration.\n\nSo I don't see it - The current submission process does NOTHING to prevent me from using a super computer to build a model. It just adds one cumbersome step to the submission process. \n\nTake a look at the leader board - when I just looked there are 11 folks who's best score is clearly a submission error. There several hundred above that who's best score is likely also a type of submission error. There are more than a handful of names on that list that I am willing to bet will not be seen as active next month. (my standard bet is a Big Ben sandwich at Parnalli's in downtown Pittsburgh - GO STEELERS)", "votes": -4 } ] }, { "id": 1167756, "postDate": "2021-01-24T13:16:53.053Z", "content": "Hey totally agree, I've seen plagiarism frequently in the past months. I guess some individuals plagiarise to get quick upvotes in the community. Nevertheless, I think it is disrispectful to the person who has dedicated so much effort and time towards publishing a notebook. It took me once 3 months to publish a notebook imagine that, how would I feel if others take credit for all that effort and sacrifice I did in those 3 months? I also had once someone who contacted me on LinkedIn so that I could sign the individual's thesis and I found out that one part of the person's thesis was exactly one of the notebooks I published in Kaggle. You really see sometimes this kind of situations but to be honest I don't know how much can Kaggle do other than investigating the notebook when someone reports plagiarism. Anyways, your notebooks are amazing and don't feel discourage, keep sharing your amazing work!
", "rawMarkdown": "Hey totally agree, I've seen plagiarism frequently in the past months. I guess some individuals plagiarise to get quick upvotes in the community. Nevertheless, I think it is disrispectful to the person who has dedicated so much effort and time towards publishing a notebook. It took me once 3 months to publish a notebook imagine that, how would I feel if others take credit for all that effort and sacrifice I did in those 3 months? I also had once someone who contacted me on LinkedIn so that I could sign the individual's thesis and I found out that one part of the person's thesis was exactly one of the notebooks I published in Kaggle. You really see sometimes this kind of situations but to be honest I don't know how much can Kaggle do other than investigating the notebook when someone reports plagiarism. Anyways, your notebooks are amazing and don't feel discourage, keep sharing your amazing work!", "votes": 4, "replies": [ { "id": 1167880, "postDate": "2021-01-24T14:45:09.603Z", "content": "Hi @janiobachmann , oh boy what a story 😊, when you open a notebook it is clear if the author has put real effort in making that, I am sure that the 3-month notebook must be awesome, we don't always get the reaction that we expect, but we know it is not only for the upvotes. Thank you for your kind words, I will keep them coming.
", "rawMarkdown": "Hi @janiobachmann , oh boy what a story 😊, when you open a notebook it is clear if the author has put real effort in making that, I am sure that the 3-month notebook must be awesome, we don't always get the reaction that we expect, but we know it is not only for the upvotes. Thank you for your kind words, I will keep them coming.", "votes": 2 } ] }, { "id": 1204415, "postDate": "2021-02-16T06:02:05.187Z", "content": "Taking credit for someone else work is really shameful. It takes hours and days of hard work to publish the good work. I think the discussion should be going forward Is there a way to make it difficult for cheaters?
", "rawMarkdown": "Taking credit for someone else work is really shameful. It takes hours and days of hard work to publish the good work. I think the discussion should be going forward Is there a way to make it difficult for cheaters? ", "votes": 1 }, { "id": 1204170, "postDate": "2021-02-15T22:33:06.103Z", "content": "I totally agree, what I see is that every day it's increasing, but I think most kagglers realizes that, so keep up the good work 👌
", "rawMarkdown": "I totally agree, what I see is that every day it's increasing, but I think most kagglers realizes that, so keep up the good work 👌", "votes": 1 }, { "id": 1183584, "postDate": "2021-02-03T05:29:47.617Z", "content": "Yes I agree with you this is shameful activity.
", "rawMarkdown": "Yes I agree with you this is shameful activity.", "votes": 1 }, { "id": 1178330, "postDate": "2021-01-30T19:44:27.420Z", "content": "hi, I just wanted to add, as a community member who often looks to notebooks to learn, this is definitely an issue on Kaggle and has been for some time. I think we struggle as a community to continue to support an open resource environment for the sake of improvement while holding every contributor, viewer and member to the appropriate levels of integrity and anti-plaigarism standards. While I am a relatively new member, I wonder if Kaggle has considered adding a \"report this post\" button similar to Instagram, etc. Realistically this would require more hands on monitoring, but I think it could be more successful than simply an upvote or downvote. Potentially it could make sense to nominate community members to help moderate competitions and postings? Again, this is more work, but I think it could potentially be worth it.
\nI also love the comment to switch to R, Etienne. I didn't think of that, but you are probably right.
", "rawMarkdown": "hi, I just wanted to add, as a community member who often looks to notebooks to learn, this is definitely an issue on Kaggle and has been for some time. I think we struggle as a community to continue to support an open resource environment for the sake of improvement while holding every contributor, viewer and member to the appropriate levels of integrity and anti-plaigarism standards. While I am a relatively new member, I wonder if Kaggle has considered adding a \"report this post\" button similar to Instagram, etc. Realistically this would require more hands on monitoring, but I think it could be more successful than simply an upvote or downvote. Potentially it could make sense to nominate community members to help moderate competitions and postings? Again, this is more work, but I think it could potentially be worth it.\n\nI also love the comment to switch to R, Etienne. I didn't think of that, but you are probably right. ", "votes": 1 }, { "id": 1169201, "postDate": "2021-01-25T11:33:45.680Z", "content": "Honestly, can someone explain why Kaggle cannot implement checks of that sort in their notebooks? I have seen these type of threads many times all over the site. I can't imagine it would require much effort for making a model to detect plagiarism in code (kinda like the already existing document plagiarism checkers).
\nThen Kaggle could show a small private message to the user listing all the sources copied from, and ask if the user would like to credit those sources in their notebooks. It seems a pretty decent \"First-Response\" method to discourage Plagarism. AFAIK Kaggle hasn't implemented anything of the sort.
\nKaggle = Google. This kind of issue should have been resolved years ago. It's alright to copy notebooks as long as you give them the due credit. IMO Kaggle/Google does not take this issue very seriously OR they don't realize how demotivating it is to have your work copied and not given due credit.
", "rawMarkdown": "Honestly, can someone explain why Kaggle cannot implement checks of that sort in their notebooks? I have seen these type of threads many times all over the site. I can't imagine it would require much effort for making a model to detect plagiarism in code (kinda like the already existing document plagiarism checkers). \n\nThen Kaggle could show a small private message to the user listing all the sources copied from, and ask if the user would like to credit those sources in their notebooks. It seems a pretty decent \"First-Response\" method to discourage Plagarism. AFAIK Kaggle hasn't implemented anything of the sort. \n\nKaggle = Google. This kind of issue should have been resolved years ago. It's alright to copy notebooks as long as you give them the due credit. IMO Kaggle/Google does not take this issue very seriously OR they don't realize how **demotivating** it is to have your work copied and not given due credit.", "votes": 1 }, { "id": 1161002, "postDate": "2021-01-20T08:59:54.827Z", "content": "Thanks to your opinion, I have a lot to think about.
", "rawMarkdown": "Thanks to your opinion, I have a lot to think about.", "votes": 1 }, { "id": 1160821, "postDate": "2021-01-20T06:42:54.443Z", "content": "I have personally used some of your setup code for Tensorflow, so thank you. Think I have upvoted your notebooks which I've made use of. I've not shared any public notebooks in this comp.
\nIt does seem to me that if Kaggle could find a way to make this more automatic in crediting original notebook, it would fix the problem both for people intentionally and unintentionally sharing copies without crediting. I suppose people could still make a copy via methods other than forking, but at least that would highlight deliberate examples.
", "rawMarkdown": "I have personally used some of your setup code for Tensorflow, so thank you. Think I have upvoted your notebooks which I've made use of. I've not shared any public notebooks in this comp.\n\nIt does seem to me that if Kaggle could find a way to make this more automatic in crediting original notebook, it would fix the problem both for people intentionally and unintentionally sharing copies without crediting. I suppose people could still make a copy via methods other than forking, but at least that would highlight deliberate examples.", "votes": 1, "replies": [ { "id": 1161152, "postDate": "2021-01-20T11:19:21.907Z", "content": "You're welcome @davidedwards1 ,
\nActually, Kaggle does have a way to give credit to the original notebook, when you fork a notebook you will see a reference on the upper right side, like this:
\n
I agree with you. Personally, I think it's very unfortunate.
", "rawMarkdown": "I agree with you. Personally, I think it's very unfortunate.", "votes": 1 }, { "id": 1160362, "postDate": "2021-01-19T21:19:08.447Z", "content": "I've seen a couple things (even that i used myself this competition) that i don't even remember who's the original author such as the GradualWarmup code. But thanks for sharing your frustration, this was my first competition so i wasn't 100% aware of this sort of problems. But thanks to your post i'll definitely be always posting credits and links on the appropriate sections from now on.
\nMaybe kaggle could implement a method, such as:
\nif notebook is forked and notebook.author != notebook.fork.author
\nas to not award points and medals to discourage people from just simply copying notebooks.
@capiru I wouldn't go as far as giving credits to every section of code (unless maybe it is something very unique), it would be too much work to find the original authors, as you said.
\nCopying code and hacking it, is a great tool to learn, and I encourage it.
", "rawMarkdown": "@capiru I wouldn't go as far as giving credits to every section of code (unless maybe it is something very unique), it would be too much work to find the original authors, as you said.\n\nCopying code and hacking it, is a great tool to learn, and I encourage it." } ] }, { "id": 1162003, "postDate": "2021-01-20T22:31:39.257Z", "content": "Olá,
\nNotei que é BR, então irei lhe responder em português. Enfim, é realmente um grande desrespeito copiar e até mesmo divulgar um notebook sem dar os devidos créditos, porém não creio que seja motivo para ficar muito zangado. Não tiro sua razão, mas no fundo, sabemos que o que importa é que você tem a capacidade, o conhecimento e fez o esforço para criar um material interessante para a comunidade, e quem só copia algo não costuma ir longe.
\nGrande abraço,
\n…
Oi @olavomendes ,
\nNa verdade eu nem fico muito chateado em ver esse tipo de coisa, venho publicando notebooks aqui ja tem um tempo, e certamente irei continuar. O que realmente importa eh isso tudo ai que voce mencionou, mas tambem acho que eh importante levantar esses topicos eventualmente, ate para que o pessoal de suporte da plataforma ficar ciente de problemas do tipo.
\nOutro abraco 😄
", "rawMarkdown": "Oi @olavomendes ,\n\nNa verdade eu nem fico muito chateado em ver esse tipo de coisa, venho publicando notebooks aqui ja tem um tempo, e certamente irei continuar. O que realmente importa eh isso tudo ai que voce mencionou, mas tambem acho que eh importante levantar esses topicos eventualmente, ate para que o pessoal de suporte da plataforma ficar ciente de problemas do tipo.\n\nOutro abraco 😄", "votes": 1 } ] }, { "id": 1160483, "postDate": "2021-01-20T00:45:00.220Z", "content": "\n\npeople doing that are novices or contributors
\n
I also want to look at it from another angle and think that some of those novices/contributors are actually unaware of properly forking or the meta-etiquettes here in kaggle (albeit its a pretty well-known morally that you should credit people's work).
\nTo remove that aspect, maybe the platform can implement reminders when it's someone's first time opening, downloading and/or forking a notebook. The contents of this comment, that I dug up a few weeks ago when I was wary about forking a notebook, would be a good start:
\nhttps://www.kaggle.com/c/siim-acr-pneumothorax-segmentation/discussion/97571#563571
I am not sure if it is detailed in the terms and conditions when signing up (who reads them?) but it is probable that someone would overlook it or even forget it by the time they would first work on a notebook.
", "rawMarkdown": "> people doing that are novices or contributors\n\nI also want to look at it from another angle and think that some of those novices/contributors are actually unaware of properly forking or the *meta-etiquettes* here in kaggle (albeit its a pretty well-known morally that you should credit people's work). \n\nTo remove that aspect, maybe the platform can implement reminders when it's someone's first time opening, downloading and/or forking a notebook. The contents of this comment, that I dug up a few weeks ago when I was wary about forking a notebook, would be a good start:\nhttps://www.kaggle.com/c/siim-acr-pneumothorax-segmentation/discussion/97571#563571\n\nI am not sure if it is detailed in the terms and conditions when signing up (who reads them?) but it is probable that someone would overlook it or even forget it by the time they would first work on a notebook.", "votes": 2 }, { "id": 1160032, "postDate": "2021-01-19T16:10:39.757Z", "content": "One of the four pillars of the Kaggle profile rating system is Notebook Contributor. You gain ranking as folks copy and fork, along with up vote your notebook.
\nIf you do not want your work copied your solution is real simple - don't share it.
\nYou can also pay attention to notebooks and NOT up vote those that are copies with no improvements. Careless up votes are a reason why \"plagiarism\" can exist. So do your part.
\nGM's became GM's because they chose to share great code.
\nKaggle has made lots of changes over the couple of years I have been active to prevent folks from gaming the ranking system - IMO lots of the changes have hurt the novice 99% who do not game the system. It's a very hard task to make rules that help more than they hurt. I think they are trying hard.
\nA version of your post has appeared in almost every competition that I have followed for the last several years - hmmm - is that plagiarism? :)
", "rawMarkdown": "One of the four pillars of the Kaggle profile rating system is Notebook Contributor. You gain ranking as folks copy and fork, along with up vote your notebook.\n\nIf you do not want your work copied your solution is real simple - don't share it.\n\nYou can also pay attention to notebooks and NOT up vote those that are copies with no improvements. Careless up votes are a reason why \"plagiarism\" can exist. So do your part.\n\nGM's became GM's because they chose to share great code.\n\nKaggle has made lots of changes over the couple of years I have been active to prevent folks from gaming the ranking system - IMO lots of the changes have hurt the novice 99% who do not game the system. It's a very hard task to make rules that help more than they hurt. I think they are trying hard.\n\nA version of your post has appeared in almost every competition that I have followed for the last several years - hmmm - is that plagiarism? :)\n", "replies": [ { "id": 1160082, "postDate": "2021-01-19T16:56:23.833Z", "content": "I don't think you gain rakings by other people copying or forking your code.
\nAlso, I don't think that not sharing is the solution here, but paying attention to code before upvoting is, as you mentioned, sharing code is what makes Kaggle what it is, in my opinion. But I agree that ending plagiarism is not an easy task.
", "rawMarkdown": "I don't think you gain rakings by other people copying or forking your code.\n\nAlso, I don't think that not sharing is the solution here, but paying attention to code before upvoting is, as you mentioned, sharing code is what makes Kaggle what it is, in my opinion. But I agree that ending plagiarism is not an easy task.", "votes": 1 }, { "id": 1160320, "postDate": "2021-01-19T20:22:04.207Z", "content": "@pcjimmmy Republishing someone's public code claiming it as one's own without giving due credits is infact plagiarism, Kaggle kernels are licensed under Apache 2.0 as well. Most people change a couple parameters and earn clout for it- it's bound to annoy the original author.
", "rawMarkdown": "@pcjimmmy Republishing someone's public code claiming it as one's own without giving due credits is infact plagiarism, Kaggle kernels are licensed under Apache 2.0 as well. Most people change a couple parameters and earn clout for it- it's bound to annoy the original author.", "votes": 3 }, { "id": 1161012, "postDate": "2021-01-20T09:05:34.017Z", "content": "Hmm - tried to find the criteria for ranking in the \"Notebooks\" pillar - it seems like up votes and fork count create metals. So your ranking in this pillar is the result of folks copying and forking, with probably a small percentage of them giving you an upvote along the way.
\nA lot of the issue I see with \"copy and paste\" kernels seems to be mostly folks who are pretty new to Kaggle. Lot's of them don't know the desired community standards for giving credit - up voting if you like and use shared code, etc.
\nI have sometimes offered a comment in a shared kernel suggesting that the person give credit and identify what they changed, but that seems later when I re-read my post to be me being a bully.
\nDon't know the answer, but IMO the last thing needed is a new set of rules that make life on Kaggle difficult for the 99% who use it in an honest manner.
", "rawMarkdown": "Hmm - tried to find the criteria for ranking in the \"Notebooks\" pillar - it seems like up votes and fork count create metals. So your ranking in this pillar is the result of folks copying and forking, with probably a small percentage of them giving you an upvote along the way.\n\nA lot of the issue I see with \"copy and paste\" kernels seems to be mostly folks who are pretty new to Kaggle. Lot's of them don't know the desired community standards for giving credit - up voting if you like and use shared code, etc.\n\nI have sometimes offered a comment in a shared kernel suggesting that the person give credit and identify what they changed, but that seems later when I re-read my post to be me being a bully.\n\nDon't know the answer, but IMO the last thing needed is a new set of rules that make life on Kaggle difficult for the 99% who use it in an honest manner.\n\n" } ] }, { "id": 2380632, "postDate": "2023-08-08T17:25:50.127Z", "content": "You might consider it as plagiarism and find it unexpectable but from a different perspective it isn't really, they copy your code and start trying better preprocessing, better models and it just gives them a better headstart. For beginners who want fast growth in the field, they don't really have great resources to understand Machine/Deep Learning Pipelines and simply the Flow. 😭
\nKaggle Novices/Contributers generally don't have a great Python sense or a sense of code, this helps them get the concepts without starting from scratch which definitely is way faster.
\nSome people do just want medals and so just ignore them and look at the bright side. 😇
I'm a noob. I'm here just to learn and I think quite a few people are here for the same. To copy and paste another's code and call it on my is not a learning process. But on that note, I looked at other codes and used the same procedure usually when it comes to preparing the data, not the modeling and training that part is where I'm trying to build a good model on my own merit. Is this wrong of me or is this cheating? Because I do not want to cheat
", "rawMarkdown": "I'm a noob. I'm here just to learn and I think quite a few people are here for the same. To copy and paste another's code and call it on my is not a learning process. But on that note, I looked at other codes and used the same procedure usually when it comes to preparing the data, not the modeling and training that part is where I'm trying to build a good model on my own merit. Is this wrong of me or is this cheating? Because I do not want to cheat" }, { "id": 1194948, "postDate": "2021-02-10T13:24:07.227Z", "content": "I agree with you and no more consent.
", "rawMarkdown": " I agree with you and no more consent." }, { "id": 1395491, "postDate": "2021-07-21T09:05:06.347Z", "content": "Unfortunately, I have found so many notebooks also that copied others' works.
\nFor example, if you check the following link, you may find an exact copy of my work.
\nhttps://www.kaggle.com/satpreetmakhija/netflix-movies-and-tv-shows-2021/code
", "rawMarkdown": "Unfortunately, I have found so many notebooks also that copied others' works. \n\nFor example, if you check the following link, you may find an exact copy of my work.\n\n\nhttps://www.kaggle.com/satpreetmakhija/netflix-movies-and-tv-shows-2021/code", "isDeleted": true } ], "comments": [ { "id": 1160160, "author_name": "Heroseo", "author_url": "", "post_date": "2021-01-19T17:45:08.300000", "content": "You're doing great work, and I think they'll be hard to see in other competitions after all. So please continue your great work. @dimitreoliveira
", "votes": 12, "replies": [] }, { "id": 1177950, "author_name": "Etienne R", "author_url": "", "post_date": "2021-01-30T15:30:18.557000", "content": "I have a solution for you : start coding in Julia or R.
\nMore seriously yes, I guess it is a pain. Citing properly work is the base of scientific demarch, especially for people that want to go deeper into details. I actually commented one of your kernels recently, which was quite usefull for me, and you even comment back to answer my question.
\nI guess we could flag also flag plagiarism in comment, even it is seems a bit mean to me (but maybe necessary). I did not even know that there were a system to report plagiarism.
Sub-question : is the commentary sections of this notebook with plagiarism are filled with \"I upvoted nananan please checked my kernel ?\". I fear that people hunting for up-vote tend to up-vote each other and persist.
", "votes": 9, "replies": [ { "id": 1184229, "author_name": "Jason Dolorso", "author_url": "", "post_date": "2021-02-03T12:58:37.273000", "content": "\n\nI fear that people hunting for up-vote tend to up-vote each other and persist.
\n
For sure. I noticed that they spam newly publish notebooks everywhere.
", "votes": 1, "replies": [] } ] }, { "id": 1175096, "author_name": "Luca Massaron", "author_url": "", "post_date": "2021-01-28T22:29:28.847000", "content": "Rampant plagiarism is simply because people think that they could faster and more easily reach to master and grandmaster title ranks in such a way (discussion and data are also other ways to do so and in such a sense they also have got their own distorsions). Also people think that if they get to that grade they can boast it out and that will bring great advantages to them and their career.
\nSincerely, I do hire people and I do look at what they do on Kaggle, but I absolutely do not consider rankings gained on discussion, notebooks and data unless I see something really unique and original. I do not know if there are around other hiring managers that have a different opinion than mine, but mine is that having good grades in discussion/notebooks/data is generally (there are exceptions, though) really not much predictive of one's ability in data science.
\nDiscussion/notebooks/data were intended to help Kagglers and support them in their learning and competiting and they are much useful without any doubt. Anyway, until you will have grades similar to competitions attached to discussion/notebooks/data, you will always experience a host of plagiars and free riders that you cannot avoid and you cannot mitigate in any way.
", "votes": 7, "replies": [ { "id": 1175169, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2021-01-29T01:14:09.643000", "content": "I agree @lucamassaron , I have been interviewed a few times and my Kaggle titles had no impact so far, what mattered was the experienced I gained here, competing or creating content, sometimes event competition medals have little impact, maybe people got lucky with a random blend.
", "votes": 4, "replies": [] }, { "id": 1197857, "author_name": "sandy1112", "author_url": "", "post_date": "2021-02-12T13:41:43.180000", "content": "100% agree with you @lucamassaron . Even in my case, when I am hiring for my team I do not place much importance on titles gained in Data, Discussions and Notebooks. In terms of hiring, I place importance on Competitions medals - when the candidate can explain the solution and subsequent downstream business value of their solution.
", "votes": 1, "replies": [] } ] }, { "id": 1170298, "author_name": "ryches", "author_url": "", "post_date": "2021-01-26T05:49:50.160000", "content": "This is a genuinely worrying trend to me. Whenever I look at the notebooks for competitions I see increasing numbers of ripoffs and it is becoming harder to find the original source. Very often the ripoffs have important notes removed but rise to the top because they beat the source work by .0001 on the lb.
", "votes": 5, "replies": [] }, { "id": 1176692, "author_name": "Andy Jian Zhou", "author_url": "", "post_date": "2021-01-29T18:20:47.423000", "content": "It is a shame to see your notebooks being plagiarized, I have seen most of your notebooks(you may even recognize me), and it's very cruel for those who have copied your beautiful original notebooks that have helped multitudes of people(including me).
\nDo realize, they won't be the ones improving, they won't have the fun experience making their own notebooks, hence they won't even go far in their career without simply crediting someone else for their work.
\nUltimately, I believe that the competition for kaggle medals has gone so far to the extent people are cheating for them. The reason why I(and many other alike), are using kaggle is a means to gain experience, to learn. As long as we learn something new, we give ourselves a medal. Sorry to hear people are taking your credit away.
", "votes": 3, "replies": [ { "id": 1176903, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2021-01-29T21:55:07.667000", "content": "Thanks for the kind words @andyjianzhou ,
\nI also believe learning should be the main drive, that was the reason I joined Kaggle and is the reason I stay :)
You are absolutely right. When i browse the top competitors in any ongoing competitions
\nAlmost all of them have exact same score.
\nHow is that even possible.
\nThis can not be coincidence definitely.
\nKaggle team should pay more attention to code plagarism.
That's actually quite normal for many people to have the same score, but not the same CV
", "votes": 1, "replies": [] } ] }, { "id": 1160093, "author_name": "Junyi Ng", "author_url": "", "post_date": "2021-01-19T17:06:52.843000", "content": "I feel you. That's why I like to only share helpful concepts and ideas that work, but not the full implementation. It's a shame, your notebook with TPU was really high quality too.
", "votes": 3, "replies": [ { "id": 1161209, "author_name": "Sajid Hussain", "author_url": "", "post_date": "2021-01-20T12:12:59.593000", "content": "I have used your notebook as a reference and mentioned that I have referred it. Will it be okay to do that.
\nhttps://www.kaggle.com/sajidhussain3/beginners-take-on-cassava-kfold-augment-keras
Please go ahead! Feel free to fork notebooks I make public :)
", "votes": 1, "replies": [] } ] }, { "id": 1159903, "author_name": "Jason Dolorso", "author_url": "", "post_date": "2021-01-19T14:34:11.653000", "content": "As someone who is new to kaggle and was listening to lots of kaggle-related podcasts, one of the consensus was really how prevalent the problem of plagiarism is in this platform and how most GMs really think it won't go away easily.
\nThat said, I made sure to develop the habit of placing credits even if i only have a notebook in private.
\nFor what it's worth, your notebooks and your contributions in the discussion are really helpful especially to competition novices like me. Let me know how can I also help report plagiarisms.
", "votes": 3, "replies": [ { "id": 1160009, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2021-01-19T15:56:30.407000", "content": "Yes @jasondolorso , I also agree that plagiarism is a tricky problem to solve, especially because it is so easy to copy code here. But having said that the fork feature is also awesome forking code from experienced people and experimented here at Kaggle was one of the main things that enabled me to learn very fast.
Forking public kernels and adding relevant content on top of it, then publish it, is one of the things the makes Kaggle awesome, and makes progress so fast here, my issue is with people re-publish the same code with no relevant modifications, and sometimes they seem to try to claim ownership of the whole thing.
", "votes": 2, "replies": [] }, { "id": 1160037, "author_name": "Jason Dolorso", "author_url": "", "post_date": "2021-01-19T16:12:31.190000", "content": "I agree. Republishing without *fork *means they really made the effort to either download the code locally and re-uploading, or copying the code cell by cell.
\nThat got me thinking.. With the thousands of data science talents here in Kaggle, especially those with NLP expertise, it would be a great project to have a model that analyzes plagiarism based on code similarity — of course, given some constraints such as not forking or giving credits.
", "votes": 2, "replies": [] }, { "id": 1160066, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2021-01-19T16:41:12.650000", "content": "Maybe that could be an idea for a playground competition.
", "votes": 1, "replies": [] }, { "id": 1160073, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2021-01-19T16:47:15.820000", "content": "Jason
\nPretty sure that the analysis that you mention already exists in Kaggle. Another of the more trouble some \"cheats\" is for folks to work as an informal team and share code between them, but without ever becoming a formal Kaggle team. A disadvantage of forming a kaggle team is the loss of daily submissions. A team of 8 only gets 5 daily submissions, while an informal cheating team of 8 would get 40. Looking for duplicated code, not published in a public kernel is one of the methods to control this cheat.
\nTo the best of my knowledge there is no rule that prohibits an exact copy of a shared kernel from being used for the final submissions and leader board ranking. Sharing a high scoring kernel near the end of a competition is considered by many as a form of cheating - your shared code will get tons of forks and upvotes - increasing your ranking in one of the 4 pillars.
\n\nA disadvantage of forming a kaggle team is the loss of daily submissions. A team of 8 only gets 5 daily submissions, while an informal cheating team of 8 would get 40.
\n
I agree with this. Some competitions, in particular, can be easily exploited by scanning through the testing data.
\nCheaters will always find a way, and we can only do so much to mitigate it. But I believe that you would know a legit great data scientist (or Kaggle Master/GM) when you see how they dispose themselves and through their works. I doubt that those who are gaming the system would really rise up to their levels in terms of popularity and skill.
\nI am also relatively new to the platform though, so take my opinion with a grain of salt.
", "votes": 0, "replies": [] }, { "id": 1160977, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2021-01-20T08:46:45.477000", "content": "\n\nCheaters will always find a way>
\n
I don't cheat. I don't publish kernels unless they are mostly my original work. I do fork and use shared kernels on my local PC and for submission - it's the greatest thing about Kaggle that I can learn from some pretty sharp folks who share their code.
\nIn the years I have been active on the site, Kaggle has implemented lots of things to prevent \"cheaters\". IMO many of the changes make it extremely difficult for novice folks to participate. One only needs to look at all the posts in the last year, of various competitions, for folks trying to solve \"submission not found\". Pretty sure that this single attempt to \"prevent cheating\" has wasted thousands of hours of novice's time and driven a pretty decent percentage of them away from the site in frustration.
\nSo my general rule about cheater discussions on Kaggle - folks - just STFU.
", "votes": -3, "replies": [] }, { "id": 1161117, "author_name": "Yovin Yahathugoda", "author_url": "", "post_date": "2021-01-20T10:42:16.510000", "content": "@pcjimmmy I don't understand why you are being so aggressive towards the topic being discussed here. Also I don't think Kaggle made any changes that prevent novices from learning, please provide some links if I am wrong. All @dimitreoliveira said was that people keep copying his notebook and publishing them as their own work without crediting the original notebook, which is frustrating because people don't even have the decency to be grateful to the original author. This behavior is also against Kaggle community guidelines.
\n\n\nfor folks trying to solve \"submission not found\". Pretty sure that this single attempt to \"prevent cheating\" has wasted thousands of hours of novice's time
\n
How is the \"submission not found\" error related to anti-cheating? This issue is faced by only a handful of participants and the community helps them solve the issue when they post it on a discussion.
", "votes": 3, "replies": [] }, { "id": 1161596, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2021-01-20T16:30:17.497000", "content": "Yovin
\nSorry if I appear aggressive. It is frustration - not aggression that was intended.
\nWhen you and I joined Kaggle we could generate a submission file on our local PC and get a LB score within a minute or two. I could process a wide range of ideas, generate the code, make a submission and get a score in rapid fashion. I almost always was using my daily allowance of submissions. I was a complete newbie at R and that system let me have a very rapid learning curve.
\nCheaters spent their time probing the private test set and other actions that took unfair use of the direct submission and test data availability.
\nThe current submission process, as practiced in this competition, is the apparent direct response by Kaggle to prevent those forms of cheating. Now to validate an idea I need to save/submit. Wait a period of time while the public test set is run. Come back and submit - wait up to 9 hours for the code to run, come back and finally get a score. For the last year or more for competitions using this cheat prevention method I almost NEVER use my daily quota of submissions - the process to turn over a single idea takes a full day. When things error out - only a very limited and almost useless bit of information is provided to help debug. I never had a submission not found error back in the days when I could make a direct submission. I never recall a post in discussion asking for help for those type of errors.
\nKaggle knows the failure rate for this type of submission error. They should know how many new folks are hitting that initial roadblock. They can probably estimate the rate at which folks stop being active on the site and how many submission failures they had before they stopped. I spent 50+ years processing data to solve problems - most of the time that needed to be done in my head, I think I got pretty good at making general guesses that would accurate enough for me to solve the issue. My guess is that it's a lot more than a \"handful\". A quick search of this completion with \"submission error\" gets me two handfuls. Not sure I want to try a search for all competitions :}
\nI am not sure the \"community helps\" as much as they can - I know you are very active as a helper.
\nKaggle has done lots of good things over my years here to address the specific issue of the poster. You can report kernels, when a kernel is forked there is an automated citation added, etc.
\nMy frustration is that every time a new form of cheating is invented or a new complaint gains traction what follows is ALWAYS pain for me.
", "votes": 0, "replies": [] }, { "id": 1161637, "author_name": "Yovin Yahathugoda", "author_url": "", "post_date": "2021-01-20T16:59:35.267000", "content": "@pcjimmmy The main reason for having a public/private re-run of the submission notebook is to make sure the solution presented to the host of the competition is capable of running efficiently. When people were allowed to submit the CSV file, each participant's code will require different computing resources, which may not be ideal for the competition host since some participants may have extremely powerful local workstations for training and inference.
\nBy setting a rule to submit your solution through the Kaggle kernel with the same amount of resources available to everyone, it produces a more real-world deployable model. This is why you see different submission run times for some competitions because the host expects our solution to work within the specified time and computing resources available. While this also reduces cheating, it is not the main reason for this change.
", "votes": 1, "replies": [] }, { "id": 1161706, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2021-01-20T17:42:19.620000", "content": "Going to follow my own advice and STFU - but two last comments. Apparently one form of whining can get you a gold metal while my form gets me down votes :) Besides - I have drifted too far from the original posters topic.
\nWhen I directly submitted a submission file you are correct in that I could have generated that file using a Supercomputer that I happened to have access to. But pretty sure that the bottom 1000 folks who clearly have submission file issues in this competition are not using super computers to generate the model that than fails. With the current GPU and TPU quota's I not sure how competition hosts are benefiting from us all having to play with the same limited toys. A real-world deployable model is hard to reach on 30 hours a week with a full day turn around for each iteration.
\nSo I don't see it - The current submission process does NOTHING to prevent me from using a super computer to build a model. It just adds one cumbersome step to the submission process.
\nTake a look at the leader board - when I just looked there are 11 folks who's best score is clearly a submission error. There several hundred above that who's best score is likely also a type of submission error. There are more than a handful of names on that list that I am willing to bet will not be seen as active next month. (my standard bet is a Big Ben sandwich at Parnalli's in downtown Pittsburgh - GO STEELERS)
", "votes": -4, "replies": [] } ] }, { "id": 1167756, "author_name": "Janio Martinez Bachmann", "author_url": "", "post_date": "2021-01-24T13:16:53.053000", "content": "Hey totally agree, I've seen plagiarism frequently in the past months. I guess some individuals plagiarise to get quick upvotes in the community. Nevertheless, I think it is disrispectful to the person who has dedicated so much effort and time towards publishing a notebook. It took me once 3 months to publish a notebook imagine that, how would I feel if others take credit for all that effort and sacrifice I did in those 3 months? I also had once someone who contacted me on LinkedIn so that I could sign the individual's thesis and I found out that one part of the person's thesis was exactly one of the notebooks I published in Kaggle. You really see sometimes this kind of situations but to be honest I don't know how much can Kaggle do other than investigating the notebook when someone reports plagiarism. Anyways, your notebooks are amazing and don't feel discourage, keep sharing your amazing work!
", "votes": 4, "replies": [ { "id": 1167880, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2021-01-24T14:45:09.603000", "content": "Hi @janiobachmann , oh boy what a story 😊, when you open a notebook it is clear if the author has put real effort in making that, I am sure that the 3-month notebook must be awesome, we don't always get the reaction that we expect, but we know it is not only for the upvotes. Thank you for your kind words, I will keep them coming.
", "votes": 2, "replies": [] } ] }, { "id": 1204415, "author_name": "Vicky Goyal", "author_url": "", "post_date": "2021-02-16T06:02:05.187000", "content": "Taking credit for someone else work is really shameful. It takes hours and days of hard work to publish the good work. I think the discussion should be going forward Is there a way to make it difficult for cheaters?
", "votes": 1, "replies": [] }, { "id": 1204170, "author_name": "Miloud Belarebia", "author_url": "", "post_date": "2021-02-15T22:33:06.103000", "content": "I totally agree, what I see is that every day it's increasing, but I think most kagglers realizes that, so keep up the good work 👌
", "votes": 1, "replies": [] }, { "id": 1183584, "author_name": "Muhammad Ali", "author_url": "", "post_date": "2021-02-03T05:29:47.617000", "content": "Yes I agree with you this is shameful activity.
", "votes": 1, "replies": [] }, { "id": 1178330, "author_name": "Elizabeth Bradford", "author_url": "", "post_date": "2021-01-30T19:44:27.420000", "content": "hi, I just wanted to add, as a community member who often looks to notebooks to learn, this is definitely an issue on Kaggle and has been for some time. I think we struggle as a community to continue to support an open resource environment for the sake of improvement while holding every contributor, viewer and member to the appropriate levels of integrity and anti-plaigarism standards. While I am a relatively new member, I wonder if Kaggle has considered adding a \"report this post\" button similar to Instagram, etc. Realistically this would require more hands on monitoring, but I think it could be more successful than simply an upvote or downvote. Potentially it could make sense to nominate community members to help moderate competitions and postings? Again, this is more work, but I think it could potentially be worth it.
\nI also love the comment to switch to R, Etienne. I didn't think of that, but you are probably right.
", "votes": 1, "replies": [] }, { "id": 1169201, "author_name": "neelg007", "author_url": "", "post_date": "2021-01-25T11:33:45.680000", "content": "Honestly, can someone explain why Kaggle cannot implement checks of that sort in their notebooks? I have seen these type of threads many times all over the site. I can't imagine it would require much effort for making a model to detect plagiarism in code (kinda like the already existing document plagiarism checkers).
\nThen Kaggle could show a small private message to the user listing all the sources copied from, and ask if the user would like to credit those sources in their notebooks. It seems a pretty decent \"First-Response\" method to discourage Plagarism. AFAIK Kaggle hasn't implemented anything of the sort.
\nKaggle = Google. This kind of issue should have been resolved years ago. It's alright to copy notebooks as long as you give them the due credit. IMO Kaggle/Google does not take this issue very seriously OR they don't realize how demotivating it is to have your work copied and not given due credit.
", "votes": 1, "replies": [] }, { "id": 1161002, "author_name": "jeeyoon_kwak", "author_url": "", "post_date": "2021-01-20T08:59:54.827000", "content": "Thanks to your opinion, I have a lot to think about.
", "votes": 1, "replies": [] }, { "id": 1160821, "author_name": "Dave E", "author_url": "", "post_date": "2021-01-20T06:42:54.443000", "content": "I have personally used some of your setup code for Tensorflow, so thank you. Think I have upvoted your notebooks which I've made use of. I've not shared any public notebooks in this comp.
\nIt does seem to me that if Kaggle could find a way to make this more automatic in crediting original notebook, it would fix the problem both for people intentionally and unintentionally sharing copies without crediting. I suppose people could still make a copy via methods other than forking, but at least that would highlight deliberate examples.
", "votes": 1, "replies": [ { "id": 1161152, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2021-01-20T11:19:21.907000", "content": "You're welcome @davidedwards1 ,
\nActually, Kaggle does have a way to give credit to the original notebook, when you fork a notebook you will see a reference on the upper right side, like this:
\n
I agree with you. Personally, I think it's very unfortunate.
", "votes": 1, "replies": [] }, { "id": 1160362, "author_name": "Gabriel Prado", "author_url": "", "post_date": "2021-01-19T21:19:08.447000", "content": "I've seen a couple things (even that i used myself this competition) that i don't even remember who's the original author such as the GradualWarmup code. But thanks for sharing your frustration, this was my first competition so i wasn't 100% aware of this sort of problems. But thanks to your post i'll definitely be always posting credits and links on the appropriate sections from now on.
\nMaybe kaggle could implement a method, such as:
\nif notebook is forked and notebook.author != notebook.fork.author
\nas to not award points and medals to discourage people from just simply copying notebooks.
@capiru I wouldn't go as far as giving credits to every section of code (unless maybe it is something very unique), it would be too much work to find the original authors, as you said.
\nCopying code and hacking it, is a great tool to learn, and I encourage it.
", "votes": 0, "replies": [] } ] }, { "id": 1162003, "author_name": "CSD", "author_url": "", "post_date": "2021-01-20T22:31:39.257000", "content": "Olá,
\nNotei que é BR, então irei lhe responder em português. Enfim, é realmente um grande desrespeito copiar e até mesmo divulgar um notebook sem dar os devidos créditos, porém não creio que seja motivo para ficar muito zangado. Não tiro sua razão, mas no fundo, sabemos que o que importa é que você tem a capacidade, o conhecimento e fez o esforço para criar um material interessante para a comunidade, e quem só copia algo não costuma ir longe.
\nGrande abraço,
\n…
Oi @olavomendes ,
\nNa verdade eu nem fico muito chateado em ver esse tipo de coisa, venho publicando notebooks aqui ja tem um tempo, e certamente irei continuar. O que realmente importa eh isso tudo ai que voce mencionou, mas tambem acho que eh importante levantar esses topicos eventualmente, ate para que o pessoal de suporte da plataforma ficar ciente de problemas do tipo.
\nOutro abraco 😄
", "votes": 1, "replies": [] } ] }, { "id": 1160483, "author_name": "Jason Dolorso", "author_url": "", "post_date": "2021-01-20T00:45:00.220000", "content": "\n\npeople doing that are novices or contributors
\n
I also want to look at it from another angle and think that some of those novices/contributors are actually unaware of properly forking or the meta-etiquettes here in kaggle (albeit its a pretty well-known morally that you should credit people's work).
\nTo remove that aspect, maybe the platform can implement reminders when it's someone's first time opening, downloading and/or forking a notebook. The contents of this comment, that I dug up a few weeks ago when I was wary about forking a notebook, would be a good start:
\nhttps://www.kaggle.com/c/siim-acr-pneumothorax-segmentation/discussion/97571#563571
I am not sure if it is detailed in the terms and conditions when signing up (who reads them?) but it is probable that someone would overlook it or even forget it by the time they would first work on a notebook.
", "votes": 2, "replies": [] }, { "id": 1160032, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2021-01-19T16:10:39.757000", "content": "One of the four pillars of the Kaggle profile rating system is Notebook Contributor. You gain ranking as folks copy and fork, along with up vote your notebook.
\nIf you do not want your work copied your solution is real simple - don't share it.
\nYou can also pay attention to notebooks and NOT up vote those that are copies with no improvements. Careless up votes are a reason why \"plagiarism\" can exist. So do your part.
\nGM's became GM's because they chose to share great code.
\nKaggle has made lots of changes over the couple of years I have been active to prevent folks from gaming the ranking system - IMO lots of the changes have hurt the novice 99% who do not game the system. It's a very hard task to make rules that help more than they hurt. I think they are trying hard.
\nA version of your post has appeared in almost every competition that I have followed for the last several years - hmmm - is that plagiarism? :)
", "votes": 0, "replies": [ { "id": 1160082, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2021-01-19T16:56:23.833000", "content": "I don't think you gain rakings by other people copying or forking your code.
\nAlso, I don't think that not sharing is the solution here, but paying attention to code before upvoting is, as you mentioned, sharing code is what makes Kaggle what it is, in my opinion. But I agree that ending plagiarism is not an easy task.
", "votes": 1, "replies": [] }, { "id": 1160320, "author_name": "Avik", "author_url": "", "post_date": "2021-01-19T20:22:04.207000", "content": "@pcjimmmy Republishing someone's public code claiming it as one's own without giving due credits is infact plagiarism, Kaggle kernels are licensed under Apache 2.0 as well. Most people change a couple parameters and earn clout for it- it's bound to annoy the original author.
", "votes": 3, "replies": [] }, { "id": 1161012, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2021-01-20T09:05:34.017000", "content": "Hmm - tried to find the criteria for ranking in the \"Notebooks\" pillar - it seems like up votes and fork count create metals. So your ranking in this pillar is the result of folks copying and forking, with probably a small percentage of them giving you an upvote along the way.
\nA lot of the issue I see with \"copy and paste\" kernels seems to be mostly folks who are pretty new to Kaggle. Lot's of them don't know the desired community standards for giving credit - up voting if you like and use shared code, etc.
\nI have sometimes offered a comment in a shared kernel suggesting that the person give credit and identify what they changed, but that seems later when I re-read my post to be me being a bully.
\nDon't know the answer, but IMO the last thing needed is a new set of rules that make life on Kaggle difficult for the 99% who use it in an honest manner.
", "votes": 0, "replies": [] } ] }, { "id": 2380632, "author_name": "Aneekesh Kumar", "author_url": "", "post_date": "2023-08-08T17:25:50.127000", "content": "You might consider it as plagiarism and find it unexpectable but from a different perspective it isn't really, they copy your code and start trying better preprocessing, better models and it just gives them a better headstart. For beginners who want fast growth in the field, they don't really have great resources to understand Machine/Deep Learning Pipelines and simply the Flow. 😭
\nKaggle Novices/Contributers generally don't have a great Python sense or a sense of code, this helps them get the concepts without starting from scratch which definitely is way faster.
\nSome people do just want medals and so just ignore them and look at the bright side. 😇
I'm a noob. I'm here just to learn and I think quite a few people are here for the same. To copy and paste another's code and call it on my is not a learning process. But on that note, I looked at other codes and used the same procedure usually when it comes to preparing the data, not the modeling and training that part is where I'm trying to build a good model on my own merit. Is this wrong of me or is this cheating? Because I do not want to cheat
", "votes": 0, "replies": [] }, { "id": 1194948, "author_name": "Haotian Huang", "author_url": "", "post_date": "2021-02-10T13:24:07.227000", "content": "I agree with you and no more consent.
", "votes": 0, "replies": [] }, { "id": 1395491, "author_name": "Saman", "author_url": "", "post_date": "2021-07-21T09:05:06.347000", "content": "Unfortunately, I have found so many notebooks also that copied others' works.
\nFor example, if you check the following link, you may find an exact copy of my work.
\nhttps://www.kaggle.com/satpreetmakhija/netflix-movies-and-tv-shows-2021/code
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1159692": "I was browsing through the notebooks section and I quickly find almost 20 notebooks that are essentially exact copies of some of my work posted here, people doing that are novices or contributors, but still, I find it disrespectful, to have your work copied and posted as someones else, after investing many hours into creating something helpful for the community, at some level, this always happens in competitions but here it was more than I expected.\n\nFor those who have reported notebooks for plagiarism, does anything happens at a reasonable time?", "1160160": "You're doing great work, and I think they'll be hard to see in other competitions after all. So please continue your great work. @dimitreoliveira ", "1177950": "I have a solution for you : start coding in Julia or R.\n\nMore seriously yes, I guess it is a pain. Citing properly work is the base of scientific demarch, especially for people that want to go deeper into details. I actually commented one of your kernels recently, which was quite usefull for me, and you even comment back to answer my question.\nI guess we could flag also flag plagiarism in comment, even it is seems a bit mean to me (but maybe necessary). I did not even know that there were a system to report plagiarism. \n\nSub-question : is the commentary sections of this notebook with plagiarism are filled with \"I upvoted nananan please checked my kernel ?\". I fear that people hunting for up-vote tend to up-vote each other and persist. ", "1175096": "Rampant plagiarism is simply because people think that they could faster and more easily reach to master and grandmaster title ranks in such a way (discussion and data are also other ways to do so and in such a sense they also have got their own distorsions). Also people think that if they get to that grade they can boast it out and that will bring great advantages to them and their career.\n\nSincerely, I do hire people and I do look at what they do on Kaggle, but I absolutely do not consider rankings gained on discussion, notebooks and data unless I see something really unique and original. I do not know if there are around other hiring managers that have a different opinion than mine, but mine is that having good grades in discussion/notebooks/data is generally (there are exceptions, though) really not much predictive of one's ability in data science. \n\nDiscussion/notebooks/data were intended to help Kagglers and support them in their learning and competiting and they are much useful without any doubt. Anyway, until you will have grades similar to competitions attached to discussion/notebooks/data, you will always experience a host of plagiars and free riders that you cannot avoid and you cannot mitigate in any way. ", "1170298": "This is a genuinely worrying trend to me. Whenever I look at the notebooks for competitions I see increasing numbers of ripoffs and it is becoming harder to find the original source. Very often the ripoffs have important notes removed but rise to the top because they beat the source work by .0001 on the lb. ", "1176692": "It is a shame to see your notebooks being plagiarized, I have seen most of your notebooks(you may even recognize me), and it's very cruel for those who have copied your beautiful original notebooks that have helped multitudes of people(including me).\n\nDo realize, they won't be the ones improving, they won't have the fun experience making their own notebooks, hence they won't even go far in their career without simply crediting someone else for their work. \n\nUltimately, I believe that the competition for kaggle medals has gone so far to the extent people are cheating for them. The reason why I(and many other alike), are using kaggle is a means to gain experience, to learn. As long as we learn something new, we give ourselves a medal. Sorry to hear people are taking your credit away. ", "1174821": "You are absolutely right. When i browse the top competitors in any ongoing competitions\nAlmost all of them have exact same score.\nHow is that even possible.\nThis can not be coincidence definitely.\nKaggle team should pay more attention to code plagarism.", "1160093": "I feel you. That's why I like to only share helpful concepts and ideas that work, but not the full implementation. It's a shame, your notebook with TPU was really high quality too.", "1159903": "As someone who is new to kaggle and was listening to lots of kaggle-related podcasts, one of the consensus was really how prevalent the problem of plagiarism is in this platform and how most GMs really think it won't go away easily.\n\nThat said, I made sure to develop the habit of placing credits even if i only have a notebook in private.\n\nFor what it's worth, your notebooks and your contributions in the discussion are really helpful especially to competition novices like me. Let me know how can I also help report plagiarisms.", "1167756": "Hey totally agree, I've seen plagiarism frequently in the past months. I guess some individuals plagiarise to get quick upvotes in the community. Nevertheless, I think it is disrispectful to the person who has dedicated so much effort and time towards publishing a notebook. It took me once 3 months to publish a notebook imagine that, how would I feel if others take credit for all that effort and sacrifice I did in those 3 months? I also had once someone who contacted me on LinkedIn so that I could sign the individual's thesis and I found out that one part of the person's thesis was exactly one of the notebooks I published in Kaggle. You really see sometimes this kind of situations but to be honest I don't know how much can Kaggle do other than investigating the notebook when someone reports plagiarism. Anyways, your notebooks are amazing and don't feel discourage, keep sharing your amazing work!", "1204415": "Taking credit for someone else work is really shameful. It takes hours and days of hard work to publish the good work. I think the discussion should be going forward Is there a way to make it difficult for cheaters? ", "1204170": "I totally agree, what I see is that every day it's increasing, but I think most kagglers realizes that, so keep up the good work 👌", "1183584": "Yes I agree with you this is shameful activity.", "1178330": "hi, I just wanted to add, as a community member who often looks to notebooks to learn, this is definitely an issue on Kaggle and has been for some time. I think we struggle as a community to continue to support an open resource environment for the sake of improvement while holding every contributor, viewer and member to the appropriate levels of integrity and anti-plaigarism standards. While I am a relatively new member, I wonder if Kaggle has considered adding a \"report this post\" button similar to Instagram, etc. Realistically this would require more hands on monitoring, but I think it could be more successful than simply an upvote or downvote. Potentially it could make sense to nominate community members to help moderate competitions and postings? Again, this is more work, but I think it could potentially be worth it.\n\nI also love the comment to switch to R, Etienne. I didn't think of that, but you are probably right. ", "1169201": "Honestly, can someone explain why Kaggle cannot implement checks of that sort in their notebooks? I have seen these type of threads many times all over the site. I can't imagine it would require much effort for making a model to detect plagiarism in code (kinda like the already existing document plagiarism checkers). \n\nThen Kaggle could show a small private message to the user listing all the sources copied from, and ask if the user would like to credit those sources in their notebooks. It seems a pretty decent \"First-Response\" method to discourage Plagarism. AFAIK Kaggle hasn't implemented anything of the sort. \n\nKaggle = Google. This kind of issue should have been resolved years ago. It's alright to copy notebooks as long as you give them the due credit. IMO Kaggle/Google does not take this issue very seriously OR they don't realize how **demotivating** it is to have your work copied and not given due credit.", "1161002": "Thanks to your opinion, I have a lot to think about.", "1160821": "I have personally used some of your setup code for Tensorflow, so thank you. Think I have upvoted your notebooks which I've made use of. I've not shared any public notebooks in this comp.\n\nIt does seem to me that if Kaggle could find a way to make this more automatic in crediting original notebook, it would fix the problem both for people intentionally and unintentionally sharing copies without crediting. I suppose people could still make a copy via methods other than forking, but at least that would highlight deliberate examples.", "1160479": "I agree with you. Personally, I think it's very unfortunate.", "1160362": "I've seen a couple things (even that i used myself this competition) that i don't even remember who's the original author such as the GradualWarmup code. But thanks for sharing your frustration, this was my first competition so i wasn't 100% aware of this sort of problems. But thanks to your post i'll definitely be always posting credits and links on the appropriate sections from now on.\nMaybe kaggle could implement a method, such as: \n`if notebook is forked and notebook.author != notebook.fork.author`\nas to not award points and medals to discourage people from just simply copying notebooks.", "1162003": "Olá,\n\nNotei que é BR, então irei lhe responder em português. Enfim, é realmente um grande desrespeito copiar e até mesmo divulgar um notebook sem dar os devidos créditos, porém não creio que seja motivo para ficar muito zangado. Não tiro sua razão, mas no fundo, sabemos que o que importa é que você tem a capacidade, o conhecimento e fez o esforço para criar um material interessante para a comunidade, e quem só copia algo não costuma ir longe.\n\nGrande abraço,\n...", "1160483": "> people doing that are novices or contributors\n\nI also want to look at it from another angle and think that some of those novices/contributors are actually unaware of properly forking or the *meta-etiquettes* here in kaggle (albeit its a pretty well-known morally that you should credit people's work). \n\nTo remove that aspect, maybe the platform can implement reminders when it's someone's first time opening, downloading and/or forking a notebook. The contents of this comment, that I dug up a few weeks ago when I was wary about forking a notebook, would be a good start:\nhttps://www.kaggle.com/c/siim-acr-pneumothorax-segmentation/discussion/97571#563571\n\nI am not sure if it is detailed in the terms and conditions when signing up (who reads them?) but it is probable that someone would overlook it or even forget it by the time they would first work on a notebook.", "1160032": "One of the four pillars of the Kaggle profile rating system is Notebook Contributor. You gain ranking as folks copy and fork, along with up vote your notebook.\n\nIf you do not want your work copied your solution is real simple - don't share it.\n\nYou can also pay attention to notebooks and NOT up vote those that are copies with no improvements. Careless up votes are a reason why \"plagiarism\" can exist. So do your part.\n\nGM's became GM's because they chose to share great code.\n\nKaggle has made lots of changes over the couple of years I have been active to prevent folks from gaming the ranking system - IMO lots of the changes have hurt the novice 99% who do not game the system. It's a very hard task to make rules that help more than they hurt. I think they are trying hard.\n\nA version of your post has appeared in almost every competition that I have followed for the last several years - hmmm - is that plagiarism? :)\n", "2380632": "You might consider it as plagiarism and find it unexpectable but from a different perspective it isn't really, they copy your code and start trying better preprocessing, better models and it just gives them a better headstart. For beginners who want fast growth in the field, they don't really have great resources to understand Machine/Deep Learning Pipelines and simply the Flow. 😭\nKaggle Novices/Contributers generally don't have a great Python sense or a sense of code, this helps them get the concepts without starting from scratch which definitely is way faster. \nSome people do just want medals and so just ignore them and look at the bright side. 😇", "1246478": "I'm a noob. I'm here just to learn and I think quite a few people are here for the same. To copy and paste another's code and call it on my is not a learning process. But on that note, I looked at other codes and used the same procedure usually when it comes to preparing the data, not the modeling and training that part is where I'm trying to build a good model on my own merit. Is this wrong of me or is this cheating? Because I do not want to cheat", "1194948": " I agree with you and no more consent.", "1395491": "Unfortunately, I have found so many notebooks also that copied others' works. \n\nFor example, if you check the following link, you may find an exact copy of my work.\n\n\nhttps://www.kaggle.com/satpreetmakhija/netflix-movies-and-tv-shows-2021/code" } }