Looks like there are a lot of concerns around folks cheating by not actually using generative models to create dog images but rather submitting existing dog images or slightly altered versions. How about we make it a requirement that all final submission kernels be made public? That way all final submissions are open to scrutiny from other competitors so that we don't have to worry about Kaggle's bandwidth in checking solutions to see if they are valid? We can have a post competition thread to notify violations. Just a thought.
", "messages": [ { "id": 566444, "postDate": "2019-07-02T07:10:31.833Z", "content": "Looks like there are a lot of concerns around folks cheating by not actually using generative models to create dog images but rather submitting existing dog images or slightly altered versions. How about we make it a requirement that all final submission kernels be made public? That way all final submissions are open to scrutiny from other competitors so that we don't have to worry about Kaggle's bandwidth in checking solutions to see if they are valid? We can have a post competition thread to notify violations. Just a thought.
", "rawMarkdown": "Looks like there are a lot of concerns around folks cheating by not actually using generative models to create dog images but rather submitting existing dog images or slightly altered versions. How about we make it a requirement that all final submission kernels be made public? That way all final submissions are open to scrutiny from other competitors so that we don't have to worry about Kaggle's bandwidth in checking solutions to see if they are valid? We can have a post competition thread to notify violations. Just a thought.", "votes": 80 }, { "id": 566496, "postDate": "2019-07-02T08:38:49.087Z", "content": "Yes. I wish I could upvote this twice.
\n\nAs Kaggle probably won't check solutions below gold, peer reviewing is an excellent solution and will discourage people trying to steal a medal without respecting the rules. \nPeople that have nothing to hide will share their code, and Kaggle's Team can check other's solutions.
", "rawMarkdown": "Yes. I wish I could upvote this twice.\n\nAs Kaggle probably won't check solutions below gold, peer reviewing is an excellent solution and will discourage people trying to steal a medal without respecting the rules. \nPeople that have nothing to hide will share their code, and Kaggle's Team can check other's solutions.", "votes": 14, "replies": [ { "id": 566566, "postDate": "2019-07-02T10:33:17.517Z", "content": "I would argue that some participants might want to hide their solutions in order to use them again in the following competitions. It may not be applicable for this particular competition though.
", "rawMarkdown": "I would argue that some participants might want to hide their solutions in order to use them again in the following competitions. It may not be applicable for this particular competition though.", "votes": -12 }, { "id": 566974, "postDate": "2019-07-02T21:28:28.020Z", "content": "@ppleskov I would argue that they might want to hide their solutions because there is something to hide... -.-'
", "rawMarkdown": "@ppleskov I would argue that they might want to hide their solutions because there is something to hide... -.-'", "votes": 18 }, { "id": 589051, "postDate": "2019-07-31T11:48:14.550Z", "content": "I agree with you @raimonds1993
", "rawMarkdown": "I agree with you @raimonds1993 ", "votes": 3 } ] }, { "id": 588274, "postDate": "2019-07-30T11:15:11.257Z", "content": "This is a nice suggestion, this is like Kaggle could crowdsource to verify the ligitimacy of the code which is difficult to automate using code at this point.\nA thing to appreciate here is that Kaggle is attempting to challenge the traditional method of competitions and bringing up something new.
", "rawMarkdown": "This is a nice suggestion, this is like Kaggle could crowdsource to verify the ligitimacy of the code which is difficult to automate using code at this point.\nA thing to appreciate here is that Kaggle is attempting to challenge the traditional method of competitions and bringing up something new.", "votes": 1 }, { "id": 568083, "postDate": "2019-07-04T11:25:01.307Z", "content": "Proper validation technique for GAN competition will certainly be very helpful. As it has been mentioned in rules, i am fully surprised that how can somebody can validate all the image manually .
", "rawMarkdown": "Proper validation technique for GAN competition will certainly be very helpful. As it has been mentioned in rules, i am fully surprised that how can somebody can validate all the image manually . ", "votes": 1 }, { "id": 567926, "postDate": "2019-07-04T06:58:34.457Z", "content": "This is the most reasonable idea for competition like this.
", "rawMarkdown": "This is the most reasonable idea for competition like this.", "votes": 1 }, { "id": 567834, "postDate": "2019-07-04T04:12:31.403Z", "content": "That's a great idea peer reviewing is also used by some competitive coding website. Its also a great way to learn
", "rawMarkdown": "That's a great idea peer reviewing is also used by some competitive coding website. Its also a great way to learn", "votes": 1 }, { "id": 567494, "postDate": "2019-07-03T15:38:16.523Z", "content": "Great Thought, I like your idea.
", "rawMarkdown": "Great Thought, I like your idea.", "votes": 1 }, { "id": 567164, "postDate": "2019-07-03T06:08:08.577Z", "content": "Definitely. I'm surprised that this isn't the case already.
", "rawMarkdown": "Definitely. I'm surprised that this isn't the case already.", "votes": 2 }, { "id": 566467, "postDate": "2019-07-02T07:57:14.713Z", "content": "Yes, I agree.\nKaggle should come up with an objective validation technique for GAN competitions.
", "rawMarkdown": "Yes, I agree.\nKaggle should come up with an objective validation technique for GAN competitions.", "votes": 2 }, { "id": 566595, "postDate": "2019-07-02T10:59:35.850Z", "content": "Good thought. I can only see some licensing problems, because all code is made open source by publishing the kernels.
", "rawMarkdown": "Good thought. I can only see some licensing problems, because all code is made open source by publishing the kernels. " }, { "id": 588374, "postDate": "2019-07-30T13:29:51.207Z", "content": "It would be good practice for that kind of competition and should be announced at the start of the competition as a mandatory rule to be eligible for medals/prizes, however, it should be made ahead. That is how people learn - by making a mistake
", "rawMarkdown": "It would be good practice for that kind of competition and should be announced at the start of the competition as a mandatory rule to be eligible for medals/prizes, however, it should be made ahead. That is how people learn - by making a mistake\n", "votes": -5, "replies": [ { "id": 589173, "postDate": "2019-07-31T15:04:39.117Z", "content": "Why learning from mistakes while you can scrap the internet and get all the knowledge you need?
", "rawMarkdown": "Why learning from mistakes while you can scrap the internet and get all the knowledge you need?" }, { "id": 589183, "postDate": "2019-07-31T15:24:28.607Z", "content": "Do you expect a meaningful answer or that is not the purpose of your comment?
", "rawMarkdown": "Do you expect a meaningful answer or that is not the purpose of your comment?" }, { "id": 589190, "postDate": "2019-07-31T15:45:57.180Z", "content": "You are definitely my favorite X5 person 😃
", "rawMarkdown": "You are definitely my favorite X5 person 😃 ", "votes": -2 }, { "id": 589278, "postDate": "2019-07-31T17:36:49.583Z", "content": "Learning to rank
", "rawMarkdown": "Learning to rank", "votes": -2 } ] }, { "id": 567993, "postDate": "2019-07-04T08:27:34.480Z", "rawMarkdown": "", "votes": -4, "isDeleted": true } ], "comments": [ { "id": 566496, "author_name": "Theo Viel", "author_url": "", "post_date": "2019-07-02T08:38:49.087000", "content": "Yes. I wish I could upvote this twice.
\n\nAs Kaggle probably won't check solutions below gold, peer reviewing is an excellent solution and will discourage people trying to steal a medal without respecting the rules. \nPeople that have nothing to hide will share their code, and Kaggle's Team can check other's solutions.
", "votes": 14, "replies": [ { "id": 566566, "author_name": "", "author_url": "", "post_date": "2019-07-02T10:33:17.517000", "content": "I would argue that some participants might want to hide their solutions in order to use them again in the following competitions. It may not be applicable for this particular competition though.
", "votes": -12, "replies": [] }, { "id": 566974, "author_name": "Federico Raimondi Cominesi", "author_url": "", "post_date": "2019-07-02T21:28:28.020000", "content": "@ppleskov I would argue that they might want to hide their solutions because there is something to hide... -.-'
", "votes": 18, "replies": [] }, { "id": 589051, "author_name": "torch", "author_url": "", "post_date": "2019-07-31T11:48:14.550000", "content": "I agree with you @raimonds1993
", "votes": 3, "replies": [] } ] }, { "id": 588274, "author_name": "Vishy", "author_url": "", "post_date": "2019-07-30T11:15:11.257000", "content": "This is a nice suggestion, this is like Kaggle could crowdsource to verify the ligitimacy of the code which is difficult to automate using code at this point.\nA thing to appreciate here is that Kaggle is attempting to challenge the traditional method of competitions and bringing up something new.
", "votes": 1, "replies": [] }, { "id": 568083, "author_name": "Raju Kumar Mishra", "author_url": "", "post_date": "2019-07-04T11:25:01.307000", "content": "Proper validation technique for GAN competition will certainly be very helpful. As it has been mentioned in rules, i am fully surprised that how can somebody can validate all the image manually .
", "votes": 1, "replies": [] }, { "id": 567926, "author_name": "Anna Novikova", "author_url": "", "post_date": "2019-07-04T06:58:34.457000", "content": "This is the most reasonable idea for competition like this.
", "votes": 1, "replies": [] }, { "id": 567834, "author_name": "Akash Singh", "author_url": "", "post_date": "2019-07-04T04:12:31.403000", "content": "That's a great idea peer reviewing is also used by some competitive coding website. Its also a great way to learn
", "votes": 1, "replies": [] }, { "id": 567494, "author_name": "Gina Degregorio", "author_url": "", "post_date": "2019-07-03T15:38:16.523000", "content": "Great Thought, I like your idea.
", "votes": 1, "replies": [] }, { "id": 567164, "author_name": "Ed", "author_url": "", "post_date": "2019-07-03T06:08:08.577000", "content": "Definitely. I'm surprised that this isn't the case already.
", "votes": 2, "replies": [] }, { "id": 566467, "author_name": "Srihari Pramod", "author_url": "", "post_date": "2019-07-02T07:57:14.713000", "content": "Yes, I agree.\nKaggle should come up with an objective validation technique for GAN competitions.
", "votes": 2, "replies": [] }, { "id": 566595, "author_name": "Pascal Pfeiffer", "author_url": "", "post_date": "2019-07-02T10:59:35.850000", "content": "Good thought. I can only see some licensing problems, because all code is made open source by publishing the kernels.
", "votes": 0, "replies": [] }, { "id": 588374, "author_name": "ValeriyBabushkin", "author_url": "", "post_date": "2019-07-30T13:29:51.207000", "content": "It would be good practice for that kind of competition and should be announced at the start of the competition as a mandatory rule to be eligible for medals/prizes, however, it should be made ahead. That is how people learn - by making a mistake
", "votes": -5, "replies": [ { "id": 589173, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2019-07-31T15:04:39.117000", "content": "Why learning from mistakes while you can scrap the internet and get all the knowledge you need?
", "votes": 0, "replies": [] }, { "id": 589183, "author_name": "ValeriyBabushkin", "author_url": "", "post_date": "2019-07-31T15:24:28.607000", "content": "Do you expect a meaningful answer or that is not the purpose of your comment?
", "votes": 0, "replies": [] }, { "id": 589190, "author_name": "Ahmet Erdem", "author_url": "", "post_date": "2019-07-31T15:45:57.180000", "content": "You are definitely my favorite X5 person 😃
", "votes": -2, "replies": [] }, { "id": 589278, "author_name": "ValeriyBabushkin", "author_url": "", "post_date": "2019-07-31T17:36:49.583000", "content": "Learning to rank
", "votes": -2, "replies": [] } ] }, { "id": 567993, "author_name": "", "author_url": "", "post_date": "2019-07-04T08:27:34.480000", "content": "", "votes": -4, "replies": [] } ], "raw_markdown_by_id": { "566444": "Looks like there are a lot of concerns around folks cheating by not actually using generative models to create dog images but rather submitting existing dog images or slightly altered versions. How about we make it a requirement that all final submission kernels be made public? That way all final submissions are open to scrutiny from other competitors so that we don't have to worry about Kaggle's bandwidth in checking solutions to see if they are valid? We can have a post competition thread to notify violations. Just a thought.", "566496": "Yes. I wish I could upvote this twice.\n\nAs Kaggle probably won't check solutions below gold, peer reviewing is an excellent solution and will discourage people trying to steal a medal without respecting the rules. \nPeople that have nothing to hide will share their code, and Kaggle's Team can check other's solutions.", "588274": "This is a nice suggestion, this is like Kaggle could crowdsource to verify the ligitimacy of the code which is difficult to automate using code at this point.\nA thing to appreciate here is that Kaggle is attempting to challenge the traditional method of competitions and bringing up something new.", "568083": "Proper validation technique for GAN competition will certainly be very helpful. As it has been mentioned in rules, i am fully surprised that how can somebody can validate all the image manually . ", "567926": "This is the most reasonable idea for competition like this.", "567834": "That's a great idea peer reviewing is also used by some competitive coding website. Its also a great way to learn", "567494": "Great Thought, I like your idea.", "567164": "Definitely. I'm surprised that this isn't the case already.", "566467": "Yes, I agree.\nKaggle should come up with an objective validation technique for GAN competitions.", "566595": "Good thought. I can only see some licensing problems, because all code is made open source by publishing the kernels. ", "588374": "It would be good practice for that kind of competition and should be announced at the start of the competition as a mandatory rule to be eligible for medals/prizes, however, it should be made ahead. That is how people learn - by making a mistake\n", "567993": "" } }