Michael, a participant of this competition, downloaded all the data and really spent $2k to label them. After that he sent the labeled data (~100k) to his buddy Lincoln.
\n\nNow is the show time of Lincoln. He combined the ImageNet training data and the labeled competition data to train a powerful ‘object classification’ ConvNet, and even wrote a paper as the only author to arXiv with title such as ‘balabala convolutional neural network for object classification’. Along with this paper, the powerful pre-trained model has also been released, and was even posted on Caffe model zoo (now publicly available). God, it is not the so called ‘object classification pre-trained model’. It is a distracted driver detection model, but nobody knows that.
\n\nMichael received the model and posted the link in the kaggle forum to announce he used it together with VGG19, ResNet, GoogleNet. However, who will really notice that. If someone notice it, who dare to use the model in a zero citation arXiv paper (who is Lincoln, never heard). Nevertheless, with the powerful over fitted model, Michael could get any performance as long as he wants, and all he should concern is not too obvious. Michael got the first place finally, but all the other kagglers think he is suspect. No evidence at all.
", "messages": [ { "id": "115680", "postDate": "04/19/2016 15:40:42", "content": "Michael, a participant of this competition, downloaded all the data and really spent $2k to label them. After that he sent the labeled data (~100k) to his buddy Lincoln.
\n\nNow is the show time of Lincoln. He combined the ImageNet training data and the labeled competition data to train a powerful ‘object classification’ ConvNet, and even wrote a paper as the only author to arXiv with title such as ‘balabala convolutional neural network for object classification’. Along with this paper, the powerful pre-trained model has also been released, and was even posted on Caffe model zoo (now publicly available). God, it is not the so called ‘object classification pre-trained model’. It is a distracted driver detection model, but nobody knows that.
\n\nMichael received the model and posted the link in the kaggle forum to announce he used it together with VGG19, ResNet, GoogleNet. However, who will really notice that. If someone notice it, who dare to use the model in a zero citation arXiv paper (who is Lincoln, never heard). Nevertheless, with the powerful over fitted model, Michael could get any performance as long as he wants, and all he should concern is not too obvious. Michael got the first place finally, but all the other kagglers think he is suspect. No evidence at all.
", "rawMarkdown": "Michael, a participant of this competition, downloaded all the data and really spent $2k to label them. After that he sent the labeled data (~100k) to his buddy Lincoln. \r\n\r\nNow is the show time of Lincoln. He combined the ImageNet training data and the labeled competition data to train a powerful ‘object classification’ ConvNet, and even wrote a paper as the only author to arXiv with title such as ‘balabala convolutional neural network for object classification’. Along with this paper, the powerful pre-trained model has also been released, and was even posted on Caffe model zoo (now publicly available). God, it is not the so called ‘object classification pre-trained model’. It is a distracted driver detection model, but nobody knows that.\r\n\r\nMichael received the model and posted the link in the kaggle forum to announce he used it together with VGG19, ResNet, GoogleNet. However, who will really notice that. If someone notice it, who dare to use the model in a zero citation arXiv paper (who is Lincoln, never heard). Nevertheless, with the powerful over fitted model, Michael could get any performance as long as he wants, and all he should concern is not too obvious. Michael got the first place finally, but all the other kagglers think he is suspect. No evidence at all.", "votes": null }, { "id": "115712", "postDate": "04/19/2016 17:32:27", "content": "Are they open to new team members? (Asking for a friend.)
", "rawMarkdown": "Are they open to new team members? (Asking for a friend.)", "votes": null }, { "id": "115798", "postDate": "04/20/2016 07:14:36", "content": "Sounds a little far-fetched and wouldn't it be discovered as soon as some new unseen data was fed into the solution? The solution which the winner is obliged to provide to the host, if I understand the rules correctly.
\n\nI guess if the competition host or organizers were concerned about this possibility, they could hold back some data (or create new data) to test the winning solution against before the award was finalized.
", "rawMarkdown": "Sounds a little far-fetched and wouldn't it be discovered as soon as some new unseen data was fed into the solution? The solution which the winner is obliged to provide to the host, if I understand the rules correctly.\r\n\r\nI guess if the competition host or organizers were concerned about this possibility, they could hold back some data (or create new data) to test the winning solution against before the award was finalized.", "votes": null }, { "id": "115991", "postDate": "04/21/2016 08:41:14", "content": "There is always a possibility for a scenario when someone finds a way to break the rules and to cheat. It is true not only for this competition but for the society as whole. However, almost always it would end up bad for Michael. \nNot only Michael would damage his karma, and his good reputation in the community, but at the end he most likely would not be able to benefit from it. If the model achieves high standing, it is very easy for organizers to see what is going on, and to disqualify it. If it does not - then yes, I guess it is possible to go undiscovered. However, there is no point to break the rules in this case, it is much easier than just use public scripts to achieve basically the same outcome ...
", "rawMarkdown": "There is always a possibility for a scenario when someone finds a way to break the rules and to cheat. It is true not only for this competition but for the society as whole. However, almost always it would end up bad for Michael. \r\nNot only Michael would damage his karma, and his good reputation in the community, but at the end he most likely would not be able to benefit from it. If the model achieves high standing, it is very easy for organizers to see what is going on, and to disqualify it. If it does not - then yes, I guess it is possible to go undiscovered. However, there is no point to break the rules in this case, it is much easier than just use public scripts to achieve basically the same outcome ...", "votes": null } ], "comments": [ { "id": 115712, "author_name": "inversion", "author_url": "", "post_date": "04/19/2016 17:32:27", "content": "Are they open to new team members? (Asking for a friend.)
", "votes": null, "replies": [] }, { "id": 115798, "author_name": "gauss256", "author_url": "", "post_date": "04/20/2016 07:14:36", "content": "Sounds a little far-fetched and wouldn't it be discovered as soon as some new unseen data was fed into the solution? The solution which the winner is obliged to provide to the host, if I understand the rules correctly.
\n\nI guess if the competition host or organizers were concerned about this possibility, they could hold back some data (or create new data) to test the winning solution against before the award was finalized.
", "votes": null, "replies": [] }, { "id": 115991, "author_name": "alexlzzz", "author_url": "", "post_date": "04/21/2016 08:41:14", "content": "There is always a possibility for a scenario when someone finds a way to break the rules and to cheat. It is true not only for this competition but for the society as whole. However, almost always it would end up bad for Michael. \nNot only Michael would damage his karma, and his good reputation in the community, but at the end he most likely would not be able to benefit from it. If the model achieves high standing, it is very easy for organizers to see what is going on, and to disqualify it. If it does not - then yes, I guess it is possible to go undiscovered. However, there is no point to break the rules in this case, it is much easier than just use public scripts to achieve basically the same outcome ...
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "115680": "Michael, a participant of this competition, downloaded all the data and really spent $2k to label them. After that he sent the labeled data (~100k) to his buddy Lincoln. \r\n\r\nNow is the show time of Lincoln. He combined the ImageNet training data and the labeled competition data to train a powerful ‘object classification’ ConvNet, and even wrote a paper as the only author to arXiv with title such as ‘balabala convolutional neural network for object classification’. Along with this paper, the powerful pre-trained model has also been released, and was even posted on Caffe model zoo (now publicly available). God, it is not the so called ‘object classification pre-trained model’. It is a distracted driver detection model, but nobody knows that.\r\n\r\nMichael received the model and posted the link in the kaggle forum to announce he used it together with VGG19, ResNet, GoogleNet. However, who will really notice that. If someone notice it, who dare to use the model in a zero citation arXiv paper (who is Lincoln, never heard). Nevertheless, with the powerful over fitted model, Michael could get any performance as long as he wants, and all he should concern is not too obvious. Michael got the first place finally, but all the other kagglers think he is suspect. No evidence at all.", "115712": "Are they open to new team members? (Asking for a friend.)", "115798": "Sounds a little far-fetched and wouldn't it be discovered as soon as some new unseen data was fed into the solution? The solution which the winner is obliged to provide to the host, if I understand the rules correctly.\r\n\r\nI guess if the competition host or organizers were concerned about this possibility, they could hold back some data (or create new data) to test the winning solution against before the award was finalized.", "115991": "There is always a possibility for a scenario when someone finds a way to break the rules and to cheat. It is true not only for this competition but for the society as whole. However, almost always it would end up bad for Michael. \r\nNot only Michael would damage his karma, and his good reputation in the community, but at the end he most likely would not be able to benefit from it. If the model achieves high standing, it is very easy for organizers to see what is going on, and to disqualify it. If it does not - then yes, I guess it is possible to go undiscovered. However, there is no point to break the rules in this case, it is much easier than just use public scripts to achieve basically the same outcome ..." }, "source": "meta" }