{ "id": 19495, "title": "The final rank and prize?", "url": "/competitions/second-annual-data-science-bowl/discussion/19495", "author_name": "Aaron Chen", "post_date": "2016-03-14T05:54:09.507000", "votes": -2, "comment_count": 13, "views": 1589, "content": "

Hi, I want to verify one thing:

\n\n

the final leader board consists the team with model uploaded and those who didn't. What about a team without a model reached top 3? Will the prize be cancelled or given to the next team with uploaded model?

", "messages": [ { "id": 111409, "postDate": "2016-03-14T15:53:53.637Z", "content": "

it is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).

", "rawMarkdown": "it is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).\r\n\r\n", "votes": 5 }, { "id": 111449, "postDate": "2016-03-14T18:35:58.737Z", "content": "

Everyone should take a deep breath and relax for five and half hours. The leaderboard means almost nothing at the moment and will be almost certainly look completely different once it is rescored on the 99% percent of the training data that is used for the final ranking. After that, feel free to complain about any weirdnesses that you see, but personally I expect that the leaderboard tonight will look a lot like it did last week before the test data was released.

", "rawMarkdown": "Everyone should take a deep breath and relax for five and half hours. The leaderboard means almost nothing at the moment and will be almost certainly look completely different once it is rescored on the 99% percent of the training data that is used for the final ranking. After that, feel free to complain about any weirdnesses that you see, but personally I expect that the leaderboard tonight will look a lot like it did last week before the test data was released.", "votes": 6 }, { "id": 111378, "postDate": "2016-03-14T09:06:46.083Z", "content": "

Im really solly to break this to you, but you guys cooperated outside teams and this is stricly forbiden. Whats happens is that you will get banned and wont get ranking points or money.

\n\n

Besides you are considering to much a dataset with only 6 samples. Having done good here and no so good on the validation set means that, almost for sure, you did overfit badly.

\n\n

Sorry :(

", "rawMarkdown": "Im really solly to break this to you, but you guys cooperated outside teams and this is stricly forbiden. Whats happens is that you will get banned and wont get ranking points or money. \r\n\r\nBesides you are considering to much a dataset with only 6 samples. Having done good here and no so good on the validation set means that, almost for sure, you did overfit badly.\r\n\r\nSorry :(", "votes": 6 }, { "id": 111371, "postDate": "2016-03-14T06:32:09.587Z", "content": "

And further:

\n\n

If a team with model submit a result generated from other model and reached a higher rank, and they also submit a result generated from origin model. How to count the rank?

\n\n

Currently, we have three models,

\n\n
    \n

  1. original one

    2. \n

  2. a slightly modified one, which is different at two points:

    3. \n
\n\n

a. a coefficient. because we started the contest just three weeks ago, we don't have time to verify every hyper-para, and unfortunately, we run out of submission chance before model deadline, so we have to left this coefficient to be determined in stage 2. Actually automatically determining it is very easy, which should be less than 5 lines, but we forget to implement it because in the stage 1 deadline we are so busy at writing document and form up a frame work. As you can see, we upload the model just 1h before deadline.

\n\n

b. data-preprocess. As you may notice, in several cases, different sax-slice has different view angle and resolution. In model 1, I abandoned all of them, but in this week, I found it caused many problems in test and validation, so I rotate and crop them to the same shape.

\n\n
    \n
  1. a highly modified one, which should be better
    2. \n
\n\n

So I want to know, whether the second model is considered legal for prize competition. And considering my last post, can I also submit model 3 for a higher rank?

", "rawMarkdown": "And further:\r\n\r\nIf a team with model submit a result generated from other model and reached a higher rank, and they also submit a result generated from origin model. How to count the rank?\r\n\r\nCurrently, we have three models,\r\n\r\n1. original one\r\n\r\n2. a slightly modified one, which is different at two points:\r\n\r\na. a coefficient. because we started the contest just three weeks ago, we don't have time to verify every hyper-para, and unfortunately, we run out of submission chance before model deadline, so we have to left this coefficient to be determined in stage 2. Actually automatically determining it is very easy, which should be less than 5 lines, but we forget to implement it because in the stage 1 deadline we are so busy at writing document and form up a frame work. As you can see, we upload the model just 1h before deadline.\r\n\r\nb. data-preprocess. As you may notice, in several cases, different sax-slice has different view angle and resolution. In model 1, I abandoned all of them, but in this week, I found it caused many problems in test and validation, so I rotate and crop them to the same shape. \r\n\r\n3. a highly modified one, which should be better\r\n\r\nSo I want to know, whether the second model is considered legal for prize competition. And considering my last post, can I also submit model 3 for a higher rank?", "votes": 1 }, { "id": 111456, "postDate": "2016-03-14T19:16:48.680Z", "content": "

Don't be too nervous guy, even now I can't reach your score in validation, so it's highly possible that you can get back. I am so curious that how can u get such good results with existence of case 595 and 599, and look forward to see your algorithm

\n\n

Well, the admin never said those behavior are banned, those teams without model can submit their results, why can't I? I fully understand if admin refuse to pay me for modified model, but I just want to see what can I achieve in 3 weeks, and chase for a higher rank for honor.

\n\n

Finally, I agree that this competition rule is over complicated yet far from complete and cause many confusions.

\n\n

[quote=woshialex;111409]

\n\n

it is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).

\n\n

[/quote]

", "rawMarkdown": "Don't be too nervous guy, even now I can't reach your score in validation, so it's highly possible that you can get back. I am so curious that how can u get such good results with existence of case 595 and 599, and look forward to see your algorithm\r\n\r\nWell, the admin never said those behavior are banned, those teams without model can submit their results, why can't I? I fully understand if admin refuse to pay me for modified model, but I just want to see what can I achieve in 3 weeks, and chase for a higher rank for honor.\r\n\r\nFinally, I agree that this competition rule is over complicated yet far from complete and cause many confusions.\r\n\r\n[quote=woshialex;111409]\r\n\r\nit is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).\r\n\r\n\r\n\r\n[/quote]\r\n", "votes": 1 }, { "id": 111444, "postDate": "2016-03-14T17:55:41.810Z", "content": "

[quote=DavidGbodiOdaibo;111437]

\n\n

Thank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!

\n\n

[/quote]

\n\n

Then night is still young ;) I think woshialex has a very good chance to get back to the top 3 after all ;)

\n\n

However, personally, I think that all this process was needlessly complicated, forcing people to spend precious time on developing bullet proof fully automated software instead of working on the science of it. Even major SW houses most of the time fail to create such level of reliability - a famous Blue Screen of Death comes to mind.

", "rawMarkdown": "[quote=DavidGbodiOdaibo;111437]\r\n\r\nThank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!\r\n\r\n[/quote]\r\n\r\nThen night is still young ;) I think woshialex has a very good chance to get back to the top 3 after all ;)\r\n\r\nHowever, personally, I think that all this process was needlessly complicated, forcing people to spend precious time on developing bullet proof fully automated software instead of working on the science of it. Even major SW houses most of the time fail to create such level of reliability - a famous Blue Screen of Death comes to mind. ", "votes": 1 }, { "id": 111437, "postDate": "2016-03-14T17:38:48.903Z", "content": "

Thank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!

", "rawMarkdown": "Thank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!" }, { "id": 111386, "postDate": "2016-03-14T09:42:35.440Z", "content": "

Please Let me clarify the issue from the beginning.

\n\n

Two week ago Liao and I saw the competition and join it without hesitation. Soon I registered a team call show_me_the_money and invite Liao. He did receive the inviting email and click the url, however forget to push the button in the html page. It has been overdue when we realized the mistake.

\n\n

Within the competition, Liao and me work together in a team, using the same team show_me_the_momey to submit. Thus, we did not cooperated outside teams, because Liao never submit in another team. We only have 3 chance a day just as a normal team.

\n\n

Hope this can eliminate the misunderstanding.

\n\n

BTW, I just want to know whether kaggle can add Liao to show_me_the_money back. You can check the invitation email in the Kaggle system and know that I told the truth.

", "rawMarkdown": "Please Let me clarify the issue from the beginning. \r\n\r\nTwo week ago Liao and I saw the competition and join it without hesitation. Soon I registered a team call **show_me_the_money** and invite Liao. He did receive the inviting email and click the url, however **forget to push the button** in the html page. It has been overdue when we realized the mistake. \r\n\r\nWithin the competition, Liao and me work together in a team, using the same team **show_me_the_momey** to submit. Thus, we **did not** cooperated outside teams, because Liao **never** submit in another team. We only have 3 chance a day just as a normal team.\r\n\r\nHope this can eliminate the misunderstanding. \r\n\r\nBTW, I just want to know whether kaggle can add Liao to **show_me_the_money** back. You can check the invitation email in the Kaggle system and know that I told the truth." }, { "id": 111382, "postDate": "2016-03-14T09:18:50.500Z", "content": "

@Leustagos

", "rawMarkdown": "@Leustagos ", "votes": -2 }, { "id": 111381, "postDate": "2016-03-14T09:17:28.687Z", "content": "

Firstly, It's not a cooperation outside teams, because I didn't submit anything in stage 1, in your system, I didn't join this competition, and the reason that I didn't join, I have clarified in the other post

\n\n

Secondly, I totally agree with you that this result doesn't make sense, but because we rank 8th in stage 1 and improves a lot in stage 2, I believe our final rank won't be too low. Further more, we submit only twice (actually only once after modifying code), this ranking could be result of randomness but surely it's not overfit

\n\n

Thank you for your reply

", "rawMarkdown": "Firstly, It's not a cooperation outside teams, because I didn't submit anything in stage 1, in your system, I didn't join this competition, and the reason that I didn't join, I have clarified in the other post\r\n\r\nSecondly, I totally agree with you that this result doesn't make sense, but because we rank 8th in stage 1 and improves a lot in stage 2, I believe our final rank won't be too low. Further more, we submit only twice (actually only once after modifying code), this ranking could be result of randomness but surely it's not overfit\r\n\r\nThank you for your reply", "votes": -2 }, { "id": 111369, "postDate": "2016-03-14T05:54:09.507Z", "content": "

Hi, I want to verify one thing:

\n\n

the final leader board consists the team with model uploaded and those who didn't. What about a team without a model reached top 3? Will the prize be cancelled or given to the next team with uploaded model?

", "rawMarkdown": "Hi, I want to verify one thing:\r\n\r\nthe final leader board consists the team with model uploaded and those who didn't. What about a team without a model reached top 3? Will the prize be cancelled or given to the next team with uploaded model?", "votes": -2 }, { "id": 111473, "postDate": "2016-03-14T21:30:21.427Z", "content": "

Ideally the organizer should allow teams to separately pick submissions for the prize and for a leader board. That will also save themselves some trouble trying to reproduce some non-reproducible results.\n They never clarified whether submitting a submission not producible from the submitted model will totally disqualify one team from winning the prize, that is, even when one's reproducible results are still top three among all the reproducible models. I think that's why many top teams including us refrain themselves from uploading better submissions.

", "rawMarkdown": "Ideally the organizer should allow teams to separately pick submissions for the prize and for a leader board. That will also save themselves some trouble trying to reproduce some non-reproducible results.\r\n They never clarified whether submitting a submission not producible from the submitted model will totally disqualify one team from winning the prize, that is, even when one's reproducible results are still top three among all the reproducible models. I think that's why many top teams including us refrain themselves from uploading better submissions.\r\n\r\n" }, { "id": 111422, "postDate": "2016-03-14T16:38:33.867Z", "content": "

To be eligible for a prize, the model has to be exactly the same than the one you uploaded.\nIf you don't care about the prize you can submit your slightly modified version, as an admin explains in the following quoted post:

\n\n

[quote=William Cukierski;110967]

\n\n

You may still submit to the leaderboard, but to be prize eligible it would need to come from the model you uploaded.

\n\n

[/quote]

", "rawMarkdown": "To be eligible for a prize, the model has to be exactly the same than the one you uploaded.\r\nIf you don't care about the prize you can submit your slightly modified version, as an admin explains in the following quoted post:\r\n\r\n[quote=William Cukierski;110967]\r\n\r\nYou may still submit to the leaderboard, but to be prize eligible it would need to come from the model you uploaded.\r\n\r\n[/quote]\r\n\r\n" }, { "id": 111447, "postDate": "2016-03-14T18:22:18.687Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ], "comments": [ { "id": 111409, "author_name": "woshialex", "author_url": "", "post_date": "2016-03-14T15:53:53.637000", "content": "

it is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).

", "votes": 5, "replies": [] }, { "id": 111449, "author_name": "Tim Hochberg", "author_url": "", "post_date": "2016-03-14T18:35:58.737000", "content": "

Everyone should take a deep breath and relax for five and half hours. The leaderboard means almost nothing at the moment and will be almost certainly look completely different once it is rescored on the 99% percent of the training data that is used for the final ranking. After that, feel free to complain about any weirdnesses that you see, but personally I expect that the leaderboard tonight will look a lot like it did last week before the test data was released.

", "votes": 6, "replies": [] }, { "id": 111378, "author_name": "Leustagos", "author_url": "", "post_date": "2016-03-14T09:06:46.083000", "content": "

Im really solly to break this to you, but you guys cooperated outside teams and this is stricly forbiden. Whats happens is that you will get banned and wont get ranking points or money.

\n\n

Besides you are considering to much a dataset with only 6 samples. Having done good here and no so good on the validation set means that, almost for sure, you did overfit badly.

\n\n

Sorry :(

", "votes": 6, "replies": [] }, { "id": 111371, "author_name": "Aaron Chen", "author_url": "", "post_date": "2016-03-14T06:32:09.587000", "content": "

And further:

\n\n

If a team with model submit a result generated from other model and reached a higher rank, and they also submit a result generated from origin model. How to count the rank?

\n\n

Currently, we have three models,

\n\n
    \n

  1. original one

    2. \n

  2. a slightly modified one, which is different at two points:

    3. \n
\n\n

a. a coefficient. because we started the contest just three weeks ago, we don't have time to verify every hyper-para, and unfortunately, we run out of submission chance before model deadline, so we have to left this coefficient to be determined in stage 2. Actually automatically determining it is very easy, which should be less than 5 lines, but we forget to implement it because in the stage 1 deadline we are so busy at writing document and form up a frame work. As you can see, we upload the model just 1h before deadline.

\n\n

b. data-preprocess. As you may notice, in several cases, different sax-slice has different view angle and resolution. In model 1, I abandoned all of them, but in this week, I found it caused many problems in test and validation, so I rotate and crop them to the same shape.

\n\n
    \n
  1. a highly modified one, which should be better
    2. \n
\n\n

So I want to know, whether the second model is considered legal for prize competition. And considering my last post, can I also submit model 3 for a higher rank?

", "votes": 1, "replies": [] }, { "id": 111456, "author_name": "Fangzhou Liao", "author_url": "", "post_date": "2016-03-14T19:16:48.680000", "content": "

Don't be too nervous guy, even now I can't reach your score in validation, so it's highly possible that you can get back. I am so curious that how can u get such good results with existence of case 595 and 599, and look forward to see your algorithm

\n\n

Well, the admin never said those behavior are banned, those teams without model can submit their results, why can't I? I fully understand if admin refuse to pay me for modified model, but I just want to see what can I achieve in 3 weeks, and chase for a higher rank for honor.

\n\n

Finally, I agree that this competition rule is over complicated yet far from complete and cause many confusions.

\n\n

[quote=woshialex;111409]

\n\n

it is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).

\n\n

[/quote]

", "votes": 1, "replies": [] }, { "id": 111444, "author_name": "==>(AL)<==", "author_url": "", "post_date": "2016-03-14T17:55:41.810000", "content": "

[quote=DavidGbodiOdaibo;111437]

\n\n

Thank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!

\n\n

[/quote]

\n\n

Then night is still young ;) I think woshialex has a very good chance to get back to the top 3 after all ;)

\n\n

However, personally, I think that all this process was needlessly complicated, forcing people to spend precious time on developing bullet proof fully automated software instead of working on the science of it. Even major SW houses most of the time fail to create such level of reliability - a famous Blue Screen of Death comes to mind.

", "votes": 1, "replies": [] }, { "id": 111437, "author_name": "DavidGbodiOdaibo", "author_url": "", "post_date": "2016-03-14T17:38:48.903000", "content": "

Thank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!

", "votes": 0, "replies": [] }, { "id": 111386, "author_name": "Aaron Chen", "author_url": "", "post_date": "2016-03-14T09:42:35.440000", "content": "

Please Let me clarify the issue from the beginning.

\n\n

Two week ago Liao and I saw the competition and join it without hesitation. Soon I registered a team call show_me_the_money and invite Liao. He did receive the inviting email and click the url, however forget to push the button in the html page. It has been overdue when we realized the mistake.

\n\n

Within the competition, Liao and me work together in a team, using the same team show_me_the_momey to submit. Thus, we did not cooperated outside teams, because Liao never submit in another team. We only have 3 chance a day just as a normal team.

\n\n

Hope this can eliminate the misunderstanding.

\n\n

BTW, I just want to know whether kaggle can add Liao to show_me_the_money back. You can check the invitation email in the Kaggle system and know that I told the truth.

", "votes": 0, "replies": [] }, { "id": 111382, "author_name": "Fangzhou Liao", "author_url": "", "post_date": "2016-03-14T09:18:50.500000", "content": "

@Leustagos

", "votes": -2, "replies": [] }, { "id": 111381, "author_name": "Fangzhou Liao", "author_url": "", "post_date": "2016-03-14T09:17:28.687000", "content": "

Firstly, It's not a cooperation outside teams, because I didn't submit anything in stage 1, in your system, I didn't join this competition, and the reason that I didn't join, I have clarified in the other post

\n\n

Secondly, I totally agree with you that this result doesn't make sense, but because we rank 8th in stage 1 and improves a lot in stage 2, I believe our final rank won't be too low. Further more, we submit only twice (actually only once after modifying code), this ranking could be result of randomness but surely it's not overfit

\n\n

Thank you for your reply

", "votes": -2, "replies": [] }, { "id": 111473, "author_name": "Wei Dong", "author_url": "", "post_date": "2016-03-14T21:30:21.427000", "content": "

Ideally the organizer should allow teams to separately pick submissions for the prize and for a leader board. That will also save themselves some trouble trying to reproduce some non-reproducible results.\n They never clarified whether submitting a submission not producible from the submitted model will totally disqualify one team from winning the prize, that is, even when one's reproducible results are still top three among all the reproducible models. I think that's why many top teams including us refrain themselves from uploading better submissions.

", "votes": 0, "replies": [] }, { "id": 111422, "author_name": "Vincent L.", "author_url": "", "post_date": "2016-03-14T16:38:33.867000", "content": "

To be eligible for a prize, the model has to be exactly the same than the one you uploaded.\nIf you don't care about the prize you can submit your slightly modified version, as an admin explains in the following quoted post:

\n\n

[quote=William Cukierski;110967]

\n\n

You may still submit to the leaderboard, but to be prize eligible it would need to come from the model you uploaded.

\n\n

[/quote]

", "votes": 0, "replies": [] }, { "id": 111447, "author_name": "", "author_url": "", "post_date": "2016-03-14T18:22:18.687000", "content": "", "votes": -1, "replies": [] } ], "raw_markdown_by_id": { "111409": "it is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).\r\n\r\n", "111449": "Everyone should take a deep breath and relax for five and half hours. The leaderboard means almost nothing at the moment and will be almost certainly look completely different once it is rescored on the 99% percent of the training data that is used for the final ranking. After that, feel free to complain about any weirdnesses that you see, but personally I expect that the leaderboard tonight will look a lot like it did last week before the test data was released.", "111378": "Im really solly to break this to you, but you guys cooperated outside teams and this is stricly forbiden. Whats happens is that you will get banned and wont get ranking points or money. \r\n\r\nBesides you are considering to much a dataset with only 6 samples. Having done good here and no so good on the validation set means that, almost for sure, you did overfit badly.\r\n\r\nSorry :(", "111371": "And further:\r\n\r\nIf a team with model submit a result generated from other model and reached a higher rank, and they also submit a result generated from origin model. How to count the rank?\r\n\r\nCurrently, we have three models,\r\n\r\n1. original one\r\n\r\n2. a slightly modified one, which is different at two points:\r\n\r\na. a coefficient. because we started the contest just three weeks ago, we don't have time to verify every hyper-para, and unfortunately, we run out of submission chance before model deadline, so we have to left this coefficient to be determined in stage 2. Actually automatically determining it is very easy, which should be less than 5 lines, but we forget to implement it because in the stage 1 deadline we are so busy at writing document and form up a frame work. As you can see, we upload the model just 1h before deadline.\r\n\r\nb. data-preprocess. As you may notice, in several cases, different sax-slice has different view angle and resolution. In model 1, I abandoned all of them, but in this week, I found it caused many problems in test and validation, so I rotate and crop them to the same shape. \r\n\r\n3. a highly modified one, which should be better\r\n\r\nSo I want to know, whether the second model is considered legal for prize competition. And considering my last post, can I also submit model 3 for a higher rank?", "111456": "Don't be too nervous guy, even now I can't reach your score in validation, so it's highly possible that you can get back. I am so curious that how can u get such good results with existence of case 595 and 599, and look forward to see your algorithm\r\n\r\nWell, the admin never said those behavior are banned, those teams without model can submit their results, why can't I? I fully understand if admin refuse to pay me for modified model, but I just want to see what can I achieve in 3 weeks, and chase for a higher rank for honor.\r\n\r\nFinally, I agree that this competition rule is over complicated yet far from complete and cause many confusions.\r\n\r\n[quote=woshialex;111409]\r\n\r\nit is very clear that only your version 1 (original model that submitted before stage 2) can enter for the final ranking. there is no excuse for modifying anything that changes the model. I think if you insist picking your v2 or v3 for your final standing you'll simply be removed from the competition... So wish you have a good luck with your original code. You can still check the performance of your updated code after the competition ends and see if it really could rank good (top 3).\r\n\r\n\r\n\r\n[/quote]\r\n", "111444": "[quote=DavidGbodiOdaibo;111437]\r\n\r\nThank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!\r\n\r\n[/quote]\r\n\r\nThen night is still young ;) I think woshialex has a very good chance to get back to the top 3 after all ;)\r\n\r\nHowever, personally, I think that all this process was needlessly complicated, forcing people to spend precious time on developing bullet proof fully automated software instead of working on the science of it. Even major SW houses most of the time fail to create such level of reliability - a famous Blue Screen of Death comes to mind. ", "111437": "Thank God I am not in the top 3, oh!! The emotional rollercoaster!! I see woshialex is not in the top 3 anymore. So close but yet so far!", "111386": "Please Let me clarify the issue from the beginning. \r\n\r\nTwo week ago Liao and I saw the competition and join it without hesitation. Soon I registered a team call **show_me_the_money** and invite Liao. He did receive the inviting email and click the url, however **forget to push the button** in the html page. It has been overdue when we realized the mistake. \r\n\r\nWithin the competition, Liao and me work together in a team, using the same team **show_me_the_momey** to submit. Thus, we **did not** cooperated outside teams, because Liao **never** submit in another team. We only have 3 chance a day just as a normal team.\r\n\r\nHope this can eliminate the misunderstanding. \r\n\r\nBTW, I just want to know whether kaggle can add Liao to **show_me_the_money** back. You can check the invitation email in the Kaggle system and know that I told the truth.", "111382": "@Leustagos ", "111381": "Firstly, It's not a cooperation outside teams, because I didn't submit anything in stage 1, in your system, I didn't join this competition, and the reason that I didn't join, I have clarified in the other post\r\n\r\nSecondly, I totally agree with you that this result doesn't make sense, but because we rank 8th in stage 1 and improves a lot in stage 2, I believe our final rank won't be too low. Further more, we submit only twice (actually only once after modifying code), this ranking could be result of randomness but surely it's not overfit\r\n\r\nThank you for your reply", "111369": "Hi, I want to verify one thing:\r\n\r\nthe final leader board consists the team with model uploaded and those who didn't. What about a team without a model reached top 3? Will the prize be cancelled or given to the next team with uploaded model?", "111473": "Ideally the organizer should allow teams to separately pick submissions for the prize and for a leader board. That will also save themselves some trouble trying to reproduce some non-reproducible results.\r\n They never clarified whether submitting a submission not producible from the submitted model will totally disqualify one team from winning the prize, that is, even when one's reproducible results are still top three among all the reproducible models. I think that's why many top teams including us refrain themselves from uploading better submissions.\r\n\r\n", "111422": "To be eligible for a prize, the model has to be exactly the same than the one you uploaded.\r\nIf you don't care about the prize you can submit your slightly modified version, as an admin explains in the following quoted post:\r\n\r\n[quote=William Cukierski;110967]\r\n\r\nYou may still submit to the leaderboard, but to be prize eligible it would need to come from the model you uploaded.\r\n\r\n[/quote]\r\n\r\n", "111447": "" } }