{ "id": 54366, "title": "Humorous take on Stage 2", "url": "/competitions/data-science-bowl-2018/discussion/54366", "author_name": "", "post_date": "2018-04-12T14:53:06.627786500Z", "votes": 81, "comment_count": 24, "views": 0, "content": "

\"enter

", "messages": [ { "id": "312938", "postDate": "04/12/2018 14:53:06", "content": "

\"enter

", "rawMarkdown": "![enter image description here][1]\n\n\n [1]: http://i.imgur.com/j84ci6q.jpg", "votes": null }, { "id": "312939", "postDate": "04/12/2018 14:58:01", "content": "

I’m looking forward to bumping this thread after the phase two leaderboard is public. 🤣

", "rawMarkdown": "I’m looking forward to bumping this thread after the phase two leaderboard is public. 🤣", "votes": null }, { "id": "313033", "postDate": "04/12/2018 17:29:08", "content": "

good one

", "rawMarkdown": "good one", "votes": null }, { "id": "313037", "postDate": "04/12/2018 17:33:36", "content": "

I agree, what a waste of time and money.

\n\n

Best summary ever of a competition.

", "rawMarkdown": "I agree, what a waste of time and money. \n\nBest summary ever of a competition.", "votes": null }, { "id": "313045", "postDate": "04/12/2018 17:44:08", "content": "

Can't agree more.

", "rawMarkdown": "Can't agree more.", "votes": null }, { "id": "313049", "postDate": "04/12/2018 18:03:12", "content": "

OMG!!!!!!!!! I love it. I can't stop laughing.

", "rawMarkdown": "OMG!!!!!!!!! I love it. I can't stop laughing.", "votes": null }, { "id": "313280", "postDate": "04/13/2018 03:30:14", "content": "

Hello，the score is hidden on leader board？ but why we also see 0.000 on ”My Submissions“ page

", "rawMarkdown": "Hello，the score is hidden on leader board？ but why we also see 0.000 on ”My Submissions“ page", "votes": null }, { "id": "313282", "postDate": "04/13/2018 03:34:48", "content": "

To prevent leaderboard probing, apparently

", "rawMarkdown": "To prevent leaderboard probing, apparently", "votes": null }, { "id": "313294", "postDate": "04/13/2018 04:18:42", "content": "

But you don't know which of your model is better

", "rawMarkdown": "But you don't know which of your model is better", "votes": null }, { "id": "313358", "postDate": "04/13/2018 06:42:42", "content": "

@xiapidan You just use all parameters from best model of Stage1. This is intentional.

", "rawMarkdown": "xiapidan You just use all parameters from best model of Stage1. This is intentional.", "votes": null }, { "id": "313444", "postDate": "04/13/2018 09:10:16", "content": "

HA! HA! I like lottery. : )

", "rawMarkdown": "HA! HA! I like lottery. : )", "votes": null }, { "id": "313673", "postDate": "04/13/2018 17:24:19", "content": "

@xiapidan

\n\n

In the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...

", "rawMarkdown": "xiapidan\n\n\nIn the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...", "votes": null }, { "id": "314158", "postDate": "04/14/2018 18:43:19", "content": "

Haven't read this in this competition's rules. Is this a general rule for two-stage competitions? If so, feels so unfair.

", "rawMarkdown": "Haven't read this in this competition's rules. Is this a general rule for two-stage competitions? If so, feels so unfair.", "votes": null }, { "id": "314180", "postDate": "04/14/2018 19:39:31", "content": "

Sure, it must generalize, but 670 images looks like too small training set for good results.

", "rawMarkdown": "Sure, it must generalize, but 670 images looks like too small training set for good results.", "votes": null }, { "id": "314187", "postDate": "04/14/2018 20:02:43", "content": "

Yes, that is the general rule for two stage competions: see two stage faq and the pinned post at the top...

", "rawMarkdown": "Yes, that is the general rule for two stage competions: see [two stage faq][1] and the pinned post at the top...\n\n\n [1]: https://www.kaggle.com/two-stage-faq", "votes": null }, { "id": "314190", "postDate": "04/14/2018 20:15:25", "content": "

How are we supposed to know that if it is not in competition rules?! I am just frustrated about this poorly organized competition.

", "rawMarkdown": "How are we supposed to know that if it is not in competition rules?! I am just frustrated about this poorly organized competition.", "votes": null }, { "id": "314729", "postDate": "04/16/2018 07:14:08", "content": "

Deep learining models should generalize!!!!

", "rawMarkdown": "Deep learining models should generalize!!!!", "votes": null }, { "id": "314789", "postDate": "04/16/2018 09:19:35", "content": "
\n

How are we supposed to know that if it is not in competition rules?!

\n
\n\n

It is clearly said in the timeline page of this competition, (re)read it.

", "rawMarkdown": "> How are we supposed to know that if it is not in competition rules?! \n\nIt is clearly said in the timeline page of this competition, (re)read it.", "votes": null }, { "id": "314904", "postDate": "04/16/2018 13:26:26", "content": "

@CPMP It is not clearly stated. It makes sense after reading two-stage faqs. However, I didn't understand that when I read it for the first time.

", "rawMarkdown": "CPMP It is not clearly stated. It makes sense after reading [two-stage faqs][1]. However, I didn't understand that when I read it for the first time.\n\n\n [1]: https://www.kaggle.com/two-stage-faq", "votes": null }, { "id": "314907", "postDate": "04/16/2018 13:39:06", "content": "

What is not clear in the following, taken from the timeline page?

\n\n
\n

April 11, 2018 - Stage one deadline and stage two data release. Your model must be finalized and uploaded to Kaggle by this deadline.

\n
", "rawMarkdown": "What is not clear in the following, taken from the timeline page?\n\n> April 11, 2018 - Stage one deadline and stage two data release. Your model must be finalized and uploaded to Kaggle by this deadline.", "votes": null }, { "id": "314927", "postDate": "04/16/2018 14:17:28", "content": "

Dude all I am saying that I did not understand:

\n\n
\n

We expect you may need to make some \"non scientific\" alterations, such as changes to path names, in order to create your submissions for the second stage. You are allowed to re-train your model (including the stage one data), but your code should not change. You should not be doing any hyper parameter tuning in the second stage. Parameter tuning is permitted as long as it is fully automated.

\n
\n\n

from what is written on the timeline:

\n\n
\n

your model must be finalized

\n
\n\n

So yes, organizers must have been more clear about this rule. At least a link for two-stage faqs would have been helpful.

", "rawMarkdown": "Dude all I am saying that I did not understand:\n\n> We expect you may need to make some \"non scientific\" alterations, such as changes to path names, in order to create your submissions for the second stage. You are allowed to re-train your model (including the stage one data), but your code should not change. You should not be doing any hyper parameter tuning in the second stage. Parameter tuning is permitted as long as it is fully automated.\n\nfrom what is written on the timeline:\n\n> your model must be finalized\n\nSo yes, organizers must have been more clear about this rule. At least a link for two-stage faqs would have been helpful.", "votes": null }, { "id": "314935", "postDate": "04/16/2018 14:29:06", "content": "
\n

Haven't read this in this competition's rules.

\n
\n\n

Yet you complain... ;)

", "rawMarkdown": "> Haven't read this in this competition's rules.\n\nYet you complain... ;)", "votes": null }, { "id": "314943", "postDate": "04/16/2018 14:35:56", "content": "

@CPMP I was referring to the above post

\n\n
\n

In the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...

\n
\n\n

And this small complaint is nothing, compared to all of your complaints throughout the competition ;)

", "rawMarkdown": "CPMP I was referring to the above post\n\n> In the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...\n\nAnd this small complaint is nothing, compared to all of your complaints throughout the competition ;)", "votes": null }, { "id": "314949", "postDate": "04/16/2018 14:38:50", "content": "
\n

So yes, organizers must have been more clear about this rule. At least\n a link for two-stage faqs would have been helpful.

\n
\n\n

I agree that it would be very helpful to have a link to the two-stage faq directly in the 'Timeline' section of 2-stage contests. I requested this exact same thing a month ago:

\n\n

Confusion about contest Timeline/April 11 deadline...

", "rawMarkdown": "> So yes, organizers must have been more clear about this rule. At least\n> a link for two-stage faqs would have been helpful.\n\nI agree that it would be very helpful to have a link to the two-stage faq directly in the 'Timeline' section of 2-stage contests. I requested this exact same thing a month ago:\n\n[Confusion about contest Timeline/April 11 deadline...][1]\n\n\n [1]: https://www.kaggle.com/c/data-science-bowl-2018/discussion/52385", "votes": null }, { "id": "315524", "postDate": "04/17/2018 08:03:19", "content": "

Waiting for the lottery

", "rawMarkdown": "Waiting for the lottery", "votes": null } ], "comments": [ { "id": 312939, "author_name": "agoodman", "author_url": "", "post_date": "04/12/2018 14:58:01", "content": "

I’m looking forward to bumping this thread after the phase two leaderboard is public. 🤣

", "votes": null, "replies": [ { "id": 313280, "author_name": "xhqlcwq", "author_url": "", "post_date": "04/13/2018 03:30:14", "content": "

Hello，the score is hidden on leader board？ but why we also see 0.000 on ”My Submissions“ page

", "votes": null, "replies": [] }, { "id": 313282, "author_name": "pointyointment", "author_url": "", "post_date": "04/13/2018 03:34:48", "content": "

To prevent leaderboard probing, apparently

", "votes": null, "replies": [] }, { "id": 313294, "author_name": "xhqlcwq", "author_url": "", "post_date": "04/13/2018 04:18:42", "content": "

But you don't know which of your model is better

", "votes": null, "replies": [] }, { "id": 313358, "author_name": "ildoonet", "author_url": "", "post_date": "04/13/2018 06:42:42", "content": "

@xiapidan You just use all parameters from best model of Stage1. This is intentional.

", "votes": null, "replies": [] }, { "id": 313673, "author_name": "romualdj", "author_url": "", "post_date": "04/13/2018 17:24:19", "content": "

@xiapidan

\n\n

In the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...

", "votes": null, "replies": [] }, { "id": 314158, "author_name": "panserbjorne", "author_url": "", "post_date": "04/14/2018 18:43:19", "content": "

Haven't read this in this competition's rules. Is this a general rule for two-stage competitions? If so, feels so unfair.

", "votes": null, "replies": [] }, { "id": 314187, "author_name": "romualdj", "author_url": "", "post_date": "04/14/2018 20:02:43", "content": "

Yes, that is the general rule for two stage competions: see two stage faq and the pinned post at the top...

", "votes": null, "replies": [] }, { "id": 314190, "author_name": "panserbjorne", "author_url": "", "post_date": "04/14/2018 20:15:25", "content": "

How are we supposed to know that if it is not in competition rules?! I am just frustrated about this poorly organized competition.

", "votes": null, "replies": [] }, { "id": 314789, "author_name": "cpmpml", "author_url": "", "post_date": "04/16/2018 09:19:35", "content": "
\n

How are we supposed to know that if it is not in competition rules?!

\n
\n\n

It is clearly said in the timeline page of this competition, (re)read it.

", "votes": null, "replies": [] }, { "id": 314904, "author_name": "panserbjorne", "author_url": "", "post_date": "04/16/2018 13:26:26", "content": "

@CPMP It is not clearly stated. It makes sense after reading two-stage faqs. However, I didn't understand that when I read it for the first time.

", "votes": null, "replies": [] }, { "id": 314907, "author_name": "cpmpml", "author_url": "", "post_date": "04/16/2018 13:39:06", "content": "

What is not clear in the following, taken from the timeline page?

\n\n
\n

April 11, 2018 - Stage one deadline and stage two data release. Your model must be finalized and uploaded to Kaggle by this deadline.

\n
", "votes": null, "replies": [] }, { "id": 314927, "author_name": "panserbjorne", "author_url": "", "post_date": "04/16/2018 14:17:28", "content": "

Dude all I am saying that I did not understand:

\n\n
\n

We expect you may need to make some \"non scientific\" alterations, such as changes to path names, in order to create your submissions for the second stage. You are allowed to re-train your model (including the stage one data), but your code should not change. You should not be doing any hyper parameter tuning in the second stage. Parameter tuning is permitted as long as it is fully automated.

\n
\n\n

from what is written on the timeline:

\n\n
\n

your model must be finalized

\n
\n\n

So yes, organizers must have been more clear about this rule. At least a link for two-stage faqs would have been helpful.

", "votes": null, "replies": [] }, { "id": 314935, "author_name": "cpmpml", "author_url": "", "post_date": "04/16/2018 14:29:06", "content": "
\n

Haven't read this in this competition's rules.

\n
\n\n

Yet you complain... ;)

", "votes": null, "replies": [] }, { "id": 314943, "author_name": "panserbjorne", "author_url": "", "post_date": "04/16/2018 14:35:56", "content": "

@CPMP I was referring to the above post

\n\n
\n

In the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...

\n
\n\n

And this small complaint is nothing, compared to all of your complaints throughout the competition ;)

", "votes": null, "replies": [] }, { "id": 314949, "author_name": "jliamfinnie", "author_url": "", "post_date": "04/16/2018 14:38:50", "content": "
\n

So yes, organizers must have been more clear about this rule. At least\n a link for two-stage faqs would have been helpful.

\n
\n\n

I agree that it would be very helpful to have a link to the two-stage faq directly in the 'Timeline' section of 2-stage contests. I requested this exact same thing a month ago:

\n\n

Confusion about contest Timeline/April 11 deadline...

", "votes": null, "replies": [] } ] }, { "id": 313033, "author_name": "yugusev", "author_url": "", "post_date": "04/12/2018 17:29:08", "content": "

good one

", "votes": null, "replies": [] }, { "id": 313037, "author_name": "ironbar", "author_url": "", "post_date": "04/12/2018 17:33:36", "content": "

I agree, what a waste of time and money.

\n\n

Best summary ever of a competition.

", "votes": null, "replies": [] }, { "id": 313045, "author_name": "abnerchou", "author_url": "", "post_date": "04/12/2018 17:44:08", "content": "

Can't agree more.

", "votes": null, "replies": [] }, { "id": 313049, "author_name": "dskswu", "author_url": "", "post_date": "04/12/2018 18:03:12", "content": "

OMG!!!!!!!!! I love it. I can't stop laughing.

", "votes": null, "replies": [] }, { "id": 313444, "author_name": "qinhui1999", "author_url": "", "post_date": "04/13/2018 09:10:16", "content": "

HA! HA! I like lottery. : )

", "votes": null, "replies": [] }, { "id": 314180, "author_name": "mcounter", "author_url": "", "post_date": "04/14/2018 19:39:31", "content": "

Sure, it must generalize, but 670 images looks like too small training set for good results.

", "votes": null, "replies": [] }, { "id": 314729, "author_name": "buyizhiyou", "author_url": "", "post_date": "04/16/2018 07:14:08", "content": "

Deep learining models should generalize!!!!

", "votes": null, "replies": [] }, { "id": 315524, "author_name": "yaozhenjie", "author_url": "", "post_date": "04/17/2018 08:03:19", "content": "

Waiting for the lottery

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "312938": "![enter image description here][1]\n\n\n [1]: http://i.imgur.com/j84ci6q.jpg", "312939": "I’m looking forward to bumping this thread after the phase two leaderboard is public. 🤣", "313033": "good one", "313037": "I agree, what a waste of time and money. \n\nBest summary ever of a competition.", "313045": "Can't agree more.", "313049": "OMG!!!!!!!!! I love it. I can't stop laughing.", "313280": "Hello，the score is hidden on leader board？ but why we also see 0.000 on ”My Submissions“ page", "313282": "To prevent leaderboard probing, apparently", "313294": "But you don't know which of your model is better", "313358": "xiapidan You just use all parameters from best model of Stage1. This is intentional.", "313444": "HA! HA! I like lottery. : )", "313673": "xiapidan\n\n\nIn the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...", "314158": "Haven't read this in this competition's rules. Is this a general rule for two-stage competitions? If so, feels so unfair.", "314180": "Sure, it must generalize, but 670 images looks like too small training set for good results.", "314187": "Yes, that is the general rule for two stage competions: see [two stage faq][1] and the pinned post at the top...\n\n\n [1]: https://www.kaggle.com/two-stage-faq", "314190": "How are we supposed to know that if it is not in competition rules?! I am just frustrated about this poorly organized competition.", "314729": "Deep learining models should generalize!!!!", "314789": "> How are we supposed to know that if it is not in competition rules?! \n\nIt is clearly said in the timeline page of this competition, (re)read it.", "314904": "CPMP It is not clearly stated. It makes sense after reading [two-stage faqs][1]. However, I didn't understand that when I read it for the first time.\n\n\n [1]: https://www.kaggle.com/two-stage-faq", "314907": "What is not clear in the following, taken from the timeline page?\n\n> April 11, 2018 - Stage one deadline and stage two data release. Your model must be finalized and uploaded to Kaggle by this deadline.", "314927": "Dude all I am saying that I did not understand:\n\n> We expect you may need to make some \"non scientific\" alterations, such as changes to path names, in order to create your submissions for the second stage. You are allowed to re-train your model (including the stage one data), but your code should not change. You should not be doing any hyper parameter tuning in the second stage. Parameter tuning is permitted as long as it is fully automated.\n\nfrom what is written on the timeline:\n\n> your model must be finalized\n\nSo yes, organizers must have been more clear about this rule. At least a link for two-stage faqs would have been helpful.", "314935": "> Haven't read this in this competition's rules.\n\nYet you complain... ;)", "314943": "CPMP I was referring to the above post\n\n> In the two stage competition the idea is to choose the two models for final submission at the end of stage 1. Then the code for generating them is uploaded to Kaggle and during stage 2 the only thing that can be done is to submit the outputs of exactly those two models (possibly retrained on both stage1 train and stage1 test data, but without changing any hyperparameters). So during stage 2 there is no choice left for choosing models - one just has to follow the decision from stage 1...\n\nAnd this small complaint is nothing, compared to all of your complaints throughout the competition ;)", "314949": "> So yes, organizers must have been more clear about this rule. At least\n> a link for two-stage faqs would have been helpful.\n\nI agree that it would be very helpful to have a link to the two-stage faq directly in the 'Timeline' section of 2-stage contests. I requested this exact same thing a month ago:\n\n[Confusion about contest Timeline/April 11 deadline...][1]\n\n\n [1]: https://www.kaggle.com/c/data-science-bowl-2018/discussion/52385", "315524": "Waiting for the lottery" }, "source": "meta" }