The rules state:
\n\nCan \"external data sources\" include compiled code?
\n\nFor instance, a notebook that contains a feature extraction program (say a c++ compiled executable) that is later called from a python script that does analytics using the features would be forbidden? Is that considered a \"custom package\"? What if that code was open sourced?
", "messages": [ { "id": "693625", "postDate": "12/12/2019 15:34:01", "content": "The rules state:
\n\nCan \"external data sources\" include compiled code?
\n\nFor instance, a notebook that contains a feature extraction program (say a c++ compiled executable) that is later called from a python script that does analytics using the features would be forbidden? Is that considered a \"custom package\"? What if that code was open sourced?
", "rawMarkdown": "The rules state: \n\n- *No using other Kaggle notebooks or utility scripts as inputs to your submission notebook. Instead, load external models or datasets as an external data source.*\n- *No custom packages enabled in your submission notebook.*\n\nCan \"external data sources\" include compiled code?\n\nFor instance, a notebook that contains a feature extraction program (say a c++ compiled executable) that is later called from a python script that does analytics using the features would be forbidden? Is that considered a \"custom package\"? What if that code was open sourced?", "votes": null }, { "id": "693650", "postDate": "12/12/2019 16:13:49", "content": "'load external models or datasets as an external data source'\n1. your model weights\n2. datasets
\n\n@batangas I think you'd better to @ competition organizer, they will give you official answer.
", "rawMarkdown": "'load external models or datasets as an external data source'\n1. your model weights\n2. datasets\n\n@batangas I think you'd better to @ competition organizer, they will give you official answer.", "votes": null }, { "id": "693854", "postDate": "12/12/2019 21:26:09", "content": "Hi @juliaelliott,
\n\nAs Diego suggests, could you give the official position on this? From what I have seen browsing randomly, some clear fakes (like intermittent ones) could be easily detected if some low level coding was allowed. Maybe \"coding competition\" should not be just another name for \"Python coding competition\" (or, less likely, R).
", "rawMarkdown": "Hi @juliaelliott,\n\nAs Diego suggests, could you give the official position on this? From what I have seen browsing randomly, some clear fakes (like intermittent ones) could be easily detected if some low level coding was allowed. Maybe \"coding competition\" should not be just another name for \"Python coding competition\" (or, less likely, R).", "votes": null }, { "id": "693869", "postDate": "12/12/2019 21:47:36", "content": "No “custom packages” specifically refers to enabling any custom packages under the “Settings” menu of the Kaggle Notebook for your submission. Anything you do offline as part of what you’re loading into your notebook is permitted, as long as you’re not using a utility script or other Kaggle notebook as the input to your submission and you’re not enabling a custom package on the settings. I hope this is clearer.
", "rawMarkdown": "No “custom packages” specifically refers to enabling any custom packages under the “Settings” menu of the Kaggle Notebook for your submission. Anything you do offline as part of what you’re loading into your notebook is permitted, as long as you’re not using a utility script or other Kaggle notebook as the input to your submission and you’re not enabling a custom package on the settings. I hope this is clearer.", "votes": null }, { "id": "694157", "postDate": "12/13/2019 08:52:28", "content": "Ok. Thank you.
", "rawMarkdown": "Ok. Thank you.", "votes": null }, { "id": "694197", "postDate": "12/13/2019 09:38:33", "content": "Just a note - you can't use feature extraction offline, because you have no full data offline.
", "rawMarkdown": "Just a note - you can't use feature extraction offline, because you have no full data offline.", "votes": null }, { "id": "694411", "postDate": "12/13/2019 14:46:07", "content": "Hi Jacek, I don't see the difference with any other submission, except for the extra work of making everything work inside the kaggle/python docker image which comes \"out of the box\" if you use the \"standard\" tools. But that, of course, only applies if I have something worth submitting. For the moment, I can just work offline on my computer with my favorite tools, just like anyone else. If the solution is worth submitting, there is some extra work to pack the code as an \"external source\" and calling it from the python notebook. I have a docker image to try it. If it works on the docker image, it should work when submitted. It will extract the features from the videos supplied by the platform, of course.
", "rawMarkdown": "Hi Jacek, I don't see the difference with any other submission, except for the extra work of making everything work inside the kaggle/python docker image which comes \"out of the box\" if you use the \"standard\" tools. But that, of course, only applies if I have something worth submitting. For the moment, I can just work offline on my computer with my favorite tools, just like anyone else. If the solution is worth submitting, there is some extra work to pack the code as an \"external source\" and calling it from the python notebook. I have a docker image to try it. If it works on the docker image, it should work when submitted. It will extract the features from the videos supplied by the platform, of course.", "votes": null }, { "id": "745761", "postDate": "02/14/2020 06:59:34", "content": "Yea the code competitions are annoying if you don't use python or r and it seems like more and more competitions are code competitions. I have stopped using python and I do everything in C++ now. I made this switch because when it comes to large data sets like this the little extra time to write the C++ code vs python ends up saving you time in the long run because C++ is just so much faster, also the memory use is way way less. In addition the the time to write c++ code decreases with every competition you enter because you are creating a library of reusable code.
", "rawMarkdown": "Yea the code competitions are annoying if you don't use python or r and it seems like more and more competitions are code competitions. I have stopped using python and I do everything in C++ now. I made this switch because when it comes to large data sets like this the little extra time to write the C++ code vs python ends up saving you time in the long run because C++ is just so much faster, also the memory use is way way less. In addition the the time to write c++ code decreases with every competition you enter because you are creating a library of reusable code.", "votes": null } ], "comments": [ { "id": 693650, "author_name": "diegojohnson", "author_url": "", "post_date": "12/12/2019 16:13:49", "content": "'load external models or datasets as an external data source'\n1. your model weights\n2. datasets
\n\n@batangas I think you'd better to @ competition organizer, they will give you official answer.
", "votes": null, "replies": [] }, { "id": 693854, "author_name": "batangas", "author_url": "", "post_date": "12/12/2019 21:26:09", "content": "Hi @juliaelliott,
\n\nAs Diego suggests, could you give the official position on this? From what I have seen browsing randomly, some clear fakes (like intermittent ones) could be easily detected if some low level coding was allowed. Maybe \"coding competition\" should not be just another name for \"Python coding competition\" (or, less likely, R).
", "votes": null, "replies": [] }, { "id": 693869, "author_name": "juliaelliott", "author_url": "", "post_date": "12/12/2019 21:47:36", "content": "No “custom packages” specifically refers to enabling any custom packages under the “Settings” menu of the Kaggle Notebook for your submission. Anything you do offline as part of what you’re loading into your notebook is permitted, as long as you’re not using a utility script or other Kaggle notebook as the input to your submission and you’re not enabling a custom package on the settings. I hope this is clearer.
", "votes": null, "replies": [] }, { "id": 694157, "author_name": "batangas", "author_url": "", "post_date": "12/13/2019 08:52:28", "content": "Ok. Thank you.
", "votes": null, "replies": [] }, { "id": 694197, "author_name": "jacekpoplawski", "author_url": "", "post_date": "12/13/2019 09:38:33", "content": "Just a note - you can't use feature extraction offline, because you have no full data offline.
", "votes": null, "replies": [] }, { "id": 694411, "author_name": "batangas", "author_url": "", "post_date": "12/13/2019 14:46:07", "content": "Hi Jacek, I don't see the difference with any other submission, except for the extra work of making everything work inside the kaggle/python docker image which comes \"out of the box\" if you use the \"standard\" tools. But that, of course, only applies if I have something worth submitting. For the moment, I can just work offline on my computer with my favorite tools, just like anyone else. If the solution is worth submitting, there is some extra work to pack the code as an \"external source\" and calling it from the python notebook. I have a docker image to try it. If it works on the docker image, it should work when submitted. It will extract the features from the videos supplied by the platform, of course.
", "votes": null, "replies": [] }, { "id": 745761, "author_name": "rilesdg3", "author_url": "", "post_date": "02/14/2020 06:59:34", "content": "Yea the code competitions are annoying if you don't use python or r and it seems like more and more competitions are code competitions. I have stopped using python and I do everything in C++ now. I made this switch because when it comes to large data sets like this the little extra time to write the C++ code vs python ends up saving you time in the long run because C++ is just so much faster, also the memory use is way way less. In addition the the time to write c++ code decreases with every competition you enter because you are creating a library of reusable code.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "693625": "The rules state: \n\n- *No using other Kaggle notebooks or utility scripts as inputs to your submission notebook. Instead, load external models or datasets as an external data source.*\n- *No custom packages enabled in your submission notebook.*\n\nCan \"external data sources\" include compiled code?\n\nFor instance, a notebook that contains a feature extraction program (say a c++ compiled executable) that is later called from a python script that does analytics using the features would be forbidden? Is that considered a \"custom package\"? What if that code was open sourced?", "693650": "'load external models or datasets as an external data source'\n1. your model weights\n2. datasets\n\n@batangas I think you'd better to @ competition organizer, they will give you official answer.", "693854": "Hi @juliaelliott,\n\nAs Diego suggests, could you give the official position on this? From what I have seen browsing randomly, some clear fakes (like intermittent ones) could be easily detected if some low level coding was allowed. Maybe \"coding competition\" should not be just another name for \"Python coding competition\" (or, less likely, R).", "693869": "No “custom packages” specifically refers to enabling any custom packages under the “Settings” menu of the Kaggle Notebook for your submission. Anything you do offline as part of what you’re loading into your notebook is permitted, as long as you’re not using a utility script or other Kaggle notebook as the input to your submission and you’re not enabling a custom package on the settings. I hope this is clearer.", "694157": "Ok. Thank you.", "694197": "Just a note - you can't use feature extraction offline, because you have no full data offline.", "694411": "Hi Jacek, I don't see the difference with any other submission, except for the extra work of making everything work inside the kaggle/python docker image which comes \"out of the box\" if you use the \"standard\" tools. But that, of course, only applies if I have something worth submitting. For the moment, I can just work offline on my computer with my favorite tools, just like anyone else. If the solution is worth submitting, there is some extra work to pack the code as an \"external source\" and calling it from the python notebook. I have a docker image to try it. If it works on the docker image, it should work when submitted. It will extract the features from the videos supplied by the platform, of course.", "745761": "Yea the code competitions are annoying if you don't use python or r and it seems like more and more competitions are code competitions. I have stopped using python and I do everything in C++ now. I made this switch because when it comes to large data sets like this the little extra time to write the C++ code vs python ends up saving you time in the long run because C++ is just so much faster, also the memory use is way way less. In addition the the time to write c++ code decreases with every competition you enter because you are creating a library of reusable code." }, "source": "meta" }