{ "id": 11027, "title": "Regarding the \"Open Source\" clause", "url": "/competitions/avazu-ctr-prediction/discussion/11027", "author_name": "", "post_date": "2014-11-23T19:22:40.430Z", "votes": null, "comment_count": 8, "views": 2534, "content": "

The rules state the following:

\n

A Submission will be ineligible to win a prize if it was developed using code containing or depending on software licensed under an open source license:

\n

* other than an Open Source Initiative-approved license (see <http://opensource.org/>); or
* an open source license that prohibits commercial use.

\n

* using Matlab or any closed source software (except BLAS) will not be qualified.

\n

My question: does using GPL and/or LGPL licensed software result in ineligibility? These licenses are obviously recognized by the OSI. If so, then a quite strange definition of "open source" is being put forward in this competition.

\n

In fact, this would rule out the vast majority of machine learning libraries. It would rule out the R language entirely, since it's GPL licensed (source). Since some of the code examples given in this competition are in R, I assume using it is fine?  (and consequently, using any GPL/LGPL software would be permitted)

", "messages": [ { "id": "58740", "postDate": "11/23/2014 19:22:40", "content": "

The rules state the following:

\n

A Submission will be ineligible to win a prize if it was developed using code containing or depending on software licensed under an open source license:

\n

* other than an Open Source Initiative-approved license (see <http://opensource.org/>); or
* an open source license that prohibits commercial use.

\n

* using Matlab or any closed source software (except BLAS) will not be qualified.

\n

My question: does using GPL and/or LGPL licensed software result in ineligibility? These licenses are obviously recognized by the OSI. If so, then a quite strange definition of "open source" is being put forward in this competition.

\n

In fact, this would rule out the vast majority of machine learning libraries. It would rule out the R language entirely, since it's GPL licensed (source). Since some of the code examples given in this competition are in R, I assume using it is fine?  (and consequently, using any GPL/LGPL software would be permitted)

", "rawMarkdown": "", "votes": null }, { "id": "58741", "postDate": "11/23/2014 19:52:29", "content": "

And . . . if I may add an additional question.

\n

What about libraries that speed up code? For example, what if I had Continuum's MKL Optimization installed? I'm assuming that's okay, since it doesn't modify any code.

\n

https://store.continuum.io/cshop/mkl-optimizations/

\n

To take it a step farther, what about NumbaPro? The only thing you're doing is adding a decorator to functions (e.g., @jit). The code still works without the decorator, just a lot slower.

\n

It's probably a grey area since you could obviously strip out the decorator before submitting a solution; thus, there is no way to enforce it.

", "rawMarkdown": "", "votes": null }, { "id": "58742", "postDate": "11/23/2014 20:36:42", "content": "

[quote=Marc Claesen;58740]

\n

The rules state the following:

\n

A Submission will be ineligible to win a prize if it was developed using code containing or depending on software licensed under an open source license:

\n

* other than an Open Source Initiative-approved license (see <http://opensource.org/>); or
* an open source license that prohibits commercial use.

\n

* using Matlab or any closed source software (except BLAS) will not be qualified.

\n

My question: does using GPL and/or LGPL licensed software result in ineligibility? These licenses are obviously recognized by the OSI. If so, then a quite strange definition of "open source" is being put forward in this competition.

\n

In fact, this would rule out the vast majority of machine learning libraries. It would rule out the R language entirely, since it's GPL licensed (source). Since some of the code examples given in this competition are in R, I assume using it is fine?  (and consequently, using any GPL/LGPL software would be permitted)

\n

[/quote]

\n

Perhaps I'm misreading your post, but I don't see the problem. Both GPL and LGPL satisfy the conditions:

\n

1. Both are recognized by OSI.

\n

2. Both allow commerical use.

", "rawMarkdown": "", "votes": null }, { "id": "58743", "postDate": "11/23/2014 20:38:37", "content": "

[quote=laserwolf;58742]

\n

2. Both allow commerical use.

\n

[/quote]

\n

That depends on how you define commercial use. GPL can be problematic if the organiser's goal is to incorporate the solution into a proprietary code base, which appears to be the case.

", "rawMarkdown": "", "votes": null }, { "id": "58753", "postDate": "11/23/2014 23:47:54", "content": "

[quote=Marc Claesen;58743]

\n

[quote=laserwolf;58742]

\n

2. Both allow commerical use.

\n

[/quote]

\n

That depends on how you define commercial use. GPL can be problematic if the organiser's goal is to incorporate the solution into a proprietary code base, which appears to be the case.

\n

[/quote]

\n

The GPL might not mesh perfectly with your business plan, but that is different than prohibiting commercial use. 

\n

Something like CC-NC or NPOSL would run afoul of the rules however.

", "rawMarkdown": "", "votes": null }, { "id": "58767", "postDate": "11/24/2014 09:12:36", "content": "

[quote=emolson;58753]

\n

Something like CC-NC or NPOSL would run afoul of the rules however.

\n

[/quote]

\n

When I read the rules I assumed that those were indeed the licenses that are being dismissed, but I prefer to make 100% sure before starting.

", "rawMarkdown": "", "votes": null }, { "id": "58770", "postDate": "11/24/2014 13:08:11", "content": "

[quote=inversion;58741]

\n

And . . . if I may add an additional question.

\n

What about libraries that speed up code? For example, what if I had Continuum's MKL Optimization installed? I'm assuming that's okay, since it doesn't modify any code.

\n

https://store.continuum.io/cshop/mkl-optimizations/

\n

To take it a step farther, what about NumbaPro? The only thing you're doing is adding a decorator to functions (e.g., @jit). The code still works without the decorator, just a lot slower.

\n

It's probably a grey area since you could obviously strip out the decorator before submitting a solution; thus, there is no way to enforce it.

\n

[/quote]

\n

That's an interesting set of upgrades! I'm personally using PyPy for the speed up, even though I have to do without NumPy. Does Anaconda+MKL offer the same level of speed? Would be greaty if you could spare some advice on that.

", "rawMarkdown": "", "votes": null }, { "id": "58771", "postDate": "11/24/2014 13:33:31", "content": "

[quote=Ivan Lobov;58770]

\n

That's an interesting set of upgrades! I'm personally using PyPy for the speed up, even though I have to do without NumPy. Does Anaconda+MKL offer the same level of speed? Would be greaty if you could spare some advice on that.

\n

[/quote]

\n

Well, actually, I'm running on an AMD machine at the moment, so I wouldn't get the speedup from MKL. I may upgrade sometime in the future, though.

\n

When I was working on the Tradeshift competition, Numba gave me a 17% improvement over PyPy. (Note, just Numba, not NumbaPro.)

The main advantage of Numba over PyPy, though, was that since I'm running on Windows, PyPy only has access to 2 Gb or RAM. Going to Anaconda/Numba gave me access to my full system RAM (32 Gb).

\n

I'm interested in trying NumbaPro for some contests, which would allow me to more easily run multi-core and GPU computations. Again, the reason I'm asking is it would be trivial to revert code back from NumbaPro to regular python before submitting a winning entry (although, obviously, it would run much slower). But, I'm a by-the-rules kind of guy, so if NumbaPro falls out of the competition rules, then I'm not going to use it.

", "rawMarkdown": "", "votes": null }, { "id": "58774", "postDate": "11/24/2014 14:11:13", "content": "

[quote=inversion;58771]

\n

I'm running on Windows

\n

[/quote]

\n

Wait a minute! I don't think Windows is open-source.

\n

Just kidding :)

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 58741, "author_name": "inversion", "author_url": "", "post_date": "11/23/2014 19:52:29", "content": "

And . . . if I may add an additional question.

\n

What about libraries that speed up code? For example, what if I had Continuum's MKL Optimization installed? I'm assuming that's okay, since it doesn't modify any code.

\n

https://store.continuum.io/cshop/mkl-optimizations/

\n

To take it a step farther, what about NumbaPro? The only thing you're doing is adding a decorator to functions (e.g., @jit). The code still works without the decorator, just a lot slower.

\n

It's probably a grey area since you could obviously strip out the decorator before submitting a solution; thus, there is no way to enforce it.

", "votes": null, "replies": [] }, { "id": 58742, "author_name": "vvarkey", "author_url": "", "post_date": "11/23/2014 20:36:42", "content": "

[quote=Marc Claesen;58740]

\n

The rules state the following:

\n

A Submission will be ineligible to win a prize if it was developed using code containing or depending on software licensed under an open source license:

\n

* other than an Open Source Initiative-approved license (see <http://opensource.org/>); or
* an open source license that prohibits commercial use.

\n

* using Matlab or any closed source software (except BLAS) will not be qualified.

\n

My question: does using GPL and/or LGPL licensed software result in ineligibility? These licenses are obviously recognized by the OSI. If so, then a quite strange definition of "open source" is being put forward in this competition.

\n

In fact, this would rule out the vast majority of machine learning libraries. It would rule out the R language entirely, since it's GPL licensed (source). Since some of the code examples given in this competition are in R, I assume using it is fine?  (and consequently, using any GPL/LGPL software would be permitted)

\n

[/quote]

\n

Perhaps I'm misreading your post, but I don't see the problem. Both GPL and LGPL satisfy the conditions:

\n

1. Both are recognized by OSI.

\n

2. Both allow commerical use.

", "votes": null, "replies": [] }, { "id": 58743, "author_name": "claesenm", "author_url": "", "post_date": "11/23/2014 20:38:37", "content": "

[quote=laserwolf;58742]

\n

2. Both allow commerical use.

\n

[/quote]

\n

That depends on how you define commercial use. GPL can be problematic if the organiser's goal is to incorporate the solution into a proprietary code base, which appears to be the case.

", "votes": null, "replies": [] }, { "id": 58753, "author_name": "emolson", "author_url": "", "post_date": "11/23/2014 23:47:54", "content": "

[quote=Marc Claesen;58743]

\n

[quote=laserwolf;58742]

\n

2. Both allow commerical use.

\n

[/quote]

\n

That depends on how you define commercial use. GPL can be problematic if the organiser's goal is to incorporate the solution into a proprietary code base, which appears to be the case.

\n

[/quote]

\n

The GPL might not mesh perfectly with your business plan, but that is different than prohibiting commercial use. 

\n

Something like CC-NC or NPOSL would run afoul of the rules however.

", "votes": null, "replies": [] }, { "id": 58767, "author_name": "claesenm", "author_url": "", "post_date": "11/24/2014 09:12:36", "content": "

[quote=emolson;58753]

\n

Something like CC-NC or NPOSL would run afoul of the rules however.

\n

[/quote]

\n

When I read the rules I assumed that those were indeed the licenses that are being dismissed, but I prefer to make 100% sure before starting.

", "votes": null, "replies": [] }, { "id": 58770, "author_name": "ivanlobov", "author_url": "", "post_date": "11/24/2014 13:08:11", "content": "

[quote=inversion;58741]

\n

And . . . if I may add an additional question.

\n

What about libraries that speed up code? For example, what if I had Continuum's MKL Optimization installed? I'm assuming that's okay, since it doesn't modify any code.

\n

https://store.continuum.io/cshop/mkl-optimizations/

\n

To take it a step farther, what about NumbaPro? The only thing you're doing is adding a decorator to functions (e.g., @jit). The code still works without the decorator, just a lot slower.

\n

It's probably a grey area since you could obviously strip out the decorator before submitting a solution; thus, there is no way to enforce it.

\n

[/quote]

\n

That's an interesting set of upgrades! I'm personally using PyPy for the speed up, even though I have to do without NumPy. Does Anaconda+MKL offer the same level of speed? Would be greaty if you could spare some advice on that.

", "votes": null, "replies": [] }, { "id": 58771, "author_name": "inversion", "author_url": "", "post_date": "11/24/2014 13:33:31", "content": "

[quote=Ivan Lobov;58770]

\n

That's an interesting set of upgrades! I'm personally using PyPy for the speed up, even though I have to do without NumPy. Does Anaconda+MKL offer the same level of speed? Would be greaty if you could spare some advice on that.

\n

[/quote]

\n

Well, actually, I'm running on an AMD machine at the moment, so I wouldn't get the speedup from MKL. I may upgrade sometime in the future, though.

\n

When I was working on the Tradeshift competition, Numba gave me a 17% improvement over PyPy. (Note, just Numba, not NumbaPro.)

The main advantage of Numba over PyPy, though, was that since I'm running on Windows, PyPy only has access to 2 Gb or RAM. Going to Anaconda/Numba gave me access to my full system RAM (32 Gb).

\n

I'm interested in trying NumbaPro for some contests, which would allow me to more easily run multi-core and GPU computations. Again, the reason I'm asking is it would be trivial to revert code back from NumbaPro to regular python before submitting a winning entry (although, obviously, it would run much slower). But, I'm a by-the-rules kind of guy, so if NumbaPro falls out of the competition rules, then I'm not going to use it.

", "votes": null, "replies": [] }, { "id": 58774, "author_name": "vvarkey", "author_url": "", "post_date": "11/24/2014 14:11:13", "content": "

[quote=inversion;58771]

\n

I'm running on Windows

\n

[/quote]

\n

Wait a minute! I don't think Windows is open-source.

\n

Just kidding :)

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "58740": "", "58741": "", "58742": "", "58743": "", "58753": "", "58767": "", "58770": "", "58771": "", "58774": "" }, "source": "meta" }