GSOC

[Ben Goodrich]

Last year we briefly mentioned the possibility of Stan participating in Google's Summer of Code. If we actually intend to do that, we should come up with a list of manageable projects ASAP because rumor has it that the 2013 call for proposals is imminent.

Ben

[Bob Carpenter]

Sounds like a good idea, but it also looks like a huge
pile of bureaucracy and I have no idea what the chance of
getting in will be.

Does anyone want to put in the application? I wish we
could get a student to do THAT for us!

They'll take proposals March 18 -- March 29:

http://www.google-melange.com/gsoc/homepage/google/gsoc2013

Here's the top-level manual for organizations (I think), but it points
around to other docs, and I haven't yet been able to find anything
that looks like a checklist of what you need to do to apply as
an organization and what the application is supposed to look like:

http://en.flossmanuals.net/GSoCMentoring/notes-for-first-year-organizations/

- Bob

[Andrew Gelman]

If it seems like a good idea, Daniel Lee could put it together. He's been wanting to write things…

> --
> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>

[Ben Goodrich]

On Saturday, March 9, 2013 5:15:57 PM UTC-5, Bob Carpenter wrote:

Sounds like a good idea, but it also looks like a huge
pile of bureaucracy and I have no idea what the chance of
getting in will be.

The odds are never good. We can propose more than one project, by PyStan seems like our best bet. It doesn't require that much in the way of a stats background, lots of college students know Python, and Google uses it heavily. A project to develop a JSON parser would be similarly good.

Ben

[Andrew Gelman]

But Pystan is already happening now, no?

[Ben Goodrich]

On Saturday, March 9, 2013 4:48:18 PM UTC-5, Ben Goodrich wrote:

Last year we briefly mentioned the possibility of Stan participating in Google's Summer of Code. If we actually intend to do that, we should come up with a list of manageable projects ASAP because rumor has it that the 2013 call for proposals is imminent.

It is next year now. The GSOC mentoring organization proposals are due the first couple weeks of February if we want to do this

https://www.google-melange.com/gsoc/document/show/gsoc_program/google/gsoc2014/help_page#2._What_should_a_mentoring_organization

Probably need to focus on projects that do not actually involve much statistics.

Ben

[Marcus Brubaker]

On Sun, Jan 19, 2014 at 2:54 PM, Ben Goodrich <goodri...@gmail.com> wrote:

On Saturday, March 9, 2013 4:48:18 PM UTC-5, Ben Goodrich wrote:

Last year we briefly mentioned the possibility of Stan participating in Google's Summer of Code. If we actually intend to do that, we should come up with a list of manageable projects ASAP because rumor has it that the 2013 call for proposals is imminent.

It is next year now. The GSOC mentoring organization proposals are due the first couple weeks of February if we want to do this

https://www.google-melange.com/gsoc/document/show/gsoc_program/google/gsoc2014/help_page#2._What_should_a_mentoring_organization

Participating would be a great idea.

 

Probably need to focus on projects that do not actually involve much statistics.

I think it's good to provide a mix.  Some of these projects get senior undergrads and graduate students as participants so it's not strictly required that they be non-statistical.  Regardless, they should be concrete projects with limited scope.  Here are a couple of ideas:

- More flexible initialization by adding an "initialization" model block

- Optimized implementations of more distributions, e.g., make Geo BUGS models easier/faster in Stan

- Vectorization of multivariate distributions

- Implement L-BFGS and start unit testing optimization

- Parallelization

There are certainly more possibilities.  We should make a page on the wiki to keep track of ideas.

Cheers,

Marcus

[Bob Carpenter]

Anyone want to volunteer to do this on behalf of Stan?

I do know lots of folks at Google if someone needs references
for the project.

I'm not interested in doing this myself because it looks like more
trouble than it's worth given the value of the grants and the
difficulty of mentoring undergrads (or anyone for that matter):

"5500 USD per accepted student developer, of which 5000 USD goes to the student and
500 USD goes to the mentoring organization."

- Bob

On 1/19/14, 9:43 PM, Marcus Brubaker wrote:

[Andrew Gelman]

I'm ccing Dave Blei in case he'd be interested in mentoring a Google summer of code student on a stan project. It might be that Dave's stan interests would fit in well with a CS student project.
A

[Ben Goodrich]

On Monday, January 20, 2014 8:04:40 AM UTC-5, Bob Carpenter wrote:

Anyone want to volunteer to do this on behalf of Stan?

To do it, we need an administrator, a backup administrator, and mentors for particular projects on the to-be-written ideas list. I doubt we would get accepted, but if we did then we might get some code out of it (I think everyone agrees the money is negligible). For projects that are already on the todo list but not a high priority that could reasonably be done in a summer by an undergrad / master's student, GSOC is a good vehicle.

Ben

[Marcus Brubaker]

I'd be willing to mentor someone but, as Ben points out, we also need an administrator.  My understanding is that the work load isn't too high for this.  I don't know what our chances are, but I do know other open source research-type projects that have been successful with these kinds of things.

Cheers,

Marcus

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Bob Carpenter]

OK, do any three people want to volunteer?
Or can all of the admins and mentors be the
same person? I don't even have the patience to read
through their application materials.

In my experience, 3 months is rarely enough time
for even a good student to be a net win in terms
of the effort you have to put in versus the code
you get out. Peter and Yuanjun were both pleasant
exceptions!

- Bob

On 1/20/14, 6:22 PM, Ben Goodrich wrote:
> On Monday, January 20, 2014 8:04:40 AM UTC-5, Bob Carpenter wrote:
>
> Anyone want to volunteer to do this on behalf of Stan?
>
>
> To do it, we need an administrator, a backup administrator, and mentors for particular projects on the to-be-written
> ideas list. I doubt we would get accepted, but if we did then we might get some code out of it (I think everyone agrees
> the money is negligible). For projects that are already on the todo list but not a high priority that could reasonably
> be done in a summer by an undergrad / master's student, GSOC is a good vehicle.
>
> Ben
>
> On 1/19/14, 9:43 PM, Marcus Brubaker wrote:

> <javascript:>.
> > For more options, visit https://groups.google.com/groups/opt_out <https://groups.google.com/groups/opt_out>.

[Ben Goodrich]

On Monday, January 20, 2014 1:07:29 PM UTC-5, Bob Carpenter wrote:

OK, do any three people want to volunteer?
Or can all of the admins and mentors be the
same person?  I don't even have the patience to read
through their application materials.

The administrator and the backup administrator have to be different people in case one is hit by a bus. The mentors can overlap in all kinds of ways.

The first hurdle we have to clear is for Google to accept us as a mentoring organization. If that happens, then we get a bunch of student applications to read in March.
 

In my experience, 3 months is rarely enough time
for even a good student to be a net win in terms
of the effort you have to put in versus the code
you get out.  Peter and Yuanjun were both pleasant
exceptions!

We need more exceptions. But the 3 month thing is why I'm not too optimistic about largely stats projects (although we should list them if we decide to apply). But I could imagine a talented student being able to do

• HTML interface to Stan models compiled via pnacl or emscripten
• Java interface
  
• One or more of the alternate parser things (JSON, csv, file-backed, multiple data statements, etc.)
• initialization {} block
• Turning off more checks if NDEBUG is defined
• Implement multidimensional_thing ~ distribution(scalar, scalar, ...); 
• L-BFGS-B
  

And maybe some others.

Ben

[Bob Carpenter]

On 1/20/14, 8:58 PM, Ben Goodrich wrote:
> On Monday, January 20, 2014 1:07:29 PM UTC-5, Bob Carpenter wrote:
>
> OK, do any three people want to volunteer?
> Or can all of the admins and mentors be the
> same person? I don't even have the patience to read
> through their application materials.
>
>
> The administrator and the backup administrator have to be different people in case one is hit by a bus. The mentors can
> overlap in all kinds of ways.

We were wondering about our bus number when Michael,
Daniel and I were on a sketchy bus to the Alps.

> The first hurdle we have to clear is for Google to accept us as a mentoring organization. If that happens, then we get a
> bunch of student applications to read in March.
>
> In my experience, 3 months is rarely enough time
> for even a good student to be a net win in terms
> of the effort you have to put in versus the code
> you get out. Peter and Yuanjun were both pleasant
> exceptions!
>
>
> We need more exceptions.

:-)

> But the 3 month thing is why I'm not too optimistic about largely stats projects (although we
> should list them if we decide to apply). But I could imagine a talented student being able to do
>

> * HTML interface to Stan models compiled via pnacl or emscripten
> * Java interface
> * One or more of the alternate parser things (JSON, csv, file-backed, multiple data statements, etc.)
> * initialization {} block
> * Turning off more checks if NDEBUG is defined
> * Implement multidimensional_thing ~ distribution(scalar, scalar, ...);
> * L-BFGS-B

I agree there's much more hope with a well-defined project.

But someone has to want to do this. And I'm not interested
and nobody else has come forward to volunteer.

- Bob

[Andrew Gelman]

I still think Dave Blei would be ideal. He has so much experience supervising CS students already!
A

[Andrew Gelman]

B-b-b-but . . . Stan!!!!

[Ben Goodrich]

On Monday, January 20, 2014 3:40:55 PM UTC-5, Bob Carpenter wrote:

> But the 3 month thing is why I'm not too optimistic about largely stats projects (although we
> should list them if we decide to apply). But I could imagine a talented student being able to do
>
>   * HTML interface to Stan models compiled via pnacl or emscripten
>   * Java interface
>   * One or more of the alternate parser things (JSON, csv, file-backed, multiple data statements, etc.)
>   * initialization {} block
>   * Turning off more checks if NDEBUG is defined
>   * Implement multidimensional_thing ~ distribution(scalar, scalar, ...);
>   * L-BFGS-B

I agree there's much more hope with a well-defined project.

But someone has to want to do this.  And I'm not interested
and nobody else has come forward to volunteer.

I could be the administrator or backup administrator. The initial work is not that much anyway. The question is whether if we were accepted as a mentoring organization whether we have enough people who are committed enough to mentor a specific project?

Ben

[Michael Betancourt]

For reference, Shogun gets GSOC students all the time, http://shogun-toolbox.org/page/Events/gsoc2013.

[Bob Carpenter]

On 1/20/14, 11:20 PM, Michael Betancourt wrote:
> For reference, Shogun gets GSOC students all the time, http://shogun-toolbox.org/page/Events/gsoc2013.

I like their web pages. Surprised I've never heard
of them before given the obvious scale of their operation.

So my question remains --- does anyone want to take
responsibility for getting Stan authorized and mentoring
students and doing the admin? I'm not interested in
doing it, but don't let me stop anyone else who wants
to volunteer.

- Bob

[Ben Goodrich]

I already said I would be the administrator or backup administrator, which is not much of a commitment. But we need mentors (plural) since each GSOC student-project requires a mentor. We will probably get at most a few project slots, but we can't accept students and then leave them hanging. So we really need to decide if we have enough mentoring commitment. Basically, what would you be willing to do this summer if it only required you to do half the work (via mentoring a student) but would be unwilling to do if you had to do all the work?

I've put the outline of a few projects up at

https://github.com/stan-dev/stan/wiki/GSOC2014Ideas

Feel free to expand upon them or add new ones or list yourself as a mentor of an idea in the idea header.

Ben

[Marcus Brubaker]

I've added a project idea and put my name down to mentor those projects that I'd be interested in/suitable for.  I should also note that I'd be willing to help with mentoring others, but, e.g., the initialization block project touches on the parser and I don't have experience with that.