Re: Digest for google-summer-of-code-mentors-list@googlegroups.com - 25 Messages in 2 Topics

[Sean Owen <sr...@gmail.com>]

+mahout-dev I think at this point

I could be misremembering (there's that word again Grant) but are we
not supposed to sign on to mentor more than 1 person without having
talked it over on code-awards? Seems like a lot of grumbling about
gaming the system and such from past years, which seems legit.

I was saying to these guys earlier that I tried to mentor 2 proposals
in the first GSoC* and it was a lot more involved that I imagined (in
my case, hard to get work out the students :( ), so, probably a good
idea to stick at 1 anyway.

One per person is a good idea, but, indeed that means we'd be
accepting at best about 1/3 of the proposals on the table. It would be
great for one more person to sign on as a mentor if possible.



On Sat, Apr 10, 2010 at 10:15 PM, Ted Dunning <te...@gmail.com> wrote:
>
> Hard to predict, but it will definitely be less if we don't mark enough
> proposals as mentorable now.
>
> On Sat, Apr 10, 2010 at 2:07 PM, Robin Anil <ro...@gmail.com> wrote:
>>
>> I am just marking Zhao. So Sisir is open for selection. Any idea on how
>> many slots Mahout will get?
>

Re: Digest for google-summer-of-code-mentors-list@googlegroups.com - 25 Messages in 2 Topics

[Benson Margulies <bi...@gmail.com>]

OK, sign me up.

On Mon, Apr 12, 2010 at 11:05 AM, Grant Ingersoll <gs...@apache.org> wrote:
>
> On Apr 12, 2010, at 10:01 AM, Benson Margulies wrote:
>
>> Robin asked me to pay attention to this mentoring question. I have
>> assumed that, to be an effective mentor here, I'd have to be capable
>> of at least keeping up with the mentee on the math behind whatever
>> algorithm is in play. Thus, absent a proposal to work on maven
>> configurations or collections, I'd classified myself as unqualified.
>> Grant's message seems to suggest that I'm wrong, but I still worry
>> that the unsuspecting victim will dig themselves into an algorithmic
>> hole from which I cannot extract them.
>
> As I told David Hall last year, "I guarantee you that you know more about the math than I do." (and I have a Math degree!) since he is living it every day in his studies and work.  In fact, David was as much a mentor to me on LDA as I was a mentor on Mahout.  I view my job as a mentor is to help the student learn the ins and outs of open source at the ASF and in Mahout.  That I feel I know quite well.   I also know that implementations don't have to be perfect and that they will evolve and get better over time.
>
> I also know that decisions/discussions about implementations are to be done on the list even when I do know the answer (anyone who has ever emailed me w/ a private question knows this about me as well) so that everyone benefits from it.  More times than not, a better solution is arrived at anyway.  So, for me, off-list mentoring, comes down to things like checking on progress, writing recommendations, filling out the evaluation, etc. and all of those things are usually minimal (< 5 hours per week)
>
> So, in other words, it is not your job to extract them from the hole.  That's the community job.  Besides, just b/c the algorithm is in a hole doesn't mean the student has failed.
>
> -Grant

Re: Digest for google-summer-of-code-mentors-list@googlegroups.com - 25 Messages in 2 Topics

[Grant Ingersoll <gs...@apache.org>]

On Apr 12, 2010, at 10:01 AM, Benson Margulies wrote:

> Robin asked me to pay attention to this mentoring question. I have
> assumed that, to be an effective mentor here, I'd have to be capable
> of at least keeping up with the mentee on the math behind whatever
> algorithm is in play. Thus, absent a proposal to work on maven
> configurations or collections, I'd classified myself as unqualified.
> Grant's message seems to suggest that I'm wrong, but I still worry
> that the unsuspecting victim will dig themselves into an algorithmic
> hole from which I cannot extract them.

As I told David Hall last year, "I guarantee you that you know more about the math than I do." (and I have a Math degree!) since he is living it every day in his studies and work.  In fact, David was as much a mentor to me on LDA as I was a mentor on Mahout.  I view my job as a mentor is to help the student learn the ins and outs of open source at the ASF and in Mahout.  That I feel I know quite well.   I also know that implementations don't have to be perfect and that they will evolve and get better over time.

I also know that decisions/discussions about implementations are to be done on the list even when I do know the answer (anyone who has ever emailed me w/ a private question knows this about me as well) so that everyone benefits from it.  More times than not, a better solution is arrived at anyway.  So, for me, off-list mentoring, comes down to things like checking on progress, writing recommendations, filling out the evaluation, etc. and all of those things are usually minimal (< 5 hours per week)

So, in other words, it is not your job to extract them from the hole.  That's the community job.  Besides, just b/c the algorithm is in a hole doesn't mean the student has failed.

-Grant

Re: Digest for google-summer-of-code-mentors-list@googlegroups.com - 25 Messages in 2 Topics

[Benson Margulies <bi...@gmail.com>]

Robin asked me to pay attention to this mentoring question. I have
assumed that, to be an effective mentor here, I'd have to be capable
of at least keeping up with the mentee on the math behind whatever
algorithm is in play. Thus, absent a proposal to work on maven
configurations or collections, I'd classified myself as unqualified.
Grant's message seems to suggest that I'm wrong, but I still worry
that the unsuspecting victim will dig themselves into an algorithmic
hole from which I cannot extract them.

Re: Digest for google-summer-of-code-mentors-list@googlegroups.com - 25 Messages in 2 Topics

[Grant Ingersoll <gs...@apache.org>]

On Apr 10, 2010, at 5:26 PM, Sean Owen wrote:

> +mahout-dev I think at this point
> 
> I could be misremembering (there's that word again Grant) but are we
> not supposed to sign on to mentor more than 1 person without having
> talked it over on code-awards? Seems like a lot of grumbling about
> gaming the system and such from past years, which seems legit.

Yes, you need to talk it over on code-awards.  Given your work status, perhaps you could do 2, but that is your call.

It is sometimes the case that students try to sign up for multiple projects at a time and get paid for both.  The stipend Google pays can be quite a lot of money in many countries, so it encourages gaming the system, unfortunately.

> 
> I was saying to these guys earlier that I tried to mentor 2 proposals
> in the first GSoC* and it was a lot more involved that I imagined (in
> my case, hard to get work out the students :( ), so, probably a good
> idea to stick at 1 anyway.

I'd stick to 1, unless you have copious free time.  That being said, the only "mentoring" I ever do privately is stuff to do with personal matters (i.e. progress, vacation, sickness, etc. stuff of that matter).  All else discussion about the project and how to implement stuff should happen on the list.  Students need to know this up front and expect my response to private questions on implementation details to say "please ask on the list" (I'm sure David and Deneche can vouch for this).  As they say in ASF land: If it didn't happen on a list, it didn't happen.

> 
> One per person is a good idea, but, indeed that means we'd be
> accepting at best about 1/3 of the proposals on the table. It would be
> great for one more person to sign on as a mentor if possible.
> 

+1, but probably too late at this point.