j***@gmail.com
2018-12-02 04:04:58 UTC
Google announced a while ago that there will be a GSOC in 2019.
My current view is that I want to skip it again as a mentor, unless
there is a strong overlap with what I would want to work on myself.
Last summer Chad was the only active mentor, and I was pretty glad
that I could take the time to work on old PRs.
GAM is about finished as a first version, but it took me two months at
limited time availability to get there. The PRs that I worked on in
the last year, including GAM, still cover only part of what we or I
started in 2015 and 2016. Some other large topics like robust and
treatment effects are still stalled and untouched and are at least one
month of work each. There are several other topics that might be less
than a month of work and can be based on 80% finished PRs.
(I thought GAM is 90% finished until I realized that I needed to
verify and fix or enhance each individual part of it.)
We got and get a lot of good code out of GSOC, also GAM is still based
on the work of a full GSOC project, but it's time consuming and most
projects need a month or two of post-GSOC work to bring them into a
good state for merging. This is largely because our GSOC projects were
most of the time big topics that cannot really be finished within one
GSOC. Most of our students provided very good work and we would not
have those models without them. But given the backlog of PRs, I think
it is better to focus on finishing up those instead of starting with
new projects.
Josef
I'm sorting through articles that I printed in the last years, and
there are many interesting topics, but those are all over the place.
https://github.com/josef-pkt/misc/wiki/0-Topics---Sorting-used
My current view is that I want to skip it again as a mentor, unless
there is a strong overlap with what I would want to work on myself.
Last summer Chad was the only active mentor, and I was pretty glad
that I could take the time to work on old PRs.
GAM is about finished as a first version, but it took me two months at
limited time availability to get there. The PRs that I worked on in
the last year, including GAM, still cover only part of what we or I
started in 2015 and 2016. Some other large topics like robust and
treatment effects are still stalled and untouched and are at least one
month of work each. There are several other topics that might be less
than a month of work and can be based on 80% finished PRs.
(I thought GAM is 90% finished until I realized that I needed to
verify and fix or enhance each individual part of it.)
We got and get a lot of good code out of GSOC, also GAM is still based
on the work of a full GSOC project, but it's time consuming and most
projects need a month or two of post-GSOC work to bring them into a
good state for merging. This is largely because our GSOC projects were
most of the time big topics that cannot really be finished within one
GSOC. Most of our students provided very good work and we would not
have those models without them. But given the backlog of PRs, I think
it is better to focus on finishing up those instead of starting with
new projects.
Josef
I'm sorting through articles that I printed in the last years, and
there are many interesting topics, but those are all over the place.
https://github.com/josef-pkt/misc/wiki/0-Topics---Sorting-used