-
-
Notifications
You must be signed in to change notification settings - Fork 199
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
DCEGM rework #1100
DCEGM rework #1100
Conversation
The current version has the rework of the main method. I still need to
|
@alanlujan91, @llorracc tells me that you might be interested in this rework. I think it is mostly done and just needs to be documented, I can do that relatively soon. We can then chat about G2EGM, which Chris said you might want to look into. I had a go at debugging the current version long, long ago. That partial and old work is here: #782. I think the nature of the bugs that still remain might have to do with the same issues I see with the current DCEGM work. If I recall correctly, there were various methods that mixed numerical tasks with model-specific details. Separating those two things might be a worthwhile endeavor if one wants to get it to work. It's what I'd do at least. Nevertheless, the numerical procedures involved in G2EGM seem much more complex. |
This is great! Let me know when you would like me to review it. |
Codecov Report
@@ Coverage Diff @@
## master #1100 +/- ##
==========================================
+ Coverage 73.58% 73.66% +0.08%
==========================================
Files 69 69
Lines 10592 10560 -32
==========================================
- Hits 7794 7779 -15
+ Misses 2798 2781 -17
Continue to review full report at Codecov.
|
This is ready for review. The updated DemARK is in econ-ark/DemARK#181 |
@alanlujan91, you are the natural reviewer, given your earlier conversations. I'm not sure whether you can merge it if you approve of it, but if not just post a comment that you think it is ready to merge and email me and I'll merge. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This PR is in great shape!
Given that we are using Econ-Forge's interpolation
I have a couple of questions about whether we could numba-fy all of the tools here. It could allow for some tasks to be done in parallel simply changing a few range
s to prange
s which could make some processes faster, although there could be drawbacks when applied to small loops.
This PR will contain work addressing #1062.
The idea of my proposed rework will be to provide general-purpose tools to speed-up the work of someone implementing DCEGM. The current tools seem to intend to apply DCEGM as presented in the original paper.
Thus, for instance, I want to replace the current tool which
With a tool that
It is my view that optimizations and specific uses of the general purpose tools are part of the specific models and should be done in their files, when they are implemented.
The version in the PR also uses
numba.jit
and Pablo's interpolation library to speed up things.