Simplifying current reverse mode functions
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 839
- Forks
- 220
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 14
Description
Description
This pull introduced a new way to write reverse mode autodiff: https://github.com/stan-dev/math/pull/2011
Because this is simple to use, it should be a good way to program all the upcoming var_value<MatrixXd> implementations.
Given it'll be used there, it would also be cool to convert all the current reverse mode stuff. This should cut a ton of code. This issue is for tracking efforts to convert the current autodiff (and make sure the new stuff is general enough and fast enough).
Current Version:
v3.3.0
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reading the new reverse-mode autodiff approach introduced in pull request 2011 and compare it with the current reverse-mode implementations. Determine which existing implementations can use the var_value pattern, then verify that the approach remains general and fast enough across the conversions.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- machine-learning
- Issue type
- Refactor
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100