stan-dev / stan-dev/math

ldlt rev mode adjoints

Open
#2,722 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
839
Forks
220
Avg merge
2d 4h
Merged PRs (30d)
14

Description

The ldlt code can basically piggy back off of the the reverse mode cholesky adjoints. In the cholesky reverse mode code it already handles the diagonal and off diagonal elements separately. Instead of combining we need to explicitly separate out the diagonal and off diagonal returns.

The difficulty I'm having in seeing how this would work today without tuples. The ldlt class has methods to return the matrixL() and diagonalD().

I think we can get the rev mode adjoints in today. And then in the near future when we have tuples expose this function into Stan.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the reverse-mode Cholesky adjoint code and the ldlt class, focusing on matrixL() and diagonalD(). Work out how diagonal and off-diagonal adjoints should be returned without tuples; done means reverse-mode ldlt adjoints are available, with future Stan exposure left for tuple support.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
backend
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.