Add DoobMatchingFlow

[murrellb]

No description provided.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 41 lines in your changes missing coverage. Please review.

Project coverage is 14.87%. Comparing base (b4525f7) to head (d300d36).
Report is 5 commits behind head on main.

Files with missing lines	Patch %	Lines
src/doob.jl	0.00%	39 Missing ⚠️
src/bridge.jl	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main       #8      +/-   ##
==========================================
- Coverage   17.20%   14.87%   -2.33%     
==========================================
  Files           6        7       +1     
  Lines         250      289      +39     
==========================================
  Hits           43       43              
- Misses        207      246      +39     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[murrellb]

Related: https://arxiv.org/pdf/2310.16834
Also noted in Generator Matching paper, and should be a special case of that.

[murrellb]

Likely related to Campbell et al 2024 too?

[AntonOresten]

The example runs. I don't know what a Doob is but the example runs..

[AntonOresten]

Why is the closed form of doob_guide only applicable to PiQ, UniformUnmasking, and UniformDiscrete? @murrellb

[murrellb]

Why is the closed form of doob_guide only applicable to PiQ, UniformUnmasking, and UniformDiscrete? @murrellb

Most of our processes don't construct an explicit Q matrix, and the closed form for Doob needs rows of the rates in the Q matrix (as well as a backward call). Those processes are the ones I worked through for this. I guess I ignored the process that actually does have a Q matrix, which would be trivial to do, so that could be added (even though it is less useful, generally).