blackjax
blackjax copied to clipboard
Adjusted MCLMC
A change is also made to partially_refresh_momentum, which allows (adjusted) MCHMC to be the limit of MCLMC with infinite $L$. This is convenient since the user may or not want partial refreshment (in the unadjusted case, we find that it is always beneficial, but don't know yet in the adjusted case)
Addresses #676.
Checklist:
- [x] If I add a new sampler, there is an issue discussing it already;
- [x] We should be able to understand what the PR does from its title only;
- [x] There is a high-level description of the changes;
- [x] There are links to all the relevant issues, discussions and PRs;
- [x] The branch is rebased on the latest
maincommit; - [x] Commit messages follow these guidelines;
- [x] The code respects the current naming conventions;
- [x] Docstrings follow the numpy style guide
- [x]
pre-commitis installed and configured on your machine, and you ran it before opening the PR; - [x] There are tests covering the changes;
- [x] The doc is up-to-date;
- [x] If I add a new sampler* I added/updated related examples
Codecov Report
Attention: Patch coverage is 96.15385% with 5 lines in your changes missing coverage. Please review.
Project coverage is 98.60%. Comparing base (
20666de) to head (f4e8064). Report is 7 commits behind head on main.
:exclamation: Current head f4e8064 differs from pull request most recent head 477b11a
Please upload reports for the commit 477b11a to get more accurate results.
| Files | Patch % | Lines |
|---|---|---|
| blackjax/mcmc/adjusted_mclmc.py | 94.54% | 3 Missing :warning: |
| blackjax/adaptation/mclmc_adaptation.py | 97.10% | 2 Missing :warning: |
Additional details and impacted files
@@ Coverage Diff @@
## main #675 +/- ##
==========================================
- Coverage 98.72% 98.60% -0.12%
==========================================
Files 59 60 +1
Lines 2746 2874 +128
==========================================
+ Hits 2711 2834 +123
- Misses 35 40 +5
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Any advice for how to improve the code coverage here? I'm not entirely sure how to interpret the report.
I think this is ready to review again @junpenglao
EDIT: actually, let me first update the tuning algorithm.
@junpenglao Now ready to review! I will extend the sampling book's MCLMC section with the adjusted option shortly.