Closed dehann closed 6 months ago
I worked through many of the papers on bandwidth selection and leave one out. Several bugs fixed in this PR. Major ones were leave one out was not being done, and scaling around LOO was not done correctly. This PR doesnt yet achieve Optim usage but does make quite a lot of progress toward:
Attention: Patch coverage is 28.57143%
with 70 lines
in your changes are missing coverage. Please review.
Project coverage is 52.10%. Comparing base (
f401955
) to head (8d47848
).:exclamation: Current head 8d47848 differs from pull request most recent head 27b45d4. Consider uploading reports for the commit 27b45d4 to get more accurate results
Files | Patch % | Lines |
---|---|---|
src/services/ManellicTree.jl | 32.89% | 51 Missing :warning: |
src/services/KernelEval.jl | 9.52% | 19 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
I had a quick look and it seems that the bandwidth scale in
AMP.entropy
is the issue, or the entropy itself, as it doesn't give the same answer as KDE.jl.