search

JuliaGaussianProcesses / AugmentedGPLikelihoods.jl

Provide all functions needed to work with augmented likelihoods (conditionally conjugate with Gaussians)
MIT License
20 stars 1 forks source link

Add example for hyperparameter tuning #89

Closed simsurace closed 1 year ago

simsurace commented 2 years ago

This is supposed to close #76

codecov[bot] commented 2 years ago

Codecov Report

Merging #89 (a8982da) into main (79cde8a) will decrease coverage by 21.99%. The diff coverage is n/a.

:exclamation: Current head a8982da differs from pull request most recent head 7938611. Consider uploading reports for the commit 7938611 to get more accurate results

@@             Coverage Diff             @@
##             main      #89       +/-   ##
===========================================
- Coverage   95.97%   73.97%   -22.00%     
===========================================
  Files          13       16        +3     
  Lines         497      661      +164     
===========================================
+ Hits          477      489       +12     
- Misses         20      172      +152
Impacted Files Coverage Δ
src/utils.jl 63.63% <0.00%> (-36.37%) :arrow_down:
src/TestUtils.jl 82.53% <0.00%> (-13.16%) :arrow_down:
src/SpecialDistributions/ntdist.jl 95.45% <0.00%> (-4.55%) :arrow_down:
src/likelihoods/studentt.jl 100.00% <0.00%> (ø)
src/likelihoods/negativebinomial.jl 100.00% <0.00%> (ø)
src/SpecialDistributions/SpecialDistributions.jl 60.00% <0.00%> (ø)
src/likelihoods/categorical.jl 1.44% <0.00%> (ø)
...cialDistributions/polyagammanegativemultinomial.jl 0.00% <0.00%> (ø)
src/SpecialDistributions/negativemultinomial.jl 0.00% <0.00%> (ø)

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 79cde8a...7938611. Read the comment docs.

theogf commented 2 years ago

Nice thanks! If that's okay with you, could we merge #70 first and adapt this PR accordingly?

simsurace commented 2 years ago

I will be making the necessary adaptations shortly.

simsurace commented 2 years ago

The figure is probably not needed, but will be auto-generated by Literate, right?

simsurace commented 2 years ago

Unfortunately after more testing it is still not working reliably. The convergence is sometimes very fast but the solution not very good. Playing around with the data size, ground truth parameters, and random seed should reveal these problems.