Bump QMC to 0.3 and fix tests #446
Conversation
UniformSample is depricated by QMC hence replaced with RandomSample
lib/SurrogatesMOE/test/runtests.jl
Outdated
| @test_broken (rbf_rmse > moe_rmse) | ||
| @test_broken (krig_rmse > moe_rmse) |
There was a problem hiding this comment.
Not sure on that. There has been no change in either Kriging or RadialBasis or SurrogatesMOE
There was a problem hiding this comment.
what do the RMSE's look like?
There was a problem hiding this comment.
┌ Info: rbf
└ rbf_rmse = 0.15820029403909985
┌ Info: krig_rmse
└ krig_rmse = 0.5506133269171736
┌ Info: moe_rmse
└ moe_rmse = 3.1404026819527395There was a problem hiding this comment.
[ Info: Initializing GMM, 2 Gaussians diag covariance 3 dimensions using 120 data points
K-means converged with 2 iterations (objv = 84.11037552263687)
┌ Info: K-means with 120 data points using 2 iterations
└ 15.0 data points per parameter
┌ Info: rbf
└ rbf_rmse = 0.06176607586612947
┌ Info: krig_rmse
└ krig_rmse = 0.23790775421661017
┌ Info: moe_rmse
└ moe_rmse = 0.011154092911771194Something definitely broke
There was a problem hiding this comment.
It does diverge (this from the current changes).
The older one was line over line
|
Instead of reviving SectionSampler, can we do the suggest fix of SciML/QuasiMonteCarlo.jl#67 (comment) ? |
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## master #446 +/- ##
===========================================
+ Coverage 65.52% 78.17% +12.65%
===========================================
Files 23 23
Lines 3115 3153 +38
===========================================
+ Hits 2041 2465 +424
+ Misses 1074 688 -386 ☔ View full report in Codecov by Sentry. |
|
Awesome! |
No description provided.