Implements a multivariate normal distribution

j-bryan commented 9 months ago

Pull Request Description

What issue does this change request address? (Use "#" before the issue to link it, i.e., #42.)

Closes #2222

What are the significant changes in functionality due to this change request?

Implements a multivariate normal distribution in python that uses scipy.stats.multivariate_normal and scipy.stats.norm classes as a backend.

For Change Control Board: Change Request Review

The following review must be completed by an authorized member of the Change Control Board.

[x] 1. Review all computer code.
[x] 2. If any changes occur to the input syntax, there must be an accompanying change to the user manual and xsd schema. If the input syntax change deprecates existing input files, a conversion script needs to be added (see Conversion Scripts).
[x] 3. Make sure the Python code and commenting standards are respected (camelBack, etc.) - See on the wiki for details.
[x] 4. Automated Tests should pass, including run_tests, pylint, manual building and xsd tests. If there are changes to Simulation.py or JobHandler.py the qsub tests must pass.
[x] 5. If significant functionality is added, there must be tests added to check this. Tests should cover all possible options. Multiple short tests are preferred over one large test. If new development on the internal JobHandler parallel system is performed, a cluster test must be added setting, in XML block, the node <internalParallel> to True.
[x] 6. If the change modifies or adds a requirement or a requirement based test case, the Change Control Board's Chair or designee also needs to approve the change. The requirements and the requirements test shall be in sync.
[x] 7. The merge request must reference an issue. If the issue is closed, the issue close checklist shall be done.
[x] 8. If an analytic test is changed/added is the the analytic documentation updated/added?
[x] 9. If any test used as a basis for documentation examples (currently found in raven/tests/framework/user_guide and raven/docs/workshop) have been changed, the associated documentation must be reviewed and assured the text matches the example.

moosebuild commented 8 months ago

Job Test qsubs sawtooth on 415048b : invalidated by @joshua-cogliati-inl

Failed with: DirectoryNotACondaEnvironmentError: The target directory exists, but it is not a conda environment.

moosebuild commented 8 months ago

Job Test qsubs sawtooth on 04f284a : invalidated by @j-bryan

Failed with DirectoryNotACondaEnvironmentError: The target directory exists, but it is not a conda environment.

j-bryan commented 8 months ago

The last several commits, starting with this commit, add a Crow MultivariateNormalDistribution object to the constructor of the MultivariateNormalPCA class added by this pull request to check the results of the SVD of the covariance matrix of the distribution against the existing Crow implementation. This is done for now as a way to check the remainder of the implemented code. The idea here is that if inconsistencies in the SVD (e.g. signs in the U and V matrices that are not unique) are the only discrepancy between my added class and the Crow implementation, all of the tests should pass as long as the Crow classes SVD matrices are used. Rather than replace the entire matrices calculated by the RAVEN python code, I only adjust the signs of the left and right singular vectors to match those from the Crow SVD.

After applying this sign correction, there are three tests that are failing across the test machines:

tests/framework/MCMC/MetropolisTenDimMVN
tests/framework/MCMC/MetropolisTenDimPropMVN
tests/framework/pca_sparseGridCollocation/attenuCorrelation

The decomposed covariance matrix in the MetropolisTenDimMVN contains several repeated singular values, and the order of their associated vectors is not consistent between my implementation and the Crow result. Something similar happens with the MetropolisTenDimPropMVN test, but the covariance matrix in question has many more repeated singular values (9 of the 10). The vector associated with the unique singular value matches between implementations but the remaining vectors from the repeated singular values do not. This is not too surprising though, since eigenvalues of multiplicity greater than 1 do not have unique eigenvectors. Both implementations produce a valid SVD. Finally, the attenuCorrelation test is failing with a very small diff in an output XML file (difference only in the 8th or 9th significant digit). I believe this may be stemming from small numerical differences in the SVD results between implementations. I believe this test is still working as intended but needs either a regold or a tolerance added to the test.

While this isn't as nice as a script showing that the tests aren't diffing, I do believe this shows that the added code is working as intended, and the diffs in the failing tests that occur when the Crow SVD sign convention is not used are the result only of that difference in sign. @joshua-cogliati-inl, in your opinion, is this thorough enough to green light regolding effected tests?

moosebuild commented 8 months ago

Job Test qsubs sawtooth on 972af8b : invalidated by @joshua-cogliati-inl

Failed with: DirectoryNotACondaEnvironmentError: The target directory exists, but it is not a conda environment.

moosebuild commented 8 months ago

Job Test qsubs sawtooth on e1eab7f : invalidated by @joshua-cogliati-inl

Failed with: DirectoryNotACondaEnvironmentError: The target directory exists, but it is not a conda environment.

moosebuild commented 8 months ago

Job Mingw Test on e1eab7f : invalidated by @joshua-cogliati-inl

failed in fetch

moosebuild commented 8 months ago

Job Mingw Test on e1eab7f : invalidated by @j-bryan

Error when fetching Raven

moosebuild commented 8 months ago

Job Test qsubs sawtooth on e1eab7f : invalidated by @joshua-cogliati-inl

FAILED: Diff tests/cluster_tests/test_mpiqsub_long

moosebuild commented 8 months ago

Job Test qsubs sawtooth on e1eab7f : invalidated by @joshua-cogliati-inl

Failed with: DirectoryNotACondaEnvironmentError: The target directory exists, but it is not a conda environment.

moosebuild commented 8 months ago

Job Test qsubs sawtooth on e1eab7f : invalidated by @joshua-cogliati-inl

Failed with: DirectoryNotACondaEnvironmentError: The target directory exists, but it is not a conda environment.

moosebuild commented 8 months ago

Job Test qsubs sawtooth on e1eab7f : invalidated by @j-bryan

Something broke but results won''t load. Same error as before? Trying again.

idaholab / raven