search

BAMresearch / bayem

Implementation and derivation of "Variational Bayesian inference for a nonlinear forward model." [Chappell et al. 2008] for arbitrary, user-defined model errors.
MIT License
2 stars 1 forks source link

Cov matrix #102

Closed ajafarihub closed 1 year ago

ajafarihub commented 1 year ago

The covariance-enhanced version of VB is implemented together with the updated formulation and few test-cases.

ajafarihub commented 1 year ago

All the checks that are not fulfilled go to the same error: AttributeError: 'multivariate_normal_frozen' object has no attribute 'cov' However, the scipy documentation asserts that such attribute (cov) is valid: https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.multivariate_normal.html So I am wondering why the checks are not passed.

joergfunger commented 1 year ago

You are not using scipy objects, but mvn is defined in bayem here. Did you run the tests on your machine before pushing?

ajafarihub commented 1 year ago

Yes I did check it on my machine and all the tests are passed. As far as I see, the MVN object has the 'dist' method in which scipy objects are used. I also wonder why/how the changes that I've made into this branch should be related to the MVN class/tests, since I made no change to either the MVN class or its respective tests.

joergfunger commented 1 year ago

I don't think this is related to your changes, in particular since python3.7 is working, only 3.8-3.10 are not working. Did you check these with your conda environment (changing python to 3.10)?

ajafarihub commented 1 year ago

No I did not test with various python versions. However, my conda environment's python version is 3.8.2 and it just normally passes all the tests !

ajafarihub commented 1 year ago

@joergfunger , the issue was something related to scipy-v.1.10: https://www.mail-archive.com/debian-bugs-dist@lists.debian.org/msg1891357.html It is now resolved by avoiding to use version 1.10 of scipy.