adding user input emulator variance for nuclear-ROSE

[mosesyhc]

No description provided.

[wildsm]

I realize this is in draft mode, but is there an example to test? For example, if the variance is provided must it be nonnegative or positive? Is this checked?

[mosesyhc]

There is currently a local jupyter notebook that tests the use of this wrapper. The ROSE group is working on tutorials in the form of jupyter notebooks.

Variance input is not currently checked. But that is an excellent point. I will add an assertion.

[mosesyhc]

A bigger question though is how should one test such integration. ROSE software tests the emulator and surmise tests several steps of an MCMC chain? This is an emulator that does not look like any of the GP ones.

[wildsm]

A bigger question though is how should one test such integration. ROSE software tests the emulator and surmise tests several steps of an MCMC chain? This is an emulator that does not look like any of the GP ones.

Yes, I think surmise would test something with nontrivial input (e.g., a brief MCMC) and then the downstream user, rose, would test their basic case to identify whether surmise (the upstream dependency) has broken something since the last test.

[mosesyhc]

@wildsm:
A test notebook that calls nuclear-ROSE in surmise is available at: https://github.com/bandframework/rose/blob/develop/tutorials/ROSE_tutorial_2_optical_potential_surmise_UQ.ipynb

[mosesyhc]

@wildsm : one ROSE notebook for a seminar today refers to this branch - will remove branch completely after.