Design application for Uncertainty Quantification

[tkoskela]

Add UQ to FB model, starting just with hypothesis testing for features

• This should be a separate executable but can be in the same purify repo.
• After getting an image out of purify, users can fiddle with it to do feature removal or whatever. This is a manual process that we're not really involved with.
• The UQ program should take the surrogate image, the purify output image, and the visibility data as inputs, and return the result of the hypothesis test.
• Run imaging first -> external (manual) processing to get alternative image
• UQ takes surrogate image, solution image, and original data
• Start with just a hypothesis test: images + data -> bool
• Telescope operator (Phi) is in purify
• Long term it would need to be distributed, but not necessarily in the first instance
• UQ should be able to be compiled with dependency on a limited set of Purify components

References

• lexci-uq-doc.pdf
• CMIC-talk.pdf

Python Code examples

• map uncertainty
• main loop
• function definitions