Smooth calibration solutions using GPR [WIP] #149
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…t for single correlation data.
|
Has this PR been superseded by #311 @landmanbester? |
|
Indeed. Sorry forgot I had this one open |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This seems to be working apart from some weirdness with the phase smoothing that I cant really explain. I get really weird results when smoothing the phases while specifying a reference antenna. The reason I though this would be useful is because there seems to be more physical structure in the solutions when setting a reference antenna. Here are the ML solutions without a reference antenna for example
and here they are with the reference antenna specified (surprising?)
Clearly, in the latter case, there is some physical trend that we pick up by eye i.e. quick variation in time and slow variation in frequency. I thought it would be useful to exploit this when manually specifying the hyper-parameters but for some reason I can't yet fathom that breaks the smoother (conjugate gradient does not converge and gives weird results). If I smooth the raw (but unwrapped) phases and specify the reference antenna only for plotting purposes I get a very sensible reconstruction and no failures (conjugate gradient always converges)
At this stage I have no idea why this happens but at least the smoothing seems to be working. For reference here are the ML amplitudes
and the smoothed version
Compared to the phases, smoothing the amplitudes is a lot more stable. I suppose this is because jhjinv is actually a lower bound on the variance of the amplitudes whereas it's just a thumb suck for the phases.