Hi everyone!
I’m exploring SimPEG with the 1D RESOLVE Bookpurnong Inversion jupyter notebook. It’s nice, because it is related to my research.
However, I cannot reproduce the inversion from the paper. I changed the notebook a little so that it solves only one inversion problem for one sample. Then I compare the outcome of the inversion (as defined in the notebook) with the outcome as stored in the booky_resolve.hdf5-file.
I get this result:
Could it be that the inversion parameter from the stored outcome is not equal to the one in the notebook?
Hmm… Were you able to fit the data?
My gut feeling, we fixed the beta value, and there were changes made in our regularization function, so that beta value is not relevant any more. Can you try few inversions with decreasing beta values, and see if you can get similar results?
I have been working through the1D Resolve Booky tutorial. I am interested in looping through a set of soundings and stitching together the models.
I am using the version that was modified in 2020 and is on the SimPEG site.
When RunIt is set to False it picks up the model data and plots the outputs.
When RunIt is True it should execute the inversions in a loop. I seems to start 1 inversion then gives me an error in the misfit term.
------------------- Inversion -------------------
51 # data misfit term
52 uncert = abs(dobs) * relative + floor
—> 53 dat = data.Data(dobs=dobs,standard_deviation=uncert)
54 dmisfit = data_misfit.L2DataMisfit(simulation=prb, data=dat)
56 # regularization
TypeError: Data.init() missing 1 required positional argument: ‘survey’
Looking at the API it seems data.Data is expecting to get an argument related to the survey geometry.
I am not sure I am interpreting that properly?
Is there a more recent example of an inversion in a loop that cycles through an entire line or survey?