As a followup to PR #63, review how regularization is used in ex2d to avoid ringing and not overfit flux bins that only have a few pixels contributing.
Perhaps minflux should be based on counting pixels rather than summing weights. Maybe it should also be based on comparison to median rather than max weights (or number of pixels).
Ringing might be better handled with constraining neighboring bins to be similar rather than a prior to 0.
Test any changes for bias and ringing with smooth spectra, emission lines, absorption lines, high S/N, and low S/N.
As a followup to PR #63, review how regularization is used in ex2d to avoid ringing and not overfit flux bins that only have a few pixels contributing.
Perhaps minflux should be based on counting pixels rather than summing weights. Maybe it should also be based on comparison to median rather than max weights (or number of pixels).
Ringing might be better handled with constraining neighboring bins to be similar rather than a prior to 0.
Test any changes for bias and ringing with smooth spectra, emission lines, absorption lines, high S/N, and low S/N.