Bedell (Flatiron) and I worked out a greedy method to optimize the regularization parameters for wobble, at least roughly. The method is necessary, because there are 72 orders and we have come to the conclusion that for every order, for every different kind of star, we are going to need a unique set of regularization parameters.
Birky (UCSD) showed that M-dwarf stars with different kinds of pathologies (fast rotation, pre-main-sequence, flaring, or binary) have larger chi-squared values against our Cannon model. We are going to leave it at that, but because we have these dependencies, we are going to be able to make data-driven spectral indicators of all these things.
Rix (MPIA) and I discussed the results that Eilers (MPIA) produced this week. We realized that we don't have to get the full density model for the tracers right; we can just show our results in the context of various sensible assumptions about those tracers. That's a simplification, and sensible, given what we have.