scalar objective

My only research contribution today (besides attending some nice talks on particle experiments) was to start the discussion on astrometry.net about what scalar objective we ought to optimize in the tweak stage. It is not completely trivial because the usual kind of objective (things like chi-squared) can be optimized simply by rejecting pesky outliers. These days, my position has been that you haven't made a measurement of anything unless you have optimized some specified—and justified—single-valued figure of merit.