Monday: Things often seem more obvious in hindsight

Case for today: for a given range of wavelength, there are three closely spaced lines.  If the spectrograph resolution is insufficient to cleanly resolve these lines, they blur together into a single mass.  However, they must be comprised of symmetric Gaussians (or near-Gaussian) profiles.  Therefore, for each observed "mass," construct a symmetry calculation that is simply the sum of values to the left minus those to the right of the highest peak.  If this symmetry value (technically the absolute value of that for obvious reasons) is above a certain threshold (which I need to investigate more), declare this mass "woefully blended."  Next, use the presumably unblended portion (the one that has the less weight in the symmetry value), and subtract that off the mass.  Repeat.

Using this method, you should be able to strip off realizations of the inputs to the blend, which you can then fit more properly later.  As stated above, it relies on components being symmetric, and at least half clean.  Adding an analytic aspect to the subtracted model could fix this, although that starts to add a lot of extra complexity.  Still, it's a nice little algorithm that should be fairly easy to implement.

And then later in the day I drew a tiny picture of BMO in my notebook.

• Team Rocket.
• I also have word problems sometimes.  Very similar to this, I have to construct new words out of old to fill in for an existing word that has suddenly disappeared.
• What are you doing, hamster?
• What a jerk.
• I was looking at phones today due to unrelated things, and saw that this is going to be a thing soon.
• This is important, but I don't feel like explaining why.  Blah blah bubbles, blah blah corruption, blah blah poor central planning.