gully

Here is a Gif version of the MP4 from the linked Pull Request: It's slightly lower res, but more easily embeddable


Section 2: "Emulating the PHOENIX Grid" I recommend restructuring paragraphs: 1. Briefly summarize the key aspects of blasé `paper1` 2. Describe any technical changes to it (e.g. pseudo Voigt ▶️...

I consulted some literature about advanced lineshapes and found papers on the "chi-factor" (papers involving Hartmann from 1989 and onwards). I got bogged down in some of the representations and...

Key idea: We should use [PySR](https://github.com/MilesCranmer/PySR) for learning the best functional form of the broad lines. This is slightly tricky because we need to censor the narrow lines that rest...

My guess at how to make PySR work: 1. Curate an example dataset by collecting H lines from many synthetic spectra, masking sharp lines, and clipping to some number of...

I should mention that Miles Cranmer pitched a related idea at CCA last August (2022). His demo with PySR was to learn a cheaper/better functional form for the Voigt profile....

I should also mention that this Issue now feels like a blocker in the critical path--- in order to fit an entire A0V calibrator spectrum we want to infer the...

4. Benchmarking and scaling We could have a plot of, say, training time versus $N_\mathrm{lines}$, or training time versus bandwidth for fixed number of lines, or training time versus data...

5. Transfer-transfer learning We don't say it explicitly, but it's implied that the semi-empirical output of a blasé transfer learning step can serve as the *input* of a new blasé...