blase
blase copied to clipboard
gully
Interpretable Machine Learning for astronomical spectroscopy in PyTorch and JAX
- New paper draft ready - New HML algorithm for spectral inference
It is possible to submit a JOSS paper alongside an AASJournals paper (e.g. an ApJ or AJ paper). There's a blog post about it here: https://blog.joss.theoj.org/2018/12/a-new-collaboration-with-aas-publishing We could/shoulfd consider either...
I'm reading the draft of paper 2. Here are some notes: 1. Interpretable --> "Adaptive" `Paper1` emphasized the interpretability of assigning lots of parameters to individual line shapes. I think...
We should have an animation of how it works, for talks. Here is a demo: https://user-images.githubusercontent.com/860227/236579123-1acf7f23-e202-4c97-95d4-fc149ecdd191.mp4
@SujayShankarUT and I had a philosophical discourse on what blasé is and could/should be, including these topics: - how to structure it - its economics - its tradeoffs - its...
Really helpful for visualizing the state change
Just saw this code (h/t @kgullikson88) https://github.com/RuneDK93/telluric-autoencoder There's some useful stuff in there! It would be useful to compare this technique compares to how blasé handles tellurics.
Currently our default is to turn off continuum warping. We should consider turning this back-on, probably by following the example in `InstrumentModel`. In the long term, we may consider moving...
SparseLinearEmulator (and all the emulators for that matter) support a neat feature-- you can hand-in a `wavelength_grid` vector that is different from the native precomputed synthetic spectral model's wavelength coodinates....
We are approaching a point where we are GPU memory-limited on our laptops. That's great, it means we are pushing the computational frontier. Let's get blasé up-and-running on our local...