search

ojscutt / sl-pitchfork

emulating individual modes of solar-like oscillators using a single branching neural network
0 stars 0 forks source link

dynesty demon: constraints from different observables #9

Closed ojscutt closed 3 months ago

ojscutt commented 4 months ago

Plot(s) here for comparison between sampling runs where I allow constraints from different subsets of the observables:

ojscutt commented 4 months ago

Plots for no constraint from any of the observables (i.e. all obs unc = 1e9) image image

ojscutt commented 4 months ago

Repeat for temperature constraint: image image

ojscutt commented 4 months ago

Repeat for Teff, L constraints: image image

ojscutt commented 4 months ago

Repeat for L, Teff, Feh: image image

ojscutt commented 4 months ago

And now (prepare for funky posteriors) we include n=25 mode: image image

ojscutt commented 4 months ago

Posterior predictives for the above: image

ojscutt commented 4 months ago

As above, now including n=24 too: image image image

ojscutt commented 4 months ago

Adding n=26... image image image

ojscutt commented 4 months ago

What about including just the high or low radial orders? I'll start with n=6,7,8: image image image

ojscutt commented 4 months ago

These look a lot better! What if I include some more of these low modes, up to 12: image image image

ojscutt commented 3 months ago

Solved by using ultranest instead of dynesty - likelihood, emulator, seem to be working as expected.