@Ameek-Sidhu has been experimenting with an asymmetry term to the Drude profile, which is working well fitting a single ISO feature:
For JWST or even Spitzer, we could enable asymmetry as a normally-zero (i.e. defaulting to a normal Drude) parameter, then turn it on for a few sub-features. This would cost one parameter vs. potentially 3 parameters (+ constraints) for a new sub-feature, which would be advantageous during training expansion.
A good exercise would be to "re-train" on the classic SINGS training sample, to see if for example the 7.7µm PAH could be "lightweighted" by removing one of the sub-features, and replacing it with 1 or perhaps 2 "asymmetry-on" terms on the remaining features.
@Ameek-Sidhu has been experimenting with an asymmetry term to the Drude profile, which is working well fitting a single ISO feature:
For JWST or even Spitzer, we could enable asymmetry as a normally-zero (i.e. defaulting to a normal Drude) parameter, then turn it on for a few sub-features. This would cost one parameter vs. potentially 3 parameters (+ constraints) for a new sub-feature, which would be advantageous during training expansion.
A good exercise would be to "re-train" on the classic SINGS training sample, to see if for example the 7.7µm PAH could be "lightweighted" by removing one of the sub-features, and replacing it with 1 or perhaps 2 "asymmetry-on" terms on the remaining features.