The current model is implemented in pixel space, whereas the underlying theory papers of Bernstein, Bernstein and Bernstein & Gruen (2014) is performed in Fourier Space. Anecdotally, Bernstein says that many of the parameters currently fit have analytic solutions in Fourier space. The implication that the code could be much faster if implemented in Fourier space.
definitely would be very cool! I've already got a lot of the code to do that because I move my PSFs around in fourier space, but then I come back to pixel space to compute the chi squared.
The current model is implemented in pixel space, whereas the underlying theory papers of Bernstein, Bernstein and Bernstein & Gruen (2014) is performed in Fourier Space. Anecdotally, Bernstein says that many of the parameters currently fit have analytic solutions in Fourier space. The implication that the code could be much faster if implemented in Fourier space.