Peter Melchior
Peter Melchior
with jax, inline updates are not allowed, so we need to replace them in proximal operators by jax-functional equivalents. In the optimization algorithms we need to make sure that these...
As is the asserts, e.g. for number of constraints in GLMM, tell if something isn't right, but they leave the user guessing what has caused the problem. Instead we should...
We've used Travis for running the tests, which also made the docs, but we've run out of credits there a while back. I suggest we move the entire testing+docs script...
There are three distinct changes in the `lite` brach that could be brought into the main branch, depending on further testing * a custom `grad` function for the FFT-based convolution...
Having a ton of arguments for the init method of various function makes the code really ugly and requires a fair bit of work when we want to expose an...
When we fit HSC + HST data, the noise model is incorrect because the pixels are correlated in HST coadds. In essence, we’re overfitting the HST data. We could kill...
The multi-resolution tutorial uses a HSC PSF with 31 pixels. This is apparently too small and creates ringing in the oversampled PSF (which defines the diff kernel): ```python plt.imshow(obs_hsc._diff_kernels.image[4], vmax=1e-6)...
The runtime for gradient updates in `LowResObservation` is ~ an order of magnitude slower than the forward path. This is unusually slow, so there's probably a bottleneck. This issue is...
We build our own coadds for source initialization. The two methods are largely identical, so we should clarify how they are different or unify them.
Detection methods occasionally produce false/spurious detection. There are several part in scarlet where we either flag that, throw an exception, or attempt to fit it. I'm not sure who well...