Fred Moolekamp

I'd like to suggest a slight twist to this issue, although I would also be fine with making it a separate issue if you two don't feel like they are...

On the HSC side there are two typical cases that cause this behavior: 1. The center location does not have a real source. This can happen with the HSC pipeline...

I'm more in favor of option 2, global variables are discouraged for a reason. While I know that you are in favor of just setting things once and not having...

There have been a lot of instances over time. I can't think of anything recently because it's such an uncommon practice these days. Matplotlib does what I suggest, namely having...

@pmelchior feels strongly about having a global config, which is why I used the global dict. If not, I would make a config class that is the same as the...

We also need to do a more thorough investigation of how to calculate our gradient step sizes. While ADAM isn't too sensitive to the choice of step_sizes, in the image...

@EiffL I think that the maximum number of iterations and maximum run time depends a lot on how this scales with the number of pixels and the number of objects...

Sorry for the delayed response, I've been out of the office. Will post an update soon after @pmelchior and I talk offline.

Thanks for opening this ticket, sorry that I didn't get around to it yet. You might also want to consider porting the implementation of parameters in proxmin as opposed to...

HSC PSFs are usually 41x41 or 43x43 pixels. So you should be able to test with the full HSC PSF to see if truncation is the issue.