Marcus Brubaker

Agreed. According to the 2022 Google Scholar Metrics, ICLR is now 8th overall and out ranks NeurIPS at 9th: https://scholar.google.com/citations?view_op=top_venues Not including ICLR in these rankings makes no sense. The...

This should still be outstanding but it's primarily a speed issue and (to a lesser extent) a robustness/stability thing. It would be an easy fix for someone, I suspect @rtrangucci...

Arguments for ICLR being included: - Based on Google Scholar rankings based on both h5-index and h5-median (https://scholar.google.ca/citations?view_op=top_venues&hl=en&vq=eng) ICLR currently out ranks (3rd) _all_ other included machine learning venues included...

Most would likely agree that ML has three top conferences at this point: NeurIPS, ICLR and ICML. This is in line with, for instance, computer vision which also has three:...

Adding the Jacobian here won't do MAP as it is already doing MAP without the Jacobian. It would only be Maximum Likelihood if there were no priors specified in the...

Right, ok, I get this. My main concern was the MLE vs MAP framing in the original issue description. Perhaps the flag could be something about "optimize posterior in constrained...

One consideration worth mentioning in this context is that the results of "MAP in the unconstrained space" will depend on exactly how we parameterize the constraints. That means that results...

Is there an easy way to determine what the maximum supported compute capability of the linked version of CUDA is? Seems like we want to use an arch which is...

Just checking on the status of this PR @emeryberger