-
http://www.hindawi.com/journals/isrn/2011/164564/
It might be too slow for regular interactive use. Need to benchmark what's reasonable.
-
two ideas to better exploit the structure of the loglikelihood for possibly higher performance or higher accuracy
Currently we compute derivatives directly from the loglike.
- curse of dimension…
-
Hi @michaelweylandt!
I've created my STA9750 website - check it out!
https://slyastrologer.github.io/STA9750-2024-FALL/
-
### Is your feature request related to a problem? Please describe.
Something I noticed while debugging #21148 is that some points are evaluated twice. The Newton method needs two things to find the n…
-
This is part of INLA roadmap #340.
From the Stan [paper](https://arxiv.org/abs/2004.12550):
>One of the main bottlenecks is differentiating the estimated mode, $\theta^* $. In theory, it is stra…
-
## Summary
With reference to the email conversation with @piperfw and @gefux, I am opening a dedicated issue pertaining to a dynamical numerical backend to support GPUs and TPUs. OQuPy utilizes Ten…
-
It would be good to have numerical integration and differentiation. Integration would be using the Newton–Cotes formulas, and differentiation by the Symmetric derivative.
-
sage does not differentiate numerically at this time; a `numerical_diff` similar to Maple's `fdiff` would have helped me implement Riemann theta functions.
sage's `numerical_integral` uses GSL, whi…
-
Need to use eps=sqrt(1e-16) for double precision and sqrt(1e-8) for single precision when computing step size h for numerical differentiation. If the step size is too small, the cancellation error in…
-
This isn't an issue per se, but more of a request for comment. First, we have two separate algorithms for calculating numerical gradients. The first is a classic two point—well, technically three poin…