Brian Ward comments

Results 396 comments of


                                            Brian Ward

update ESS calculation code to be consistent with bayes-kit

I can't really speak on the bayes_kit code, just the behavior I expect from tools like `stansummary` (which I believe is calculated with [this code](https://github.com/stan-dev/stan/blob/f67090af676f1ecdd413db93d3655fcd54bb4ae9/src/stan/mcmc/chains.hpp#L564))

_rng() functions that match all existing signatures available for corresponding _lpdf() functions

I think the primary work for this issue would be in the math library. Based on the compiler, here are a list of distributions for which I believe the RNG...

virtual declarations missing for `get_constrained_sizedtypes()` and `get_constrained_sizedtypes()` in `model_base`

I’m not sure why these methods exist in the generated model to be honest, since nothing currently seems to use them. Do you have a use in mind for them?

virtual declarations missing for `get_constrained_sizedtypes()` and `get_constrained_sizedtypes()` in `model_base`

They were added back in 2019 but never exposed or used anywhere: https://github.com/stan-dev/stanc3/pull/226 Using `std::vector` for objects is more or less singlehandedly holding up the tuples PR at this point,...

report condition number of metric for HMC adaptive samplers

@andrjohns implemented a nice way of doing finite differences only for the functions which don't have full forward mode support: https://github.com/stan-dev/math/pull/2929 This is essentially a "best of both worlds" approach....

Move from int to long int?

This is blocked partially by our current IO mechanisms, since the model classes' `write_array` serializes everything as a double. Every int32 is representable in a double more-or-less exactly, but not...

Exception: In serializer: Storage capacity [763] exceeded while writing value of size [16] from position [759]

Sounds like this is another example of https://github.com/stan-dev/stanc3/issues/1295. One of the optimizations we currently have enabled prevents initializing a variable in Stan if it is immediately assigned to. However, this...

uses Eigen's internal multiindexing for multi indexing on vectors and matrices

@SteveBronder did you get a chance to try this in your stan-benchmark playground?

uses Eigen's internal multiindexing for multi indexing on vectors and matrices

I am still not sure if the benchmark you posted represents @bob-carpenter's original case he was concerned with, since it doesn't ever end up calling a lpdf. The concern as...

Cleanup varmat multi-index to use a hashmap instead of a set

I'd like to see some more benchmarks on this before merging. For the model @mike-lawrence was using on the forums, I was able to re-create the following numbers | Matrix...