stan
stan copied to clipboard
stan-dev
Stan development repository. The master branch contains the current release. The develop branch contains the latest stable development. See the Developer Process Wiki for details.
#### Summary: Tail effective sample size (ESS) is one of the latest [convergence diagnostics](https://arxiv.org/abs/1903.08008). While the packages RStan, posterior, and ArviZ use this new calculation, Stan/CmdStan does not yet. Tail...
#### Summary: Rhatv5 is one of the latest [convergence diagnostics](https://arxiv.org/abs/1903.08008). While the packages RStan, posterior, and ArviZ use this new calculation, Stan/CmdStan does not yet. The following functionality needs to...
#### Summary: In analyze service - [ ] Offer interfaces a unified calculation of rhat - [x] Update effective sample size calculation - [x] Updating ess requires a small change...
Right now, `mcmc/chains.hpp` is abusing Eigen to create Eigen vectors of `std::string` and Eigen vectors of Eigen vectors. - [x] replace Eigen vectors of strings with `std::vector` - [ ]...
The base class `variational::base_family` has no virtual methods and the one function it does define uses no member variables or functions, so can be made standalone outside of the class....
[`write_iteration_csv(...)`](https://github.com/stan-dev/stan/blob/develop/src/stan/services/io/write_iteration_csv.hpp) currently requires specifying the log probability `lp`. This is unneeded for variational inference, and as a hack currently writes 0.0 there: https://github.com/stan-dev/stan/blob/develop/src/stan/variational/advi.hpp#L315-L317 From [stan-dev mailing list](https://groups.google.com/forum/#!topic/stan-dev/Bo5fWHI12iA): ``` Let's look...
#### Summary: For the code in for instance [`base_nuts::transition()`](https://github.com/stan-dev/stan/blob/develop/src/stan/mcmc/hmc/nuts/base_nuts.hpp#L79) there's a bunch of ps_point's that are being used, but the type of the derived class of ps_point can come from...
#### Summary: In a lot of `mcmc` we have `get_*` and `set_*` methods. I think it would be nice to just have const overloaded mutator and accessor methods. For instance...
#### Summary: I'm looking at `base_mcmc` and it looks like it's virtualness is only used in a few functions ``` run_sampler generate_transitions write_adapt_finish ``` I think we could rewrite that...
#### Summary: see discussion https://discourse.mc-stan.org/t/generated-quantities-for-0-dim-vectors-matrices/11514/8 and here https://github.com/stan-dev/rstan/issues/708 #### Description: before calling `transform_inits`, create var_context where all params are present with default value 0 - because `transform_inits` tries to unconstrained...
