Re: [eigen] Stan project using Eigen

[Bob Carpenter <carp@xxxxxxxxxxx>]

On 10/10/12 1:51 PM, Gael Guennebaud wrote:
> Hi,
>
> congratulation for the release. I added a link in our project list.

Thanks!

Anything in the Stan reference manual using vectors
or matrices is implemented on top of Eigen.  We
use determinants, inverses, Cholesky decompositions,
and eigevalues/vectors, as well as the usual matrix
ops.

We also used the unsupported FFT module for computing
effective sample size in MCMC samples.  Speed benefited
tremendously from padding the input (this was obviously
when we profiled then looked at the FFT source).

> Is there a simple example of using stan/agrad? I could not find
> anything in the manual.

Sorry, no -- I haven't written an API manual
yet, but it's near the top of our very long
to-do list.  I feel bad telling the Eigen developers
this because the Eigen manual's so good from a user
perspective.

There are simple examples of agrad in the unit tests:
  src/test/agrad/{agrad_test.cpp, matrix_test.cpp}

There's also a large pile of impenetrable
doxygen-generated doc.

- Bob