bayesSSM 0.6.1
- Improved reproducibility of
pmmh
: Setting a seed now
ensures consistent results regardless of the number of cores used.
- Improvement performance of
pmmh
slightly.
- Rewrote resampling step in
particle_filter
to C++ for
improved performance.
- Added new article
detailed-overview
which goes over
some theory, implementation details, and tips for using the package
effectively.
- Improved the documentation of the package, including the README and
articles, to provide clearer explanations and examples.
bayesSSM 0.5.0
- The
particles
argument in init_fn
, which
is passed to particle_filter
and pmmh
, is
deprecated. Please use num_particles
instead. A warning will be issued if particles
is
used.
- Added support for time dependency in functions. You can now use
t
in transition_fn
and
likelihood_fn
when passing them to
particle_filter
and pmmh
. This allows for
time-varying transition and likelihood functions.
- Fixed a bug in
particle_filter
in likelihood
calculation causing it to be shifted by a constant.
- Improved robustness of
pmmh
when encountering very low
log-likelihood values.
- Added scaling for the proposal covariance when using
"logit"
in pmmh
.
- Improved documentation: updated package description, clarified text
in the README and vignette, and added more unit tests.
bayesSSM 0.4.7