- Fixed bug: bbricks::pdsInverse returns wrong value when the input is not a matrix.
Add method dAllIndicators
to all DP family distributions. dAllIndicators
will return the probabilities of all the possible values of the hidden indicator variables.
- Fixed bug: DP, HDP and HDP2 objects don't initialize observation distribution when created.
- Fixed bug:
rPosteriorPredictive.HDP
andrPosteriorPredictive.HDP2
return samples with unexpected name such as "z.z". - Fixed bug: fix the error
Error in obj$Z$Z2[[j]] : subscript out of bounds
whenrPosteriorPredictive.HDP
try to draw samples from an unrecorded group.
- Fix graphical model representation error in package vignette.
- Fix wrong conditional probability distributions in DP related documentations.
- Add HMM and HDP-HMM (iHMM) examples to the vignette
Enhance documentations.
- fix some display errors in function documentations.
- add example "hierarchical Bayesian linear regression" to the package vignette.
Add features and enhance documentations. Fix minor bugs.
New Model Structures:
- Linear Gaussian system.
- Gaussian and Gaussian conjugate structure.
- Gaussian and Inverse-Wishart conjugate structure.
New Inference Tasks:
- Sample from the posterior distribution.
- Calculate density values from the posterior distribution.
New Distributions:
- Inverse Gamma distribution.
- Wishart and Inverse-Wishart distributions.
Fixed bug: rGaussian returns "non-conformable arrays" error when sample mean "mu" is a 1 column matrix.
Reformat all the function documentations, now they are more readable.
Fix bug.
Fixed bug: posterior.CatHDP, posteriorDiscard.CatHDP, posterior.CatHDP2,posteriorDiscard.CatHDP2 update posterior improperly when sample weight "w" is not NULL.
This is the first release of bbricks.
New Model Structures:
- Gaussian and NIW conjugate.
- Gaussian and NIG conjugate.
- Categorical and Dirichlet conjugate.
- Categorical and Dirichlet process.
- Categorical and
- Dirichlet process.
- Hierarchical Dirichlet process.
- Hierarchical Dirichlet process with two layers of hierarchies.
New Inference Tasks:
- Update posterior info.
- Calculate sample sufficient statistics.
- Calculate MAP estimate.
- Calculate marginal likelihood.
- Calculate posterior predictive density.
- Sample from posterior predictive distribution.
None
None