Examples
Two complete high-dimensional examples accompany the project.
GLMM with random intercept and slope
A Gaussian mixed model with group-specific random intercepts and slopes, followed by alternative response families. The model demonstrates parameter-local group updates, multivariate random-effect priors, mean-only storage for a large latent state, and a shared predictor cache.
High-dimensional sparse variable selection
A Gaussian sparse regression with binary inclusion indicators and 50,000 candidate predictors. The model demonstrates bounded discrete parameters, state-dependent likelihood evaluation, exact rank-one cache updates, and posterior inclusion probabilities.