missSBM
Handling Missing Data in Stochastic Block Models
missSBM returns after four dormant years with a stricter API and a new refinement step.
◆Recent moves
- 26d ago
Replaces raw control lists with missSBM_param(), adds polish()
Breaks the control argument in favour of a dedicated constructor with named options, and adds a post-convergence node-swap refinement that fixes misclassifications the split and merge moves cannot reach. The exploration logic also moves into testable instance methods without changing the search itself.
View source ↗ - 3y ago
Adapts to Matrix 1.4-2 and fixes HTML5 documentation
A compatibility release for a new Matrix version plus HTML5 documentation fixes and typos. The last activity before a four-year gap in the feed.
View source ↗ - 4y ago
Fixes linking against nloptR 2.0.0
Fixes linking against a new nloptR release and adds a reference to the package's JSS paper. Dependency housekeeping.
View source ↗ - 5y ago
Relaxes CRAN test tolerances to avoid random failures
Relaxes test strictness to stop random CRAN check failures and trims storage in the network representation. Infrastructure only.
View source ↗ - 5y ago
Rewrites optimisation in C++ armadillo with sparse matrices
The 1.0 release rewrote the variational optimisation steps in C++ with armadillo, moved to sparse matrices for encoding zeros and missing values, and adopted a user-controlled parallel backend. It set the performance baseline that the current release refines rather than replaces.
View source ↗ - 5y ago
Renames core functions and interfaces with the sbm package
An interface overhaul following journal review, renaming the main entry points and delegating shared structure to the sbm package. An early instance of the same instinct behind the current release: fix the interface before extending the method.
View source ↗