v0.1.6

SelectedInversion v0.1.6

Diff since v0.1.5

Features

• Add support for Julia's built-in ldlt
• Add specialised algorithms for selinv diagonal of symmetric tridiagonal matrices, based on Kalman filtering & smoothing

v0.1.5

SelectedInversion v0.1.5

Diff since v0.1.4

Bugfixes

• Fix incorrect use of Symmetric leading to wrong sparsity pattern of simplicial SelInv
• Add tests to shield against this issue in the future

v0.1.4

SelectedInversion v0.1.4

Diff since v0.1.3

Performance enhancements

• Faster depermutation in simplicial SelInv

v0.1.3

SelectedInversion v0.1.3

Diff since v0.1.2

New Features

• Add selinv_diag() function for efficient computation of diagonal entries of
  the selected inverse
• Optimized implementation avoids expensive sparse matrix depermutation by
  computing permuted diagonal first
• Particularly beneficial for applications like Gaussian Markov Random Fields
  where only marginal variances (diagonal entries) are needed
• Full compatibility with both CHOLMOD and LDLFactorizations.jl
  factorizations
• Supports same API as selinv() with depermute keyword argument (defaults to
  true)

v0.1.2

SelectedInversion v0.1.2

Diff since v0.1.1

Bugfixes

• Enforce symmetry of diagonal blocks manually to improve numerical stability

v0.1.1

SelectedInversion v0.1.1

Diff since v0.1.0

Closed issues:

• TagBot trigger issue (#4)

v0.1.0

Initial release.

Supports simplicial and supernodal SelInv of sparse symmetric positive definite matrices, based on CHOLMOD.