CHOMPACK is a library of algorithms for matrix computations with chordal sparsity patterns. It includes routines for Cholesky factorization and maximum determinant positive definite completion of chordal matrices, evaluation of the gradient, Hessian, and inverse Hessian of the logarithmic barrier function of a cone of positive definite matrices with chordal sparsity pattern, and evaluation of gradient, Hessian, and inverse Hessian of the conjugate barrier.
The library provides efficient multifrontal implementations of the algorithms in the paper Covariance selection for non-chordal graphs via chordal embedding by J. Dahl, L. Vandenberghe, V. Roychowdhury (Optimization Methods and Software 23 (4), 501-520, 2008).
CHOMPACK is available as a source distribution, and precompiled binary distributions are available for a small number of platforms (see the Download page).
CHOMPACK is developed by Joachim Dahl (firstname.lastname@example.org), Lieven Vandenberghe (email@example.com), and Martin Andersen (firstname.lastname@example.org).