Expand description
Linear algebra utilities: matrix decompositions and correlation handling.
decomp holds the general-purpose factorizations (Cholesky, QR,
SVD, symmetric eigen). On top of them sit the correlation tools:
empirical correlation matrices (estimated pairwise, stressed by hand,
or copied from a term sheet) frequently fail positive
semi-definiteness; nearest_correlation
repairs them with the alternating-projections algorithm of Higham
(2002), and cholesky factorizes the result for
correlated simulation.
Re-exports§
pub use cholesky::cholesky;pub use decomp::cholesky_factor;pub use decomp::cholesky_solve;pub use decomp::least_squares;pub use decomp::pseudo_solve;pub use decomp::qr;pub use decomp::qr;pub use decomp::svd;pub use decomp::svd;pub use decomp::symmetric_eigen;pub use nearest_correlation::nearest_correlation;
Modules§
- cholesky
- Cholesky factorization of correlation matrices (PSD-tolerant): the
unit-diagonal validation layer over the general factorization in
decomp::cholesky. - decomp
- Matrix decompositions, one per file — general-purpose kernels for future use across the library (regression, calibration, PCA, factor models), independent of any finance semantics.
- nearest_
correlation - Higham’s alternating-projections algorithm for the nearest correlation matrix (Higham, IMA J. Numer. Anal. 2002).