Positive-semidefinite matrix algebra #
This file supplements Mathlib's Matrix.PosSemidef API for matrices indexed by arbitrary types.
It provides rank-one and constant matrices, finite Schur products, Schur powers,
products of weights over unions of finite sets, and the quadratic-form characterization.
The results apply in particular to positive-definite kernels, represented directly as matrices, but do not depend on Tau Ceti's positive-definite-function theory.
Main declarations #
TauCeti.posSemidef_rankOne: rank-one positive-semidefinite matrices.TauCeti.posSemidef_const_oneandTauCeti.posSemidef_const_of_nonneg: constant matrices.TauCeti.posSemidef_iff_finite_sum: the quadratic-form characterization.TauCeti.posSemidef_schur_finset_prodandMatrix.PosSemidef.hadamard_pow: finite Schur products and Schur powers.TauCeti.posSemidef_prod_union: the matrix(i, j) ↦ ∏_{a ∈ L i ∪ L j} w afor weights in[0, 1].
References #
- C. Berg, J. P. R. Christensen, P. Ressel, Harmonic Analysis on Semigroups (GTM 100, 1984), Chapter 3.
The rank-one matrix (a, b) ↦ star (g a) · g b is positive semidefinite for an arbitrary
index type. Such matrices are elementary building blocks for positive-semidefinite matrices;
taking g ≡ 1 gives the constant matrix 1.
The constant matrix with value 1 is positive semidefinite.
A nonnegative constant gives a positive-semidefinite constant matrix.
The quadratic-form characterization of an arbitrary-index positive-semidefinite matrix: conjugate symmetry and nonnegativity on every finite family, allowing repeated indices.
Finite pointwise Schur products of positive-semidefinite matrices are positive semidefinite.
For weights w in [0, 1] and finite sets L i, the matrix
(i, j) ↦ ∏_{a ∈ L i ∪ L j} w a is positive semidefinite.
Schur powers of a positive-semidefinite matrix are positive semidefinite.