DeterminantalRepresentations -- computing determinantal representations of polynomials

The goal of this package is to compute symmetric determinantal representations of real polynomials. A polynomial $f$ in $\mathbb{R}[x_1, \ldots, x_n]$ of total degree $d$ (not necessarily homogeneous) is called determinantal if $f$ is the determinant of a matrix of linear forms - in other words, there exist matrices $A_0, \ldots, A_n\in \mathbb{R}^{d\times d}$ such that $f(x_1, \ldots, x_n) = det(A_0 + x_1A_1 + \ldots + x_nA_n)$. The matrix pencil $A_0 + x_1A_1 + \ldots + x_nA_n$ is said to give a determinantal representation of $f$ of size $d$. If the matrices $A_i$ can be chosen to be all symmetric, then the determinantal representation is called symmetric. The determinantal representation is called definite if $A_0$ is positive definite, and monic if $A_0 = I_d$ is the identity matrix.

Deciding whether or not a degree $d$ polynomial has a determinantal representation of size $d$ is in general difficult, and computing such a representation even more so. Computing (monic) symmetric determinantal representations (even in 2 variables) is of interest owing to a connection with real-zero and hyperbolic polynomials, due to a celebrated theorem of Helton-Vinnikov. In general, determinantal polynomials also have connections to convex algebraic geometry and semidefinite programming.

Currently, the functions in this package are geared towards computing monic symmetric determinantal representations of quadrics, as well as plane curves of low degree (i.e., cubics and quartics). The algorithms implemented in this package can be found in [1], [2].

Additionally, a number of helper functions are included for creating/working with various classes of matrices, which may be of general interest (and are not limited to the scope of determinantal representations). These include: creating/testing orthogonal, symmetric, doubly stochastic, unipotent, and positive semidefinite matrices, Hadamard products, Cholesky decomposition, and lifting/rounding matrices from CC to RR/QQ.

References:

Authors

• Justin Chen <[email protected]>
• Papri Dey <[email protected]>

Certification

Version 1.3.0 of this package was accepted for publication in volume 10 of The Journal of Software for Algebra and Geometry on 5 December 2019, in the article Computing symmetric determinantal representations. That version can be obtained from the journal or from the Macaulay2 source code repository.

Version

This documentation describes version 1.3.1 of DeterminantalRepresentations.

Source code

The source code from which this documentation is derived is in the file DeterminantalRepresentations.m2.

Exports