Characterizing Real-Valued Multivariate Complex Polynomials and Their Symmetric Tensor Representations

Bo Jiang, Bo Jiang 0007, Zhening Li, Shuzhong ZhangSIAM J. Matrix Anal. Appl 2016 · DOI

TL;DR

This paper introduces a framework for characterizing complex polynomials that always yield real values, a crucial property in various engineering applications. It achieves this by defining new types of symmetric complex tensors and extending the concept of eigenvalues to these tensors, offering a powerful tool for optimization problems involving complex variables.

What Problem Does This Paper Address?

Many real-world engineering problems, particularly in signal processing and control theory, involve functions of complex variables that are constrained to always produce real-valued outputs. While these functions are not analytic in the traditional sense, their real-valued nature is critical for their application. Existing complex analysis tools are not directly suited for these functions. This paper aims to bridge this gap by providing a rigorous mathematical characterization of these "real-valued conjugate complex polynomials" and developing a tensor-based framework to analyze and optimize them. This is achieved by extending concepts from real symmetric tensors to their complex counterparts, specifically focusing on symmetric tensor representations of these polynomials.

Methods

The paper introduces two main classes of complex polynomials: symmetric conjugate forms and general conjugate forms. These forms are then linked to specific types of symmetric complex tensors: partial-symmetric tensors and symmetric tensors, respectively. A key contribution is the definition of new types of eigenvalues and eigenvectors for these complex tensors, namely C-eigenvalues and G-eigenvalues. These are generalizations of existing tensor eigenvalue concepts and are shown to be always real, mirroring a property of Hermitian matrices. The paper also extends Banach's theorem, which relates the largest singular value of a tensor to its largest eigenvalue, to these new classes of complex tensors. This extension is crucial for developing optimization algorithms.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Real-valued complex polynomials are not trivial: The paper demonstrates that complex polynomials can consistently yield real values, but this requires the inclusion of conjugate terms.
  • Tensors as a natural representation: Symmetric tensors provide a powerful and structured way to represent these real-valued conjugate complex polynomials.
  • Generalizing eigenvalues: The introduction of C-eigenvalues and G-eigenvalues extends the spectral theory of tensors to complex domains, opening up new avenues for analysis and computation.
  • Banach's theorem extension: Extending Banach's theorem to complex tensors provides a vital link between tensor norms (singular values) and tensor eigenvalues, which is fundamental for optimization.
  • Connection to optimization: The developed framework directly supports the formulation and solution of complex polynomial optimization problems, which are prevalent in engineering.

Key Results

  • Characterization of real-valued conjugate complex polynomials based on coefficient properties.
  • Definition of two classes of symmetric complex tensors (partial-symmetric and symmetric) corresponding to these polynomials.
  • Introduction of C-eigenvalues and G-eigenvalues for these tensor classes, with proofs that they are always real.
  • Extension of Banach's theorem to conjugate partial-symmetric and conjugate super-symmetric tensors.
  • Demonstration of how these concepts can be applied to problems like ambiguity function shaping for radar waveforms and the best rank-one approximation of complex tensors.

Limitations and Future Work

The paper acknowledges that while the Q-eigenvalue and US-eigenvalue are related to C-eigenvalues, the precise relationships and their implications require further investigation. The authors also suggest that the new notions of symmetric complex tensors and their eigenvalues/eigenvectors could lead to new modeling opportunities and improved solution methods for complex polynomial optimization. A specific future research direction mentioned is exploring the nonnegativity of complex polynomials and its relationship to sums of squares (SOS), a problem that has been extensively studied for real polynomials but remains less explored for complex ones.

Summary updated

BibTeX
@article{jiang2016characterizing,
  author = {Bo Jiang and Bo Jiang and Zhening Li and Shuzhong Zhang},
  title = {{Characterizing Real-Valued Multivariate Complex Polynomials and Their Symmetric Tensor Representations}},
  journal = {SIAM J. Matrix Anal. Appl},
  year = {2016},
  doi = {10.1137/141002256},
}
Figure