Global uniqueness and solvability of tensor complementarity problems for \(\mathcal{H}_+\)-tensors (Q2181671)

Define an \(m\)-th order \(n\)-dimensional tensor as a multi-array \[ \mathcal{A}=(a_{i_1\dots i_m}),\quad a_{i_1\dots i_m}\in\mathbb{R}, \quad 1\le i_1,\dots,i_m\le n. \] Let \(\mathbb{R}^{[m,n]}\) be the set of all such tensors. For \(x=(x_i)\in\mathbb{R}^n\), define the vector \(\mathcal{A}x^{m-1}\in\mathbb{R}^n\) with \(i\)th component \[ (\mathcal{A}x^{m-1})_i=\sum_{i_2,\dots,i_m=1}^na_{ii_2\dots i_m}x_{i_2}\cdots x_{i_m},\quad i=1,\dots,n. \] Given \(q\in\mathbb{R}^n\), the tensor complementarity problem TCP(\(\mathcal{A},q\)) is to find \(x\in\mathbb{R}^n\) satisfying (entrywise) \[ x\ge 0,\quad \mathcal{A}x^{m-1}+q\ge 0,\quad x^T(\mathcal{A}x^{m-1}+q)=0. \] A number \(\lambda\in\mathbb{C}\) is an eigenvalue of \(\mathcal{A}\) if there is \(x\in\mathbb{R}^n\setminus\{0\}\) such that \(\mathcal{A}x^{m-1}=\lambda x^{[m-1]}\), where \(x^{[m-1]}=(x_i^{m-1})\). The spectral radius of \(\mathcal{A}\), denoted by \(\rho(\mathcal{A})\), is the largest absolute value of eigenvalues. A tensor \(\mathcal{A}\in\mathbb{R}^{[m,n]}\) is a nonsingular \(\mathcal{M}\)-tensor if \(\mathcal{A}=s\mathcal{I}-\mathcal{B}\), where \(\mathcal{I}\in\mathbb{R}^{[m,n]}\) is the identity tensor, \(\mathcal{B}\in\mathbb{R}^{[m,n]}\) is (entrywise) nonnegative, and \(s>\rho(\mathcal{B})\). The comparison tensor of \(\mathcal{A}\in\mathbb{R}^{[m,n]}\) is \(\mathcal{M}(\mathcal{A})\in\mathbb{R}^{[m,n]}\) with diagonal entries \(|a_{i_1\dots i_1}|\) and off-diagonal entries \(-|a_{i_1\dots i_m}|\). The tensor \(\mathcal{A}\) is an \(\mathcal{H}_+\)-tensor if \(\mathcal{M}(\mathcal{A})\) is a nonsingular \(\mathcal{M}\)-tensor and all diagonal entries of \(\mathcal{A}\) are positive. The authors give a sufficient condition for the global unique solvability of TCP(\(\mathcal{A},q\)), where \(\mathcal{A}\) is an \(\mathcal{H}_+\)-tensor. Under certain assumptions, they also present a nonlinear dynamical system model for solving TCP(\(\mathcal{A},q\)) and show its Lyapunov stability. Computer experiments illustrate its effectiveness.