On the convergence of iterative methods for nonsymmetric linear complementarity problems

From MaRDI portal





The author considers the solution of the linear complementarity problem \(LCP(M,q)\): finding \(x\in\mathbb{R}^ n\) such that \(Mx+q>0\), \(x\geq 0\) and \(x^ T(Mx+q)=0\), where \(M\) is an \(n\times n\) matrix and \(q\) an \(n\)- vector. A new type of matrix splittings called \(T\)-splittings for nonsymmetric matrices is introduced. The advantage over previously defined splittings for nonsymmetric matrices is that they exist for a considerable wider class of matrices, including singular ones, that is, for any nonzero positive semidefinite plus matrix, \(T\)-splittings exist. Based upon the \(T\)-splittings an algorithm for finding the solution of \(LCP(M,q)\) is proved to converge. Let \(S(M,q)\) denote the set of the solution of \(LCP(M,q)\), and \(M=B+C\) be a \(T\)-splitting of the matrix \(M\), then the sequence \(\{x^{(k)}\}\) generated by \(x^{(0)}\in\mathbb{R}^ n\), \(x^{(k+1)}\in S(B,Cx^{(k)}+q)\) converges if and only if \(S(M,q)\) is nonempty, and the limit \(x^*\) of \(\{x^{(k)}\}\) belongs to \(S(M,q)\).











This page was built for publication: On the convergence of iterative methods for nonsymmetric linear complementarity problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3985452)