The quasi-complementarity problem (Q1103530): Difference between revisions

Given \(M\in R^{n\times m}\), \(q\in R^ n\), a point-to-point mapping m(u): \(R^ n\to R^ n\), a nonlinear transformation A(u): \(R^ n\to R^ n\), and a convex cone K(u) for each \(u\in R^ n\), the generalized quasi-complementarity problem is to find a \(u\in K(u)\) satisfying \((Mu+q+A(u))\in K^*(u)\) and \(<u-m(u)\), \(Mu+q+A(u)>=0\), where \(K^*(u)\) is the polar cone of K(u). In many important applications K(u) has the form \(m(u)+K\). For this case, a unified general iterative algorithm has been suggested for the problem. It begins with a \(u_ 0\in K\) and uses the iteration \[ u_{n+1}=m(u_ n)+P_ k[u_ n-\delta E\{Mu_ n+q+A(u_ n)+L(u_{n+1}-u_ n)\}-m(u_ n)],\quad n=0,1,..., \] where \(\delta >0\) is a constant, E is a positive diagonal matrix, L is either a strictly lower or upper triangular matrix, and \(P_ k\) is the projection operator into K. Convergence properties of this algorithm are considered in the special case when \(K=[0,b]\).