Constrained games and symmetric duality with pseudo-invexity (Q913671)

A constrained game \(G=(X,Y,f)\) is given by \(X=\{x\in R^ n_+\), \(p_ k(x)\leq 0\), \(k=1,...,s\}\) \((R^ n_+\) is the positive orthant in n- dimensional Euclidean space) and \(Y=\{y\in R^ m_+\), \(q_ r(y)\leq 0\), \(r=1,...,t\}\). Note that the authors reverse the usual roles of the players so that I is the minimizer, II is the maximizer. The \(p_ k\) and \(q_ r\) are assumed to be twice continuously differentiable. Sufficient conditions are given for the equivalence of G to a given dual pair of nonlinear programming problems. Convexity conditions of \textit{B. Mond}, \textit{S. Chandra} and \textit{M. V. Durga Prasad} [Opsearch 24, 69-77 (1987; Zbl 0619.90064)] are weakened to pseudo-invexity. A real valued function h on \(S\subset R^ n\) is pseudo-invex with respect to \(\eta\) on \(S\times S\) to \(R^ n\) if \(\eta (x,u)^ T\nabla g(u)\geq 0\) implies that g(x)- g(u)\(\geq 0\). The result is applied to the class of constrained games with \(f(x,y)=g(x,y)/h(x,y)\) where g is convex on X and concave on Y, h is linear, the \(p_ k\) are concave, and the \(q_ r\) are convex.