Second order duality in multiobjective programming (Q688526): Difference between revisions

In the introduction, the authors present the definition of the second order dual problem (formulated by Mangasarian) and points out its essential differences to the first order dual problem (Wolfe). Considering the concept of proper efficiency for Pareto optimization problems (MP) Minimize \((f_ i(x)\mid i\in Q)\), subject to \(g(x)\leq 0\), where \(f_ i: R^ n\to R\), \(i\in Q=\{1,2,\dots,q\}\) and \(g: R^ n\to R^ m\) are twice differentiable, the second order dual problem for (MP) is defined as: (D4) Maximize \((f_ i(u)+ y^ t g(u)- 1/2 p'\nabla^ 2(f_ i(u)+ y^ t g(u))p\mid i\in Q)\) subject to \(\nabla[\tau^ t f(u)+ y^ t g(u)]+ \nabla^ 2[\tau^ t f(u)ty^ t g(u)]p= 0\), \(\tau>0\), \(\tau^ t e= 1\), \(y\geq 0\). With various conditions as invexity and quasiconvexity, the author obtains three theorems for the weak duality and two for duality in the classical sense of duality.