No label defined (Q3977261)

Weak and strong duality theorems are proved for the following pair of symmetric dual multiobjective programming problems: Primal: minimize \((n_ 1(x,y)/d_ 1(x,y),\ldots,n_ k(x,y)/d_ k(x,y))'\) \[ \text{subject to }\sum_{i=1}^ k\omega_ i(d_ i\nabla_ y n_ i- n_ i\nabla_ y d_ i)\leq 0, \] \[ y^ t\sum_{i=1}^ k\omega_ i(d_ i\nabla_ y n_ i-n_ i\nabla_ y d_ i)\geq 0,\quad \omega>0,\quad \omega^ t e=1,\quad x\geq 0; \] Dual: maximize \((n_ 1(u,v)/d_ 1(u,v),\ldots,n_ k(u,v)/d_ k(u,v))^ t\) \[ \text{subject to }\sum_{i=1}^ k\omega_ i(d_ i\nabla_ x n_ i- n_ i\nabla_ x d_ i)\geq 0, \] \[ u^ t\sum_{i=1}^ k\omega_ i(d_ i\nabla_ x n_ i-n_ i\nabla_ x d_ i)\leq 0,\quad \omega >0,\quad \omega^ t e=1,\quad v\geq 0, \] where \(e=(1,1,\ldots,1)^ t\in R^ k\). All functions are defined on \(R^ n\times R^ m\) and are assumed to be twice differentiable, with the partial mappings \(n_ i(.,y)\), \(d_ i(x,.)\), \(-n_ i(x,.)\), \(-d_ i(.,y)\) being convex. This duality theory can be applied to the standard problem: minimize \((f_ 1(x)/g_ 1(x),\ldots,f_ k(x)/g_ k(x))'\), subject to \(h(x)\leq 0\), \(x\geq 0\), since it is equivalent to the primal problem obtained by taking \(n_ i(x,y)=f_ i(x)+y^ t h(x)\) and \(d_ i(x,y)=g_ i(x)\) \((i=1,\ldots,k)\), with \(h: R^ n\to R^ m\) and \(g_ i: R^ n\to R\).