Duality in generalized nonlinear fractional programming (Q1206878): Difference between revisions

Considered is the following generalized nonlinear fractional programming problem \[ \overline\theta=\inf\left\{\max_{1\leq i\leq p} f_ i(x)/g_ i(x)\mid h(x)\leq 0,\;x\in K\right\},\tag{P} \] where \(K\) is a subset of \(R^ n\), \(f_ i: K\to R\), \(g_ i: K\to R\) \((i=1,\dots,p)\), \(h: K\to R^ m\); all \(g_ i\) are assumed to be positive on \(K\). The author proposes two types of duals for (P): \[ \sup_{u\geq 0,\| u\|=1,v\geq 0}\inf_{x\in K} \bigl(u^ T F(x)+v^ T h(x)\bigr)/u^ T G(x),\leqno(D_ 1) \] \[ \sup_{v\geq 0}\inf_{x\in K}\left(\max_{1\leq i\leq p} f_ i(x)/g_ i(x)+\sum^ m_{j=1} v_ j\max_{1\leq i\leq p} h_ j(x)/g_ i(x)\right),\leqno(D_ 2) \] where \(F(x)=\bigl(f_ 1(x),\dots,f_ p(x)\bigr)^ T\), \(G(x)=\bigl(g_ 1(x),\dots,g_ p(x)\bigr)^ T\), \(u\in R^ p\) and \(v\in R^ m\). Four theorems of strong duality are established with the help of alternative theorems in the convex case.