Second order invexity and duality in mathematical programming (Q1322041)

For the nonlinear programming primal problem `Minimize \(f(x)\) subject to \(g(x)\leq 0\)', where \(x\in\mathbb{R}^ n\) and \(f\) and \(g\) are twice differentiable functions from \(\mathbb{R}^ n\) to \(\mathbb{R}\) and \(\mathbb{R}^ m\) respectively, the second order dual is defined as `Maximze \(f(u)+ y^ t g(u)- {1\over 2} q^ t\nabla^ 2[f(u)+ y^ t g(u)]r\) subject to \(\nabla[f(u)+ y^ t g(u)]+ \nabla^ 2[f(u)+ y^ t g(u)]p= 0\) and \(y\leq 0\),' where \(p,q,r\in \mathbb{R}^ n\), where \(p\), \(q\), and \(r\) can be regarded as functions although the operators \(\nabla\) and \(\nabla^ 2\) operate only on \(f\) and \(g\). Correspondingly, second order type I invex objective and constraint functions with respect to the functions \(y(x,u)\), \(p(x,u)\), \(q(x,u)\), and \(r(x,u)\) are defined by \[ f(x)- f(u)\geq \eta(x,u)^ t\nabla f(u)+ \eta(x,u)^ t\nabla^ 2 f(u)p(x,u)-\textstyle{{1\over 2}} q(x,u)^ t \nabla^ 2 f(u)r(x,u) \] and \[ -g_ i(u)\geq \eta(x,u)^ t\nabla g_ i(u)+ \eta(x,u)^ t\nabla^ 2 g_ i(u)p(x,u)-\textstyle{{1\over 2}} q(x,u)^ t\nabla^ 2 g_ i(u)r(x,u),\;i=1,2,\dots,m. \] Second order duality is then established for problems involving such functions. An example is given.