The search for a global minimum in problems of nonconvex programming with dependences that are biseparable superpositions of convex functions (Q1385863): Difference between revisions

The authors consider a mixed integer nonlinear programming problem \(F(x) \to \min\), \(x \in Q\), where \(Q= \{ x \in D:f_{i}(x) \leq 0\), \(i \in I\), \(g_{j}(x)=0\), \(j \in J\}\), \(D= \{ x \in X:x_{k}\), \(k \in K\) are integers\(\}\); \(X\) is a convex closed set from \(R^N\), \(K \subset \{ 1, \cdots ,N \}\), \(I= \{ 1,\dots ,m_{1} \}\), \(J=\{ 1,\dots,m_{2}\}\). The functions \(F(x)\) and \(f_{i}(x)\), \(i \in I\) have special biseparable superposition form \(\sum_{s \in S} h_{s}(p_{s}(x),q_{s}(x))\), where \(h_s (p,q)\) are continuous in \(p, q\) and satisfy appropriate monotonicity assumptions, \(S\) is a finite set of indexes. Here \(p_{s}(x), q_{s}(x)\), \(s \in S\) are supposed to be convex on \(X\). The functions \(g_{j}(x), j\in J\) have similar form with affine \(p_{s}(x)\) and \(q_{s}(x)\). The authors describe a finite algorithm for the determination of an \(\varepsilon\)-approximate solution for the original problem.