On a discrete nonlinear and nonseparable knapsack problem

From MaRDI portal





The author considers a problem of the type \[ (EXI):\;\text{ minimize } f(\hat x)=\sum_{j\in A}(p^*_ j-p_ j)\cdot \exp (-\sum_{k\in K}d_{jk}x_ k) \] subject to \(\sum \beta_ kx_ k\leq B\), \(x_ k=0,1,2,...\), \(\forall k\in K\), where \(p_ j\) is the initial probability and \(p^*_ j\) is the limiting probability \((p^*_ j>p_ j)\). This problem with \(x_ k\geq 0\), \(\forall k\in K\), and \(\sum \beta_ kx_ k=B\), is of the type (EX). To solve problem (EX), the quadratic knapsack problem (EXQ) is considered. On the basis of an algorithm solving (EX) a branch-and-bound algorithm for the solution of problem (EXI) is developed and computational results are presented. The algorithm obtained is an efficient algorithm to solve discrete nonlinear and nonseparable knapsack problems.











This page was built for publication: On a discrete nonlinear and nonseparable knapsack problem

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q920849)