An efficient algorithm for solving convex-convex quadratic fractional programs

This paper is concerned with an efficient algorithm for solving a convex-convex type quadratic fractional program whose objective function is defined as the ratio of two convex quadratic functions and whose constraints are linear. This is a typical nonconcave maximization problem with multiple local maxima. The algorithm to be proposed here is a combination of (i) the classical Dinkelbach approach, (ii) the integer programming approach for solving nonconvex quadratic programming problems and (iii) the standard nonlinear programming algorithm. It will be shown that an exact algorithm which is a combination of (i) and (ii) above can solve problems much larger than those solved by an earlier algorithm based on a branch and bound algorithm. It addition, the combination of (i)-(iii) can solve much larger problems within a practical amount of time.