부산대학교 도서관

A convex cone approach to linear programming problems is outlined. Moredetails of this work appeared earlier in a technical report. Accordingto the authors, this approach allows a reduction of the dimensions of theproblem in certain cases; it provides opportunities for parallel computingand the possibility of quick recomputation of an optimal solution aftercertain changes in the parameters. The method based on cones isto find tangency conditions of an associated affine space. A method isdetailed to study feasibility, boundedness and thus optimality.Relative positions of a linear subspace using concepts of stricttangency, internality and tangency are introduced. Search for extremerays of a cone is tied to the linear independence. The authors stressthe point that the theory of strict tangent relaxation is independentof any particular method used to solve a linear programming problem.