Experiments with Active-Set LP Algorithms Allowing Basis Deficiency

An interesting question for linear programming (LP) algorithms is how to deal with solutions in which the number of nonzero variables less than rows matrix standard form. approach that basis deficiency-allowing (BDA) simplex variations, work a subset independent columns coefficient form, wherein not necessarily represented by square matrix. We describe one such algorithm several variants. The research deals studying computational behaviour using small, extreme cases. For these instances, we must wonder parameter setting or variants are more appropriate. compare two nonsimplex active-set methods Holmström’s TomLab LpSimplex v3.0 commercial sparse primal implementation. All them update QR factorization Matlab. first implementations require fewer iterations and provide better solution quality running time.