Partially normalized pivot selection in linear programming

An algorithm is presented for pivot selection in the simplex method using the partial norm of the updated column to dynamically scale the reduced costs. The geometry of this partial norm method is compared with several known methods including the usual simplex pivot selection rule and Harris’ Devex procedure. Computational comparisons between the partial norm method and other methods show our algorithm to be a surprisingly effective procedure. We show the partial norm procedure to be amenable to simplex routines using multiple pricing.