The use of floating-point calculations limits the accuracy of solutions obtained by standard LP software. We present a simplex-based algorithm that returns exact rational solutions, taking advantage of the speed of floating-point calculations and attempting to minimize the operations performed in rational arithmetic. Extensive computational results are presented. Key words: linear programming, simplex algorithm, rational arithmetic 1991 MSC: 90C05, 1991 MSC: 90C49
David Applegate, William J. Cook, Sanjeeb Dash, Da