Improved dual network simplex

This paper improves the number of pivot steps required for the dual network simplex algorithm. We develop a simplified version of Orlin’s strongly polynomial minimum-cost flow algorithm [lo], and show how to convert it to a dual network simplex. Our pivoting strategy leads to an O(m2 logn) bound1 on the number of pivots, which is better by a factor of m compared to the previously best pivoting strategy due to Orlin [B].