On computing a best discrete \(L_ 1\) approximation using the method of vanishing Jacobians (Q1096541)

The work of Edgeworth in the 1880's gave the first systematic methods for solving the linear discrete \(L_ 1\) approximation. Wagner in 1959 showed that the linear discrete \(L_ 1\) approximation problem could be formulated as a linear programming problem. Recent improvements in linear programming based methods have been concerned with modifying the simplex method to accelerate the calculation of the optimal solution. This paper shows how the method of vanishing Jacobians which has been used to optimize quadratic programming problems can also be used to solve the special linear programming problem associated with computing the linear discrete \(L_ 1\) approximation.