Numerical Techniques for Finding Estimates which Minimize the Upper Bound of the Absolute Deviation

The paper deals with the numerical techniques for finding the special type of parameter estimates based on the minimization of L_{1}-norm of error. More specifically, these estimates are derived by minimization of the upper bound of the error, which is evaluated similarly to the upper bounds on the solution of stochastic optimization problem in WP-86-72. The research reported in this paper was performed in the Adaptation and Optimization Project of the System and Decision Sciences Program