Solution of the linear regression problem using matrix correction methods in the l1 metric

Аннотация

The linear regression problem is considered as an improper interpolation problem. The metric l₁ is used to correct (approximate) all the initial data. A probabilistic justification of this metric in the case of the exponential noise distribution is given. The original improper interpolation problem is reduced to a set of a finite number of linear programming problems. The corresponding computational algorithms are implemented in MATLAB.