论文信息 - A new descent algorithm for the least absolute value regression problem

A new descent algorithm for the least absolute value regression problem

This paper presents a new and simple algorithm for the least absolute value regression problem. It is based on the notion of “edge” descent along the surface of the objective function. It is comparable or better in computational efficiency to current linear programming approaches for roughly 4 or fewer independent variables.

G. O. Wesolowsky

[1] CXIII.On the use of medians for reducing observations relating to several quantities , 1923 .

[2] O. J. Karst,et al. Linear Curve Fitting Using Least Deviations , 1958 .

[3] Walter D. Fisher. A Note on Curve Fitting with Minimum Deviations by Linear Programming , 1961 .

[4] J. Rice,et al. Norms for Smoothing and Estimation , 1964 .

[5] Karl H. Usow,et al. On L_1 Approximation II: Computation for Discrete Functions and Discretization Effects , 1967 .

[6] J. G. Hunt,et al. The Small Sample Properties of Simultaneous Equation Least Absolute Estimators vis-a-vis Least Squares Estimators , 1970 .

[7] E. Schlossmacher. An Iterative Technique for Absolute Deviations Curve Fitting , 1973 .

[8] Ian Barrodale,et al. Algorithm 478: Solution of an Overdetermined System of Equations in the l1 Norm [F4] , 1974, Commun. ACM.

[9] J. Gentle. Least absolute values estimation: an introduction , 1977 .

[10] Ronald D. Armstrong,et al. Least Absolute Value Estimates for a Simple Linear Regression Problem , 1978 .

[11] A. Conn,et al. Minimization Techniques for Piecewise Differentiable Functions: The l_1 Solution to an Overdetermined Linear System , 1978 .

[12] Edward L. Frome,et al. A revised simplex algorithm for the absolute deviation curve fitting problem , 1979 .