pyacs.lib.robustestimators module

RobustEstimators.py includes a number of robust estimators to solve linear problems.

exception pyacs.lib.robustestimators.Error[source]

Bases: Exception

Base class for exceptions in module robustestimators.py

exception pyacs.lib.robustestimators.UnboundedFunctionError[source]

Bases: Exception

Exception raised for unbounded objective function.

pyacs.lib.robustestimators.Dikin(A, y, W, eps=0.003)[source]

L1-norm estimation of parameters x using the Dikin’s method using Linear optimization in linear model y=Ax+e

Parameters

  • A – Model matrix

  • y – observed values (observation vector)

  • W – Weight matrix of observables (of DIAGONAL type)

  • eps – small value for iteration (default: eps = 1e-6)

:return : x,e: vector of estimated parameters and residuals

Note

translated from Matlab code kindly provided by Amir Khodabandeh june.2009

reference:Recursive Algorithm for L1 Norm Estimation in Linear Models, A. Khodabandeh and A. R. Amiri-Simkooei, JOURNAL OF SURVEYING ENGINEERING ASCE / FEBRUARY 2011 / 1 doi:10.1061/ASCESU.1943-5428.0000031 Translated to python from Matlab original by J.-M. Nocquet July 2011

pyacs.lib.robustestimators.Dik_m(c, A, b, er)[source]

subject:solve the standard form of linear programming by affine/Dikin’s method(“an interior point method”) minimize z=c’*x; subject to Ax=b; input:(c):coefficients of objective function(z) as n-vector(hint:a column vector) (A):matrix of constraint set with size m*n (b):m-vector of constraint set (er):maximum discrepancy between two iteration.(“stopping criterion”) output:(X):unknown variables (z):optimal value of objective function (D):Centering transformer “D”(a diagonal matrix) Amir khodabandeh Oct.2008