Module solve_sum_abs

source code

Solves various equations involving minimizing the sum of absolute values of things.

Functions

solve1(y0, y1)
We minimize the equation sum_over_i(abs(y0_i*(1-x) + y1_i*x)) to find the best x.

source code

tuple(float,float)

solve_fit(x, y, epsx=1e-07, epsm=1e-07)
Solves for the line yhat = m*x + b that minimizes sum(abs(y - yhat)).

source code

tuple(float,float)

solve_fit_wt(x, y, wt, epsx=1e-07, epsm=1e-07)
Solves for the line yhat = m*x + b that minimizes sum(abs(y - yhat)).

source code

test1()

source code

test2()

source code

Variables
	__package__ = `'gmisclib'`

Imports: math, numpy, Numeric_gpk, root, WP

Function Details

solve1(y0, y1)

source code

We minimize the equation sum_over_i(abs(y0_i*(1-x) + y1_i*x)) to find the best x. Returns x. I don't know if this is strictly correct.

solve_fit(x, y, epsx=1e-07, epsm=1e-07)

source code

Solves for the line yhat = m*x + b that minimizes sum(abs(y - yhat)). In other words, it's a fit to a straight line, but a highly robust fit that will not be disturbed by outliers. The algorithm is from Numerical Recipes, Volume 2.

Parameters:

epsx (float) - a tolerance used in the iterative solution. Eps is roughly the required accuracy of the x-intercept of the solution, expressed as a fraction of the x-range of the data.
epsm (float) - a tolerance used in the iterative solution. Roughly, it is the accuracy of m in the solution. Note the tangent() call!

Returns: tuple(float,float)

(mhat, bhat), where mhat*x+bhat is the best fit to the data.

solve_fit_wt(x, y, wt, epsx=1e-07, epsm=1e-07)

source code

Parameters:

epsx (float) - a tolerance used in the iterative solution. Eps is roughly the required accuracy of the x-intercept of the solution, expressed as a fraction of the x-range of the data.
epsm (float) - a tolerance used in the iterative solution. Roughly, it is the accuracy of m in the solution. Note the tangent() call!

Returns: tuple(float,float)

(mhat, bhat), where mhat*x+bhat is the best fit to the data.