Polyfit

class sofia_redux.toolkit.fitting.polynomial.Polyfit(*args, error=1, mask=None, covar=True, stats=True, robust=0.0, eps=0.01, maxiter=10, solver='gaussj', set_exponents=False, ignorenans=True, **kwargs)

Bases: Model

Fits and evaluates polynomials in N-dimensions.

Examples

>>> y, x = np.mgrid[:5, :5]
>>> z = 1 + x * y + x ** 2
>>> poly = Polyfit(x, y, z, 2)
>>> assert np.allclose(poly(x, y), z)
>>> print(poly.get_coefficients())
[[ 1.  0.  1.]
 [ 0.  1.  0.]
 [-0.  0.  0.]]

Attributes:

coefficientsnumpy.ndarray (ncoeffs,)

Polynomial fit coefficients

covariancenumpy.ndarray (ncoeffs, ncoeffs)

Polynomial fit covariance. Will be None if the covariance matrix was not calculated.

exponentsnumpy.ndarray (ncoeffs, ndimensions)

Polynomial exponents

Methods Summary

evaluate(independent_samples[, dovar])	Evaluate the polynomial at independent values
get_coefficients([covar])	Return coefficients and covariance suitable for general polynd use.
initial_fit()	Initial fits of polynomial to samples
refit_data(*args[, mask, error, covar])	Refit samples
refit_mask(mask[, covar])	Place holder

Methods Documentation

evaluate(independent_samples, dovar=False)

Evaluate the polynomial at independent values

get_coefficients(covar=False)

Return coefficients and covariance suitable for general polynd use.

Parameters:

covarbool, optional

If True, return the covariance matrix as well.

Returns:

coefficients, [covariance]numpy.ndarray, (numpy.ndarray or None)

Notes

The (row, col) ordering is reversed in this case since we wish to order coefficients in the same order as the data features, which the general case usage of polynd expects, but numpy does not.

initial_fit()

Initial fits of polynomial to samples

refit_data(*args, mask=None, error=None, covar=False)

Refit samples

Parameters:

argstuple of array_like

If one argument is supplied, it is assumed to be the dependent variable. Otherwise, all independent and dependent variables should be supplied. No other arguments should be supplied.

maskarray_like of bool, optional

New user mask. If none is supplied then one will be created based on the ignorenans.

errorfloat or array_like, optional

New error. If None is supplied, the old error will be used.

covarbool, optional

If True, calculate the covariance.

refit_mask(mask, covar=False)

Place holder