LCOGT / banzai-nres / 590

pending completion

Build # 590

Build Type

Pull #34

travis-ci

Committed by

web-flow

Commit Message

<a href="https://github.com/LCOGT/banzai-nres/commit/<a class=hub.com/LCOGT/banzai-nres/commit/d04f16e0926acadbc161b254a19b474cd2e1a2c2">d04f16e09<a href="https://github.com/LCOGT/banzai-nres/commit/d04f16e0926acadbc161b254a19b474cd2e1a2c2">">Merge </a><a class="double-link" href="https://github.com/LCOGT/banzai-nres/commit/<a class="double-link" href="https://github.com/LCOGT/banzai-nres/commit/69b1a6b331c784c8da2c5868bd34e59295e54a13">69b1a6b33</a>">69b1a6b33</a><a href="https://github.com/LCOGT/banzai-nres/commit/d04f16e0926acadbc161b254a19b474cd2e1a2c2"> into 263178421">263178421</a>

Pull Request Pull Request #34: Add the ability to normalize out the continuum.

Run Details

110 of 110 new or added lines in 9 files covered. (100.0%)

86369 of 290515 relevant lines covered (29.73%)

0.3 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

75.0

/banzai_nres/fitting.py

import numpy as np
from scipy.interpolate import SmoothBivariateSpline, UnivariateSpline
from statsmodels.robust.norms import HuberT
from astropy.modeling import fitting, polynomial


def fit_smooth_spline(y, error, mask=None, n_iter=1, x=None, sigma=5):
    if mask is None:
        mask = np.zeros(y.shape, dtype=bool)
    y_to_fit = y[np.logical_not(mask)]
    error_to_fit = error[np.logical_not(mask)]
    # Note that from the docs, the weights should be about 1 over the standard deviation for the smoothing to
    # work well
    weights = 1.0 / error_to_fit
    huber_t = HuberT(t=sigma)
    if len(y.shape) == 2 and x is None:
        spline = SmoothBivariateSpline
        x1d, y1d = np.arange(y.shape[0], dtype=np.float), np.arange(y.shape[1], dtype=np.float)
        x2d, y2d = np.meshgrid(x1d, y1d)
        x_to_fit = x2d.T[np.logical_not(mask)], y2d.T[np.logical_not(mask)]
        kwargs = {'grid': False}

    elif len(y.shape) == 2 or len(x) == 2:
        spline = SmoothBivariateSpline
        x_to_fit = (x[0][np.logical_not(mask)], x[1][np.logical_not(mask)])
        kwargs = {'grid': False}

    elif x is not None:
        spline = UnivariateSpline
        x_to_fit = (x[np.logical_not(mask)], )
        kwargs = {}
    else:
        x_to_fit = (x[0][np.logical_not(mask)],)
        spline = UnivariateSpline
        kwargs = {}

    for i in range(n_iter):
        best_fit = spline(*x_to_fit, y_to_fit, w=weights)

        # Iteratively reweight to reject outliers
        weights = huber_t.weights(np.abs(y_to_fit - best_fit(*x_to_fit, **kwargs)) / error_to_fit) / error_to_fit
    return best_fit


def fit_polynomial(y, error, mask=None, n_iter=5, x=None, sigma=5, order=3, domain=None):
    if mask is None:
        mask = np.zeros(y.shape, dtype=bool)

    if x is None:
        x = np.arange(len(y), dtype=float)
    y_to_fit = y[np.logical_not(mask)]
    error_to_fit = error[np.logical_not(mask)]
    x_to_fit = x[np.logical_not(mask)]
    # Start with variance weighting
    weights = error_to_fit ** -2.0
    huber_t = HuberT(t=sigma)

    fitter = fitting.LinearLSQFitter()
    model = polynomial.Legendre1D(order, domain=domain)

    for i in range(n_iter):
        best_fit = fitter(model, x_to_fit, y_to_fit, weights=weights)

        # Iteratively reweight to reject outliers
        weights = huber_t.weights(np.abs(y_to_fit - best_fit(x_to_fit)) / error_to_fit) / error_to_fit ** 2.0
    return best_fit

1	import numpy as np	1✔
2	from scipy.interpolate import SmoothBivariateSpline, UnivariateSpline	1✔
3	from statsmodels.robust.norms import HuberT	1✔
4	from astropy.modeling import fitting, polynomial	1✔
5
6
7	def fit_smooth_spline(y, error, mask=None, n_iter=1, x=None, sigma=5):	1✔
8	if mask is None:	1✔
9	mask = np.zeros(y.shape, dtype=bool)	1✔
10	y_to_fit = y[np.logical_not(mask)]	1✔
11	error_to_fit = error[np.logical_not(mask)]	1✔
12	# Note that from the docs, the weights should be about 1 over the standard deviation for the smoothing to
13	# work well
14	weights = 1.0 / error_to_fit	1✔
15	huber_t = HuberT(t=sigma)	1✔
16	if len(y.shape) == 2 and x is None:	1✔
17	spline = SmoothBivariateSpline	×
18	x1d, y1d = np.arange(y.shape[0], dtype=np.float), np.arange(y.shape[1], dtype=np.float)	×
19	x2d, y2d = np.meshgrid(x1d, y1d)	×
20	x_to_fit = x2d.T[np.logical_not(mask)], y2d.T[np.logical_not(mask)]	×
21	kwargs = {'grid': False}	×
22
23	elif len(y.shape) == 2 or len(x) == 2:	1✔
24	spline = SmoothBivariateSpline	×
25	x_to_fit = (x[0][np.logical_not(mask)], x[1][np.logical_not(mask)])	×
26	kwargs = {'grid': False}	×
27
28	elif x is not None:	1✔
29	spline = UnivariateSpline	1✔
30	x_to_fit = (x[np.logical_not(mask)], )	1✔
31	kwargs = {}	1✔
32	else:
33	x_to_fit = (x[0][np.logical_not(mask)],)	×
34	spline = UnivariateSpline	×
35	kwargs = {}	×
36
37	for i in range(n_iter):	1✔
38	best_fit = spline(*x_to_fit, y_to_fit, w=weights)	1✔
39
40	# Iteratively reweight to reject outliers
41	weights = huber_t.weights(np.abs(y_to_fit - best_fit(x_to_fit, *kwargs)) / error_to_fit) / error_to_fit	1✔
42	return best_fit	1✔
43
44
45	def fit_polynomial(y, error, mask=None, n_iter=5, x=None, sigma=5, order=3, domain=None):	1✔
46	if mask is None:	1✔
47	mask = np.zeros(y.shape, dtype=bool)	1✔
48
49	if x is None:	1✔
50	x = np.arange(len(y), dtype=float)	×
51	y_to_fit = y[np.logical_not(mask)]	1✔
52	error_to_fit = error[np.logical_not(mask)]	1✔
53	x_to_fit = x[np.logical_not(mask)]	1✔
54	# Start with variance weighting
55	weights = error_to_fit ** -2.0	1✔
56	huber_t = HuberT(t=sigma)	1✔
57
58	fitter = fitting.LinearLSQFitter()	1✔
59	model = polynomial.Legendre1D(order, domain=domain)	1✔
60
61	for i in range(n_iter):	1✔
62	best_fit = fitter(model, x_to_fit, y_to_fit, weights=weights)	1✔
63
64	# Iteratively reweight to reject outliers
65	weights = huber_t.weights(np.abs(y_to_fit - best_fit(x_to_fit)) / error_to_fit) / error_to_fit ** 2.0	1✔
66	return best_fit	1✔

LCOGT / banzai-nres / 590

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous