11294204788

Committed 11 Oct 2024 02:30PM UTC coverage: 66.032%. First build

Build # 11294204788

Build Type

Pull #395

github

Committed by

web-flow

Commit Message

Merge 79a05365c into 18ce85989

Pull Request Pull Request #395: np.Inf and np.array(obj, copy = False) changed

Run Details

86 of 104 new or added lines in 26 files covered. (82.69%)

9543 of 14452 relevant lines covered (66.03%)

0.66 hits per line

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

96.43

/EXOSIMS/util/InverseTransformSampler.py

import numpy as np
from scipy.interpolate import interp1d
import numbers
from EXOSIMS.util._numpy_compat import copy_if_needed


class InverseTransformSampler:
    """
    Approximate Inverse Transform Sampler for arbitrary distributions
    defined via a PDF encoded as a function (or lambda function)

    Args:
        f (function):
            Probability density function.  Must be able to operate
            on numpy ndarrays.  Function does not need to be
            normalized over the sampling interval.
        xMin (float):
            Minimum of interval to sample (inclusive).
        xMax (float):
            Maximum of interval to sample (inclusive).
        nints (int):
            Number of intervals to use in approximating CDF.
            Defaults to 10000

    Attributes:
        f, xMin, xMax
            As above


    Notes:
        If xMin == xMax, return values will all exactly equal xMin.
        To sample call the object with the desired number of samples.
    """

    def __init__(self, f, xMin, xMax, nints=10000):

        # validate inputs
        assert (
            isinstance(xMin, numbers.Number)
            and isinstance(xMax, numbers.Number)
            and isinstance(nints, numbers.Number)
        ), "xMin, xMax, and nints must be numbers."
        self.xMin = float(xMin)
        self.xMax = float(xMax)
        nints = int(nints)

        assert hasattr(f, "__call__"), "f must be callable."

        if self.xMin != self.xMax:
            ints = np.linspace(self.xMin, self.xMax, nints + 1)  # interval edges
            x = np.diff(ints) / 2.0 + ints[:-1]  # interval midpoints
            fX = f(x)
            if not isinstance(fX, np.ndarray):
                fX = np.array(fX, copy=copy_if_needed, ndmin=1)

            if len(fX) == 1:
                fX = float(fX) * np.ones(x.shape)
            F = np.hstack([0, np.cumsum(fX)])
            F /= F[-1]

            self.Finv = interp1d(F, ints)

    def __call__(self, numTest=1):
        """
        A call to the object with the number of samples will
        return the sampled distribution.
        """

        assert isinstance(numTest, numbers.Number), "numTest must be an integer."
        numTest = int(numTest)

        if self.xMin == self.xMax:
            return np.zeros(numTest) + self.xMin

        return self.Finv(np.random.uniform(size=numTest))

1	import numpy as np	1✔
2	from scipy.interpolate import interp1d	1✔
3	import numbers	1✔
4	from EXOSIMS.util._numpy_compat import copy_if_needed	1✔
5
6
7	class InverseTransformSampler:	1✔
8	"""
9	Approximate Inverse Transform Sampler for arbitrary distributions
10	defined via a PDF encoded as a function (or lambda function)
11
12	Args:
13	f (function):
14	Probability density function. Must be able to operate
15	on numpy ndarrays. Function does not need to be
16	normalized over the sampling interval.
17	xMin (float):
18	Minimum of interval to sample (inclusive).
19	xMax (float):
20	Maximum of interval to sample (inclusive).
21	nints (int):
22	Number of intervals to use in approximating CDF.
23	Defaults to 10000
24
25	Attributes:
26	f, xMin, xMax
27	As above
28
29
30	Notes:
31	If xMin == xMax, return values will all exactly equal xMin.
32	To sample call the object with the desired number of samples.
33	"""
34
35	def __init__(self, f, xMin, xMax, nints=10000):	1✔
36
37	# validate inputs
38	assert (	1✔
39	isinstance(xMin, numbers.Number)
40	and isinstance(xMax, numbers.Number)
41	and isinstance(nints, numbers.Number)
42	), "xMin, xMax, and nints must be numbers."
43	self.xMin = float(xMin)	1✔
44	self.xMax = float(xMax)	1✔
45	nints = int(nints)	1✔
46
47	assert hasattr(f, "__call__"), "f must be callable."	1✔
48
49	if self.xMin != self.xMax:	1✔
50	ints = np.linspace(self.xMin, self.xMax, nints + 1) # interval edges	1✔
51	x = np.diff(ints) / 2.0 + ints[:-1] # interval midpoints	1✔
52	fX = f(x)	1✔
53	if not isinstance(fX, np.ndarray):	1✔
NEW 54	fX = np.array(fX, copy=copy_if_needed, ndmin=1)	×
55
56	if len(fX) == 1:	1✔
57	fX = float(fX) * np.ones(x.shape)	1✔
58	F = np.hstack([0, np.cumsum(fX)])	1✔
59	F /= F[-1]	1✔
60
61	self.Finv = interp1d(F, ints)	1✔
62
63	def __call__(self, numTest=1):	1✔
64	"""
65	A call to the object with the number of samples will
66	return the sampled distribution.
67	"""
68
69	assert isinstance(numTest, numbers.Number), "numTest must be an integer."	1✔
70	numTest = int(numTest)	1✔
71
72	if self.xMin == self.xMax:	1✔
73	return np.zeros(numTest) + self.xMin	1✔
74
75	return self.Finv(np.random.uniform(size=numTest))	1✔

dsavransky / EXOSIMS / 11294204788

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