#
# Copyright (C) 2010, 2016, 2018 - 2024
# Smithsonian Astrophysical Observatory
#
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
#
import logging
import numpy
from sherpa.utils.err import ModelErr
from sherpa.utils import bool_cast, get_position, guess_amplitude, \
guess_amplitude_at_ref, \
guess_amplitude2d, guess_bounds, guess_fwhm, guess_position, \
guess_reference, interpolate, linear_interp, param_apply_limits, \
sao_fcmp
from .parameter import Parameter, tinyval
from .model import ArithmeticModel, modelCacher1d, CompositeModel, \
ArithmeticFunctionModel, RegriddableModel2D, RegriddableModel1D
from . import _modelfcts # type: ignore
warning = logging.getLogger(__name__).warning
__all__ = ('Box1D', 'Const1D', 'Cos', 'Delta1D', 'Erf', 'Erfc', 'Exp', 'Exp10',
'Gauss1D', 'Log', 'Log10', 'LogParabola', 'NormGauss1D', 'Poisson',
'Polynom1D', 'PowLaw1D', 'Scale1D', 'Sin', 'Sqrt', 'StepHi1D',
'StepLo1D', 'Tan',
'Box2D', 'Const2D', 'Delta2D', 'Gauss2D', 'SigmaGauss2D',
'NormGauss2D', 'Polynom2D', 'Scale2D', 'UserModel', 'TableModel',
'Integrate1D')
DBL_EPSILON = numpy.finfo(float).eps
# This relies on the PyArg_ParseTypleAndKeywords calls in model_extension.hh
#
ALLOWED_KEYWORDS_1D = set(["pars", "xlo", "xhi", "integrate"])
ALLOWED_KEYWORDS_2D = set(["pars", "x0lo", "x1lo", "x0hi", "x1hi", "integrate"])
def clean_kwargs(allowed, model, kwargs):
"""Remove un-supported keywords for these models.
Parameters
----------
model : Model instance
It must have an integrate field.
kwargs : dict
The input keyword arguments
allowed : set of str
The allowed keyword names.
Returns
-------
allowed : dict
The keyword arguments (names and values) that are allowed.
"""
# TODO: remove the use of bool_cast here.
#
out = {"integrate": bool_cast(model.integrate)}
for key in [k for k in kwargs.keys() if k in allowed]:
out[key] = kwargs[key]
return out
def clean_kwargs1d(model, kwargs):
"""Remove un-supported keywords for these models.
The supported arguments are ALLOWED_KEYWORDS_1D.
"""
return clean_kwargs(ALLOWED_KEYWORDS_1D, model, kwargs)
def clean_kwargs2d(model, kwargs):
"""Remove un-supported keywords for these models.
The supported arguments are ALLOWED_KEYWORDS_2D.
"""
return clean_kwargs(ALLOWED_KEYWORDS_2D, model, kwargs)
[docs]
class Box1D(RegriddableModel1D):
"""One-dimensional box function.
The model is flat between ``xlow`` and ``xhi`` (both limits are
inclusive), where it is set to the ``ampl`` parameter. Outside
this range the model is zero.
.. versionchanged:: 4.16.0
The default value for the xhi parameter has been changed from 0
to 1, and the range of ampl now matches xlow and xhi rather
than being set to -1 to 1.
Attributes
----------
xlow
Coordinate of the lower cut-off.
xhi
Coordinate of the upper cut-off.
ampl
The amplitude of the box.
See Also
--------
Box2D, Const1D, Delta1D, Scale1D, StepLo1D, StepHi1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl if xlow <= x <= xhi
= 0 otherwise
and for an integrated grid it is::
f(lo,hi) = ampl if lo >= xlow and hi <= xhi
= 0 if hi <= xlow or lo >= xhi
= ampl * g where g is the fraction of lo,hi
that falls within xlo,xhi
This behavior is different to how the amplitude is handled in
other models, such as ``Const1D``.
"""
def __init__(self, name='box1d'):
self.xlow = Parameter(name, 'xlow', 0)
self.xhi = Parameter(name, 'xhi', 1)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name, (self.xlow, self.xhi, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
lo, hi = guess_bounds(args[0])
norm = guess_amplitude(dep, *args)
param_apply_limits(lo, self.xlow, **kwargs)
param_apply_limits(hi, self.xhi, **kwargs)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.box1d(p, *args, **kwargs)
class Const(ArithmeticModel):
def __init__(self, name='const'):
self.c0 = Parameter(name, 'c0', 1)
ArithmeticModel.__init__(self, name, (self.c0,))
def guess(self, dep, *args, **kwargs):
min = dep.min()
max = dep.max()
ylo = 0
if numpy.abs(min - 0) > DBL_EPSILON:
ylo = min / 100.
if min < 0:
ylo = 0
yhi = 0
if numpy.abs(max - 0) > DBL_EPSILON:
yhi = -10 * max
if max > 0:
yhi = 100 * max
param_apply_limits({'val': (max + min) / 2.,
'min': ylo, 'max': yhi},
self.c0, **kwargs)
[docs]
class Const1D(RegriddableModel1D, Const):
"""A constant model for one-dimensional data.
Attributes
----------
c0
The amplitude of the model.
See Also
--------
Box1D, Const2D, Delta1D, Scale1D, StepLo1D, StepHi1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl
and for an integrated grid it is::
f(xlo,xhi) = ampl * (xhi - xlo)
"""
def __init__(self, name='const1d'):
Const.__init__(self, name)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.const1d(p, *args, **kwargs)
[docs]
class Cos(RegriddableModel1D):
"""One-dimensional cosine function.
Attributes
----------
period
The period of the cosine, in units of the independent axis.
offset
The offset (related to the phase) of the cosine.
ampl
The amplitude of the cosine.
See Also
--------
Sin, Tan
Notes
-----
The functional form of the model for points is::
f(x) = ampl * cos (2 * pi * (x - offset) / period)
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='cos'):
self.period = Parameter(name, 'period', 1, 1e-10, 10, tinyval)
self.offset = Parameter(name, 'offset', 0, 0, hard_min=0)
self.ampl = Parameter(name, 'ampl', 1, 1e-05, hard_min=0)
ArithmeticModel.__init__(self, name,
(self.period, self.offset, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.cos(p, *args, **kwargs)
[docs]
class Delta1D(RegriddableModel1D):
"""One-dimensional delta function.
The delta function model is only non-zero at a single point
(or bin for integrated grids).
Attributes
----------
pos
The position of the signal.
ampl
The amplitude.
See Also
--------
Box1D, Const1D, Delta2D, StepLo1D, StepHi1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl if x == pos
= 0 otherwise
and for an integrated grid it is::
f(lo,hi) = ampl if lo <= pos <= hi
= 0 otherwise
This behavior is different to how the amplitude is handled in
other models, such as ``Const1D``.
"""
def __init__(self, name='delta1d'):
self.pos = Parameter(name, 'pos', 0)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name, (self.pos, self.ampl))
[docs]
def get_center(self):
return (self.pos.val,)
[docs]
def set_center(self, pos, *args, **kwargs):
self.pos.set(pos)
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
pos = get_position(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
param_apply_limits(pos, self.pos, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.delta1d(p, *args, **kwargs)
[docs]
class Erf(RegriddableModel1D):
"""One-dimensional error function.
The function is described at [1]_.
Attributes
----------
ampl
The amplitude of the model.
offset
The offset of the model.
sigma
The scaling factor.
See Also
--------
Erfc
Notes
-----
The functional form of the model for points is::
f(x) = ampl * erf((x - offset) / sigma)
erf(y) = (2 / sqrt(pi)) Int_0^y exp(-t^2) dt
and for an integrated grid it is the integral of this over
the bin.
References
----------
.. [1] http://mathworld.wolfram.com/Erf.html
"""
def __init__(self, name='erf'):
self.ampl = Parameter(name, 'ampl', 1, 0)
self.offset = Parameter(name, 'offset', 0, 0, hard_min=0)
self.sigma = Parameter(name, 'sigma', 1, 1e-10, 10, tinyval)
ArithmeticModel.__init__(self, name,
(self.ampl, self.offset, self.sigma))
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.erf(p, *args, **kwargs)
[docs]
class Erfc(RegriddableModel1D):
"""One-dimensional complementary error function.
The function is described at [1]_.
Attributes
----------
ampl
The amplitude of the model.
offset
The offset of the model.
sigma
The scaling factor.
See Also
--------
Erf
Notes
-----
The functional form of the model for points is::
f(x) = ampl * erfc((x - offset) / sigma)
erfc(y) = (2 / sqrt(pi)) Int_y^infinity exp(-t^2) dt
and for an integrated grid it is the integral of this over
the bin.
References
----------
.. [1] http://mathworld.wolfram.com/Erfc.html
"""
def __init__(self, name='erfc'):
self.ampl = Parameter(name, 'ampl', 1, 0)
self.offset = Parameter(name, 'offset', 0, 0, hard_min=0)
self.sigma = Parameter(name, 'sigma', 1, 1e-10, 10, tinyval)
ArithmeticModel.__init__(self, name,
(self.ampl, self.offset, self.sigma))
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.erfc(p, *args, **kwargs)
[docs]
class Exp(RegriddableModel1D):
"""One-dimensional exponential function.
Attributes
----------
offset
The offset of the model.
coeff
The scaling factor.
ampl
The amplitude of the model.
See Also
--------
Exp10, Log, Log10, LogParabola, Sqrt
Notes
-----
The functional form of the model for points is::
f(x) = ampl * exp(coeff * (x - offset))
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='exp'):
self.offset = Parameter(name, 'offset', 0)
self.coeff = Parameter(name, 'coeff', -1)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name,
(self.offset, self.coeff, self.ampl))
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.exp(p, *args, **kwargs)
[docs]
class Exp10(RegriddableModel1D):
"""One-dimensional exponential function, base 10.
Attributes
----------
offset
The offset of the model.
coeff
The scaling factor.
ampl
The amplitude of the model.
See Also
--------
Exp, Log, Log10, LogParabola, Sqrt
Notes
-----
The functional form of the model for points is::
f(x) = ampl * 10^(coeff * (x - offset))
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='exp10'):
self.offset = Parameter(name, 'offset', 0)
self.coeff = Parameter(name, 'coeff', -1)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name,
(self.offset, self.coeff, self.ampl))
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.exp10(p, *args, **kwargs)
[docs]
class Gauss1D(RegriddableModel1D):
"""One-dimensional gaussian function.
Attributes
----------
fwhm
The Full-Width Half Maximum of the gaussian. It is related to
the sigma value by: FWHM = sqrt(8 * log(2)) * sigma.
pos
The center of the gaussian.
ampl
The amplitude refers to the maximum peak of the model.
See Also
--------
Gauss2D, NormGauss1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl * exp(-4 * log(2) * (x - pos)^2 / fwhm^2)
and for an integrated grid it is the integral of this over
the bin.
Examples
--------
Compare the gaussian and normalized gaussian models:
>>> from sherpa.models.basic import Gauss1D, NormGauss1D
>>> m1 = Gauss1D()
>>> m2 = NormGauss1D()
>>> m1.pos, m2.pos = 10, 10
>>> m1.ampl, m2.ampl = 10, 10
>>> m1.fwhm, m2.fwhm = 5, 5
>>> m1(10)
10.0
>>> m2(10)
1.8788745573993026
>>> m1.fwhm, m2.fwhm = 1, 1
>>> m1(10)
10.0
>>> m2(10)
9.394372786996513
The normalised version will sum to the amplitude when given
an integrated grid - i.e. both low and high edges rather than
points - that covers all the signal (and with a bin size a lot
smaller than the FWHM):
>>> import numpy as np
>>> m1.fwhm, m2.fwhm = 12.2, 12.2
>>> grid = np.arange(-90, 110, 0.01)
>>> glo, ghi = grid[:-1], grid[1:]
>>> m1(glo, ghi).sum()
129.86497637060958
>>> m2(glo, ghi).sum()
10.000000000000002
"""
def __init__(self, name='gauss1d'):
self.fwhm = Parameter(name, 'fwhm', 10, tinyval, hard_min=tinyval)
self.pos = Parameter(name, 'pos', 0)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name, (self.fwhm, self.pos, self.ampl))
[docs]
def get_center(self):
return (self.pos.val,)
[docs]
def set_center(self, pos, *args, **kwargs):
self.pos.set(pos)
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
pos = get_position(dep, *args)
fwhm = guess_fwhm(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
param_apply_limits(pos, self.pos, **kwargs)
param_apply_limits(fwhm, self.fwhm, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.gauss1d(p, *args, **kwargs)
[docs]
class Log(RegriddableModel1D):
"""One-dimensional natural logarithm function.
Attributes
----------
offset
The offset of the model.
coeff
The scaling factor.
ampl
The amplitude of the model.
See Also
--------
Exp, Exp10, Log10, LogParabola, Sqrt
Notes
-----
The functional form of the model for points is::
f(x) = ampl * log (coeff * (x - offset))
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='log'):
self.offset = Parameter(name, 'offset', 0)
self.coeff = Parameter(name, 'coeff', -1)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name,
(self.offset, self.coeff, self.ampl))
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.log(p, *args, **kwargs)
[docs]
class Log10(RegriddableModel1D):
"""One-dimensional logarithm function, base 10.
Attributes
----------
offset
The offset of the model.
coeff
The scaling factor.
ampl
The amplitude of the model.
See Also
--------
Exp, Exp10, Log, LogParabola, Sqrt
Notes
-----
The functional form of the model for points is::
f(x) = ampl * log_10 (coeff * (x - offset))
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='log10'):
self.offset = Parameter(name, 'offset', 0)
self.coeff = Parameter(name, 'coeff', -1)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name,
(self.offset, self.coeff, self.ampl))
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.log10(p, *args, **kwargs)
[docs]
class LogParabola(RegriddableModel1D):
"""One-dimensional log-parabolic function.
Attributes
----------
ref
The reference point for the normalization.
c1
The power-law index (gamma).
c2
The curvature of the parabola (beta).
ampl
The amplitude of the model.
See Also
--------
Exp, Exp10, Log, Log10, Sqrt
Notes
-----
The functional form of the model for points is::
f(x) = ampl * (x / ref) ^ (-c1 - c2 * log_10 (x / ref))
The grid version is evaluated by numerically intgerating the
function over each bin using a non-adaptive Gauss-Kronrod scheme
suited for smooth functions [1]_, falling over to a simple
trapezoid scheme if this fails.
References
----------
.. [1] https://www.gnu.org/software/gsl/manual/html_node/QNG-non_002dadaptive-Gauss_002dKronrod-integration.html
"""
def __init__(self, name='logparabola'):
self.ref = Parameter(name, 'ref', 1, alwaysfrozen=True)
self.c1 = Parameter(name, 'c1', 1)
self.c2 = Parameter(name, 'c2', 1)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name, (self.ref, self.c1,
self.c2, self.ampl))
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.logparabola(p, *args, **kwargs)
_gfactor = numpy.sqrt(numpy.pi / (4 * numpy.log(2)))
[docs]
class NormGauss1D(RegriddableModel1D):
"""One-dimensional normalised gaussian function.
Attributes
----------
fwhm
The Full-Width Half Maximum of the gaussian. It is related to
the sigma value by: FWHM = sqrt(8 * log(2)) * sigma.
pos
The center of the gaussian.
ampl
The amplitude refers to the integral of the model over the
range -infinity to infinity.
See Also
--------
Gauss1D, NormGauss2D
Notes
-----
The functional form of the model for points is::
f(x) = ampl * exp(-4 * log(2) * (x - pos)^2 / fwhm^2)
----------------------------------------------
sqrt(pi / (4 * log(2))) * fwhm
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='normgauss1d'):
self.fwhm = Parameter(name, 'fwhm', 10, tinyval, hard_min=tinyval)
self.pos = Parameter(name, 'pos', 0)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name, (self.fwhm, self.pos, self.ampl))
[docs]
def get_center(self):
return (self.pos.val,)
[docs]
def set_center(self, pos, *args, **kwargs):
self.pos.set(pos)
[docs]
def guess(self, dep, *args, **kwargs):
ampl = guess_amplitude(dep, *args)
pos = get_position(dep, *args)
fwhm = guess_fwhm(dep, *args)
param_apply_limits(pos, self.pos, **kwargs)
param_apply_limits(fwhm, self.fwhm, **kwargs)
# Apply normalization factor to guessed amplitude
norm = numpy.sqrt(numpy.pi / _gfactor) * self.fwhm.val
for key in ampl.keys():
if ampl[key] is not None:
ampl[key] *= norm
param_apply_limits(ampl, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.ngauss1d(p, *args, **kwargs)
[docs]
class Poisson(RegriddableModel1D):
"""One-dimensional Poisson function.
A model expressing the ratio of two Poisson distributions of mean
mu, one for which the random variable is x, and the other for
which the random variable is equal to mu itself.
Attributes
----------
mean
The mean of the first distribution.
ampl
The amplitude of the model.
Notes
-----
The functional form of the model for points is::
f(x) = ampl * mean! exp((x - mean) * log(mean)) / x!
The grid version is evaluated by numerically intgerating the
function over each bin using a non-adaptive Gauss-Kronrod scheme
suited for smooth functions [1]_, falling over to a simple
trapezoid scheme if this fails.
References
----------
.. [1] https://www.gnu.org/software/gsl/manual/html_node/QNG-non_002dadaptive-Gauss_002dKronrod-integration.html
"""
def __init__(self, name='poisson'):
self.mean = Parameter(name, 'mean', 1, 1e-05, hard_min=tinyval)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name, (self.mean, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
pos = get_position(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
param_apply_limits(pos, self.mean, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.poisson(p, *args, **kwargs)
[docs]
class Polynom1D(RegriddableModel1D):
"""One-dimensional polynomial function of order 8.
The maximum order of the polynomial is 8. The default setting has
all parameters frozen except for ``c0``, which means that the
model acts as a constant.
Attributes
----------
c0
The constant term.
c1
The amplitude of the (x-offset) term.
c2
The amplitude of the (x-offset)^2 term.
c3
The amplitude of the (x-offset)^3 term.
c4
The amplitude of the (x-offset)^4 term.
c5
The amplitude of the (x-offset)^5 term.
c6
The amplitude of the (x-offset)^6 term.
c7
The amplitude of the (x-offset)^7 term.
c8
The amplitude of the (x-offset)^8 term.
offset
There is a degeneracy between ``c0`` and ``offset``, so it
is recommended that at least one of these remains frozen.
See Also
--------
PowLaw1D, Polynom2D
Notes
-----
The functional form of the model for points is::
f(x) = sum_(i=0)^(i=8) c_i * (x - offset)^i
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='polynom1d'):
pars = []
for i in range(9):
pars.append(Parameter(name, 'c%d' % i, 0, frozen=True))
pars[0].val = 1
pars[0].frozen = False
for p in pars:
setattr(self, p.name, p)
self.offset = Parameter(name, 'offset', 0, frozen=True)
pars.append(self.offset)
ArithmeticModel.__init__(self, name, pars)
[docs]
def guess(self, dep, *args, **kwargs):
xmin = args[0].min()
xmax = args[0].max()
ymin = dep.min()
ymax = dep.max()
dydx = (ymax - ymin) / (xmax - xmin)
dydx2 = (ymax - ymin) / ((xmax - xmin) * (xmax - xmin))
xlo = 0
if numpy.abs(xmin - 0) >= DBL_EPSILON:
xlo = -xmin
if xmin < 0:
xlo = xmin
xhi = 0
if numpy.abs(xmax - 0) >= DBL_EPSILON:
xhi = -xmax
if xmax > 0:
xhi = xmax
ylo = 0
if numpy.abs(ymin - 0) >= DBL_EPSILON:
ylo = -ymin
if ymin < 0:
ylo = ymin
yhi = 0
if numpy.abs(ymax - 0) >= DBL_EPSILON:
yhi = -ymax
if ymax > 0:
yhi = ymax
c0 = {'val': (ymax + ymin) / 2.0, 'min': ylo, 'max': yhi}
c1 = {'val': 0.0, 'min': -100 * dydx, 'max': 100 * dydx}
c2 = {'val': 0.0, 'min': -100 * dydx2, 'max': 100 * dydx2}
c3 = {'val': 0.0, 'min': ylo, 'max': yhi}
off = {'val': 0.0, 'min': xlo, 'max': xhi}
param_apply_limits(c0, self.c0, **kwargs)
param_apply_limits(c1, self.c1, **kwargs)
param_apply_limits(c2, self.c2, **kwargs)
param_apply_limits(c3, self.c3, **kwargs)
param_apply_limits(c3, self.c4, **kwargs)
param_apply_limits(c3, self.c5, **kwargs)
param_apply_limits(c3, self.c6, **kwargs)
param_apply_limits(c3, self.c7, **kwargs)
param_apply_limits(c3, self.c8, **kwargs)
param_apply_limits(off, self.offset, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.poly1d(p, *args, **kwargs)
[docs]
class PowLaw1D(RegriddableModel1D):
"""One-dimensional power-law function.
It is assumed that the independent axis is positive at all points.
Attributes
----------
gamma
The photon index of the power law.
ref
As the reference point is degenerate with the amplitude, the
``alwaysfrozen`` attribute of the reference point is set so
that it can not be varied during a fit.
ampl
The amplitude of the model.
See Also
--------
Polynom1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl * (x / ref)^(-gamma)
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='powlaw1d'):
self.gamma = Parameter(name, 'gamma', 1, -10, 10)
self.ref = Parameter(name, 'ref', 1, alwaysfrozen=True)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name, (self.gamma, self.ref, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
ref = guess_reference(self.ref.min, self.ref.max, *args)
param_apply_limits(ref, self.ref, **kwargs)
norm = guess_amplitude_at_ref(self.ref.val, dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
if kwargs['integrate']:
# avoid numerical issues with C pow() function close to zero,
# 0.0 +- ~1.e-14. PowLaw1D integrated has multiple calls to
# pow(X, 1.0 - gamma). So gamma values close to 1.0 +- 1.e-10
# should be be 1.0 to avoid errors propagating in the calculated
# model.
if sao_fcmp(p[0], 1.0, 1.e-10) == 0:
# pars { gamma, ref, ampl }
p[0] = 1.0
return _modelfcts.powlaw(p, *args, **kwargs)
[docs]
class Scale1D(Const1D):
"""A constant model for one-dimensional data.
This is a specialized form of the ``Const1D`` model where the
``integrate`` flag is turned off.
Attributes
----------
c0
The amplitude of the model.
See Also
--------
Box1D, Const1D, Delta1D, Scale2D, StepLo1D, StepHi1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl
and for an integrated grid it is::
f(xlo,xhi) = ampl * (xhi - xlo)
This later case will only be used if the ``integrate``
attribute is set to ``True``.
"""
def __init__(self, name='scale1d'):
Const1D.__init__(self, name)
self.integrate = False
[docs]
class Sin(RegriddableModel1D):
"""One-dimensional sine function.
Attributes
----------
period
The period of the sine, in units of the independent axis.
offset
The offset (related to the phase) of the sine.
ampl
The amplitude of the sine.
See Also
--------
Cos, Tan
Notes
-----
The functional form of the model for points is::
f(x) = ampl * sin (2 * pi * (x - offset) / period)
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='sin'):
self.period = Parameter(name, 'period', 1, 1e-10, 10, tinyval)
self.offset = Parameter(name, 'offset', 0, 0, hard_min=0)
self.ampl = Parameter(name, 'ampl', 1, 1e-05, hard_min=0)
ArithmeticModel.__init__(self, name,
(self.period, self.offset, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.sin(p, *args, **kwargs)
[docs]
class Sqrt(RegriddableModel1D):
"""One-dimensional square root function.
Attributes
----------
offset
The offset of the model.
ampl
The amplitude of the model.
See Also
--------
Exp, Exp10, Log, Log10, LogParabola
Notes
-----
The functional form of the model for points is::
f(x) = ampl * sqrt (x - offset)
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='sqrt'):
self.offset = Parameter(name, 'offset', 0)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name, (self.offset, self.ampl))
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.sqrt(p, *args, **kwargs)
[docs]
class StepHi1D(RegriddableModel1D):
"""One-dimensional step function.
The model is flat above ``xcut``, where it is set to the ``ampl``
parameter, and zero below this.
Attributes
----------
xcut
The position of the step.
ampl
The amplitude of the step.
See Also
--------
Box1D, Const1D, Delta1D, StepLo1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl if x >= xcut
= 0 otherwise
and for an integrated grid it is::
f(xlo,xhi) = ampl * (xhi - xlo) if xlo >= xcut
= ampl * (xhi - xcut) if xlo < xcut and xhi >= xcut
= 0 if xhi < xcut
"""
def __init__(self, name='stephi1d'):
self.xcut = Parameter(name, 'xcut', 0)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name, (self.xcut, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
cut = guess_bounds(args[0], False)
norm = guess_amplitude(dep, *args)
param_apply_limits(cut, self.xcut, **kwargs)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.stephi1d(p, *args, **kwargs)
[docs]
class StepLo1D(RegriddableModel1D):
"""One-dimensional step function.
The model is flat below ``xcut``, where it is set to the ``ampl``
parameter, and zero above this.
Attributes
----------
xcut
The position of the step.
ampl
The amplitude of the step.
See Also
--------
Box1D, Const1D, Delta1D, StepHi1D
Notes
-----
The functional form of the model for points is::
f(x) = ampl if x <= xcut
= 0 otherwise
and for an integrated grid it is::
f(xlo,xhi) = ampl * (xhi - xlo) if xhi <= xcut
= ampl * (xcut - xlo) if xlo <= xcut and xhi > xcut
= 0 if xlo > xcut
"""
def __init__(self, name='steplo1d'):
self.xcut = Parameter(name, 'xcut', 0)
self.ampl = Parameter(name, 'ampl', 1, 0)
ArithmeticModel.__init__(self, name, (self.xcut, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
cut = guess_bounds(args[0], False)
norm = guess_amplitude(dep, *args)
param_apply_limits(cut, self.xcut, **kwargs)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.steplo1d(p, *args, **kwargs)
[docs]
class Tan(RegriddableModel1D):
"""One-dimensional tan function.
Attributes
----------
period
The period of the tangent, in units of the independent axis.
offset
The offset (related to the phase) of the tangent.
ampl
The amplitude of the tangent.
See Also
--------
Cos, Sin
Notes
-----
The functional form of the model for points is::
f(x) = ampl * tan (2 * pi * (x - offset) / period)
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='tan'):
self.period = Parameter(name, 'period', 1, 1e-10, 10, tinyval)
self.offset = Parameter(name, 'offset', 0, 0, hard_min=0)
self.ampl = Parameter(name, 'ampl', 1, 1e-05, hard_min=0)
ArithmeticModel.__init__(self, name,
(self.period, self.offset, self.ampl))
[docs]
def guess(self, dep, *args, **kwargs):
norm = guess_amplitude(dep, *args)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
@modelCacher1d
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs1d(self, kwargs)
return _modelfcts.tan(p, *args, **kwargs)
[docs]
class Box2D(RegriddableModel2D):
"""Two-dimensional box function.
The model is flat between the limits, where it is set to the
``ampl`` parameter. Outside this range the model is zero.
Attributes
----------
xlow
The lower edge of the box (``x0`` axis).
xhi
The upper edge of the box (``x0`` axis).
ylow
The lower edge of the box (``x1`` axis).
yhi
The upper edge of the box (``x1`` axis).
ampl
The amplitude of the box.
See Also
--------
Box1D, Const2D, Delta2D
Notes
-----
The functional form of the model for points is::
f(x0,x1) = ampl if xlow <= x0 <= xhi
ylow <= x1 <= yhi
= 0 otherwise
and for an integrated grid it is::
f(x0lo,x0hi,x1lo,x1hi)
= 0 if x0hi <= xlow or x0lo >= xhi
or x1hi <= ylow or x1lo >= yhi
= ampl * g where g is the fraction of the pixel
that falls within the region
This behavior is different to how the amplitude is handled in
other models, such as ``Const2D``.
"""
def __init__(self, name='box2d'):
self.xlow = Parameter(name, 'xlow', 0)
self.xhi = Parameter(name, 'xhi', 0)
self.ylow = Parameter(name, 'ylow', 0)
self.yhi = Parameter(name, 'yhi', 0)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name,
(self.xlow, self.xhi, self.ylow, self.yhi,
self.ampl))
self.cache = 0
[docs]
def guess(self, dep, *args, **kwargs):
xlo, xhi = guess_bounds(args[0])
ylo, yhi = guess_bounds(args[1])
norm = guess_amplitude2d(dep, *args)
param_apply_limits(xlo, self.xlow, **kwargs)
param_apply_limits(xhi, self.xhi, **kwargs)
param_apply_limits(ylo, self.ylow, **kwargs)
param_apply_limits(yhi, self.yhi, **kwargs)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs2d(self, kwargs)
return _modelfcts.box2d(p, *args, **kwargs)
[docs]
class Const2D(RegriddableModel2D, Const):
"""A constant model for two-dimensional data.
Attributes
----------
c0
The amplitude of the model.
See Also
--------
Box2D, Const1D, Delta2D, Scale2D
Notes
-----
The functional form of the model for points is::
f(x0, x1) = ampl
and for an integrated grid it is::
f(x0lo, x1lo, x0hi, x1hi = ampl * (x0hi - x0lo)
* (x1hi - x1lo)
"""
def __init__(self, name='const2d'):
Const.__init__(self, name)
self.cache = 0
[docs]
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs2d(self, kwargs)
return _modelfcts.const2d(p, *args, **kwargs)
[docs]
class Scale2D(Const2D):
"""A constant model for two-dimensional data.
This is a specialized form of the ``Const2D`` model where the
``integrate`` flag is turned off.
Attributes
----------
c0
The amplitude of the model.
See Also
--------
Box2D, Const2D, Delta2D, Scale1D
Notes
-----
The functional form of the model for points is::
f(x0, x1) = ampl
and for an integrated grid it is::
f(x0lo, x1lo, x0hi, x1hi) = ampl * (x0hi - x0lo)
* (x1hi - x1lo)
This later case will only be used if the ``integrate``
attribute is set to ``True``.
"""
def __init__(self, name='scale2d'):
Const2D.__init__(self, name)
self.integrate = False
self.cache = 0
[docs]
class Delta2D(RegriddableModel2D):
"""Two-dimensional delta function.
The delta function model is only non-zero at a single point
(or pixel for integrated grids).
Attributes
----------
xpos
The coordinate of the signal on the x0 axis.
ypos
The coordinate of the signal on the x1 axis.
ampl
The amplitude.
See Also
--------
Box2D, Const2D, Delta1D
Notes
-----
The functional form of the model for points is::
f(x0, x1) = ampl if x0 == xpos and x1 == ypos
= 0 otherwise
and for an integrated grid it is::
f(x0lo, x1lo, x0hi, x1hi) = ampl if x0lo <= xpos <= x0hi
x1lo <= ypos <= x1hi
= 0 otherwise
This behavior is different to how the amplitude is handled in
other models, such as ``Const2D``.
"""
def __init__(self, name='delta2d'):
self.xpos = Parameter(name, 'xpos', 0)
self.ypos = Parameter(name, 'ypos', 0)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name, (self.xpos, self.ypos, self.ampl))
self.cache = 0
[docs]
def get_center(self):
return (self.xpos.val, self.ypos.val)
[docs]
def set_center(self, xpos, ypos, *args, **kwargs):
self.xpos.set(xpos)
self.ypos.set(ypos)
[docs]
def guess(self, dep, *args, **kwargs):
xpos, ypos = guess_position(dep, *args)
norm = guess_amplitude2d(dep, *args)
param_apply_limits(xpos, self.xpos, **kwargs)
param_apply_limits(ypos, self.ypos, **kwargs)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs2d(self, kwargs)
return _modelfcts.delta2d(p, *args, **kwargs)
[docs]
class Gauss2D(RegriddableModel2D):
"""Two-dimensional gaussian function.
Attributes
----------
fwhm
The Full-Width Half Maximum of the gaussian along the major
axis. It is related to the sigma value by:
FWHM = sqrt(8 * log(2)) * sigma.
xpos
The center of the gaussian on the x0 axis.
ypos
The center of the gaussian on the x1 axis.
ellip
The ellipticity of the gaussian.
theta
The angle of the major axis. It is in radians, measured
counter-clockwise from the X0 axis (i.e. the line X1=0).
ampl
The amplitude refers to the maximum peak of the model.
See Also
--------
Gauss1D, NormGauss2D, SigmaGauss2D
Notes
-----
The functional form of the model for points is::
f(x0,x1) = ampl * exp(-4 * log(2) * r(x0,x1)^2)
r(x0,x1)^2 = xoff(x0,x1)^2 * (1-ellip)^2 + yoff(x0,x1)^2
-------------------------------------------
fwhm^2 * (1-ellip)^2
xoff(x0,x1) = (x0 - xpos) * cos(theta) + (x1 - ypos) * sin(theta)
yoff(x0,x1) = (x1 - ypos) * cos(theta) - (x0 - xpos) * sin(theta)
The grid version is evaluated by adaptive multidimensional
integration scheme on hypercubes using cubature rules, based
on code from HIntLib ([1]_) and GSL ([2]_).
References
----------
.. [1] HIntLib - High-dimensional Integration Library
http://mint.sbg.ac.at/HIntLib/
.. [2] GSL - GNU Scientific Library
http://www.gnu.org/software/gsl/
"""
def __init__(self, name='gauss2d'):
self.fwhm = Parameter(name, 'fwhm', 10, tinyval, hard_min=tinyval)
self.xpos = Parameter(name, 'xpos', 0)
self.ypos = Parameter(name, 'ypos', 0)
self.ellip = Parameter(name, 'ellip', 0, 0, 0.999, 0, 0.9999,
frozen=True)
self.theta = Parameter(name, 'theta', 0, -2 * numpy.pi, 2 * numpy.pi,
-2 * numpy.pi, 4 * numpy.pi, 'radians',
frozen=True)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name,
(self.fwhm, self.xpos, self.ypos, self.ellip,
self.theta, self.ampl))
self.cache = 0
[docs]
def get_center(self):
return (self.xpos.val, self.ypos.val)
[docs]
def set_center(self, xpos, ypos, *args, **kwargs):
self.xpos.set(xpos)
self.ypos.set(ypos)
[docs]
def guess(self, dep, *args, **kwargs):
xpos, ypos = guess_position(dep, *args)
norm = guess_amplitude2d(dep, *args)
fwhm = guess_fwhm(dep, *args)
param_apply_limits(xpos, self.xpos, **kwargs)
param_apply_limits(ypos, self.ypos, **kwargs)
param_apply_limits(norm, self.ampl, **kwargs)
param_apply_limits(fwhm, self.fwhm, **kwargs)
[docs]
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs2d(self, kwargs)
return _modelfcts.gauss2d(p, *args, **kwargs)
[docs]
class SigmaGauss2D(Gauss2D):
"""Two-dimensional gaussian function (varying sigma).
Attributes
----------
sigma_a
The sigma of the gaussian along the major axis.
sigma_b
The sigma of the gaussian along the minor axis.
xpos
The center of the gaussian on the x0 axis.
ypos
The center of the gaussian on the x1 axis.
theta
The angle of the major axis. It is in radians, measured
counter-clockwise from the X0 axis (i.e. the line X1=0).
ampl
The amplitude refers to the maximum peak of the model.
See Also
--------
Gauss2D, NormGauss2D
Notes
-----
The functional form of the model for points is::
f(x0,x1) = ampl * exp(-r(x0,x1)^2 / 2)
r(x0,x1)^2 = xoff(x0,x1)^2 + yoff(x0,x1)^2
------------- -------------
sigma_a^2 sigma_b^2
xoff(x0,x1) = (x0 - xpos) * cos(theta) + (x1 - ypos) * sin(theta)
yoff(x0,x1) = (x1 - ypos) * cos(theta) - (x0 - xpos) * sin(theta)
The grid version is evaluated by adaptive multidimensional
integration scheme on hypercubes using cubature rules, based
on code from HIntLib ([1]_) and GSL ([2]_).
References
----------
.. [1] HIntLib - High-dimensional Integration Library
http://mint.sbg.ac.at/HIntLib/
.. [2] GSL - GNU Scientific Library
http://www.gnu.org/software/gsl/
"""
def __init__(self, name='sigmagauss2d'):
self.sigma_a = Parameter(name, 'sigma_a', 10, tinyval, hard_min=tinyval)
self.sigma_b = Parameter(name, 'sigma_b', 10, tinyval, hard_min=tinyval)
self.xpos = Parameter(name, 'xpos', 0)
self.ypos = Parameter(name, 'ypos', 0)
self.theta = \
Parameter(name, 'theta', 0, -2 * numpy.pi, 2 * numpy.pi,
-2 * numpy.pi, 4 * numpy.pi, 'radians', frozen=True)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name,
(self.sigma_a, self.sigma_b, self.xpos,
self.ypos, self.theta, self.ampl))
self.cache = 0
[docs]
def guess(self, dep, *args, **kwargs):
xpos, ypos = guess_position(dep, *args)
norm = guess_amplitude2d(dep, *args)
fwhm = guess_fwhm(dep, *args)
param_apply_limits(xpos, self.xpos, **kwargs)
param_apply_limits(ypos, self.ypos, **kwargs)
param_apply_limits(fwhm, self.sigma_a, **kwargs)
param_apply_limits(fwhm, self.sigma_b, **kwargs)
param_apply_limits(norm, self.ampl, **kwargs)
[docs]
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs2d(self, kwargs)
return _modelfcts.sigmagauss2d(p, *args, **kwargs)
[docs]
class NormGauss2D(RegriddableModel2D):
"""Two-dimensional normalised gaussian function.
Attributes
----------
fwhm
The Full-Width Half Maximum of the gaussian along the major
axis. It is related to the sigma value by:
FWHM = sqrt(8 * log(2)) * sigma.
xpos
The center of the gaussian on the x0 axis.
ypos
The center of the gaussian on the x1 axis.
ellip
The ellipticity of the gaussian.
theta
The angle of the major axis. It is in radians, measured
counter-clockwise from the X0 axis (i.e. the line X1=0).
ampl
The amplitude refers to the integral of the model over the
range -infinity to infinity for both axes.
See Also
--------
Gauss2D, NormGauss1D, SigmaGauss2D
Notes
-----
The functional form of the model for points is::
f(x0,x1) = 4 * log(2) * ampl * exp(-4 * log(2) * r(x0,x1)^2)
-------------------------------------------------
pi * fwhm * fwhm * sqrt(1 - ellip * ellip)
r(x0,x1)^2 = xoff(x0,x1)^2 * (1-ellip)^2 + yoff(x0,x1)^2
-------------------------------------------
fwhm^2 * (1-ellip)^2
xoff(x0,x1) = (x0 - xpos) * cos(theta) + (x1 - ypos) * sin(theta)
yoff(x0,x1) = (x1 - ypos) * cos(theta) - (x0 - xpos) * sin(theta)
The grid version is evaluated by adaptive multidimensional
integration scheme on hypercubes using cubature rules, based
on code from HIntLib ([1]_) and GSL ([2]_).
References
----------
.. [1] HIntLib - High-dimensional Integration Library
http://mint.sbg.ac.at/HIntLib/
.. [2] GSL - GNU Scientific Library
http://www.gnu.org/software/gsl/
"""
def __init__(self, name='normgauss2d'):
self.fwhm = Parameter(name, 'fwhm', 10, tinyval, hard_min=tinyval)
self.xpos = Parameter(name, 'xpos', 0)
self.ypos = Parameter(name, 'ypos', 0)
self.ellip = Parameter(name, 'ellip', 0, 0, 0.999, 0, 0.9999,
frozen=True)
self.theta = Parameter(name, 'theta', 0, -2 * numpy.pi, 2 * numpy.pi,
-2 * numpy.pi, 4 * numpy.pi, 'radians',
frozen=True)
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name,
(self.fwhm, self.xpos, self.ypos, self.ellip,
self.theta, self.ampl))
self.cache = 0
[docs]
def get_center(self):
return (self.xpos.val, self.ypos.val)
[docs]
def set_center(self, xpos, ypos, *args, **kwargs):
self.xpos.set(xpos)
self.ypos.set(ypos)
[docs]
def guess(self, dep, *args, **kwargs):
xpos, ypos = guess_position(dep, *args)
ampl = guess_amplitude2d(dep, *args)
fwhm = guess_fwhm(dep, *args)
param_apply_limits(xpos, self.xpos, **kwargs)
param_apply_limits(ypos, self.ypos, **kwargs)
param_apply_limits(fwhm, self.fwhm, **kwargs)
# Apply normalization factor to guessed amplitude
norm = (numpy.pi / _gfactor) * self.fwhm.val * self.fwhm.val * \
numpy.sqrt(1.0 - (self.ellip.val * self.ellip.val))
for key in ampl.keys():
if ampl[key] is not None:
ampl[key] *= norm
param_apply_limits(ampl, self.ampl, **kwargs)
[docs]
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs2d(self, kwargs)
return _modelfcts.ngauss2d(p, *args, **kwargs)
[docs]
class Polynom2D(RegriddableModel2D):
"""Two-dimensional polynomial function.
The maximum order of the polynomial is 2.
Attributes
----------
c
The constant term.
cy1
The coefficient for the x1 term.
cy2
The coefficient for the x1^2 term.
cx1
The coefficient for the x0 term.
cx1y1
The coefficient for the x0 x1 term.
cx1y2
The coefficient for the x0 x1^2 term.
cx2
The coefficient for the x0^2 term.
cx2y1
The coefficient for the x0^2 x1 term.
cx2y2
The coefficient for the x0^2 x1^2 term.
See Also
--------
Polynom1D
Notes
-----
The functional form of the model for points is::
f(x,x1) = c + cx1 * x0 + cx2 * x0^2 +
cy1 * x1 + cy2 * x1^2 +
cx1y1 * x0 * x1 +
cx1y2 * x0 * x1^2 +
cx2y1 * x0^2 * x1 +
cx2y2 * x0^2 * x1^2
and for an integrated grid it is the integral of this over
the bin.
"""
def __init__(self, name='polynom2d'):
self.c = Parameter(name, 'c', 1)
self.cx1 = Parameter(name, 'cx1', 0)
self.cx2 = Parameter(name, 'cx2', 0)
self.cy1 = Parameter(name, 'cy1', 0)
self.cy2 = Parameter(name, 'cy2', 0)
self.cx1y1 = Parameter(name, 'cx1y1', 0)
self.cx1y2 = Parameter(name, 'cx1y2', 0)
self.cx2y1 = Parameter(name, 'cx2y1', 0)
self.cx2y2 = Parameter(name, 'cx2y2', 0)
ArithmeticModel.__init__(self, name,
(self.c, self.cy1, self.cy2, self.cx1,
self.cx1y1, self.cx1y2, self.cx2,
self.cx2y1, self.cx2y2))
self.cache = 0
[docs]
def guess(self, dep, *args, **kwargs):
x0min = args[0].min()
x0max = args[0].max()
x1min = args[1].min()
x1max = args[1].max()
ymin = dep.min()
ymax = dep.max()
ylo = 0
if numpy.abs(ymin - 0) > DBL_EPSILON:
ylo = -ymin
if ymin < 0:
ylo = ymin
yhi = 0
if numpy.abs(ymax - 0) > DBL_EPSILON:
yhi = -ymax
if ymax > 0:
yhi = ymax
dydx0 = (ymax - ymin) / (x0max - x0min)
dydx1 = (ymax - ymin) / (x1max - x1min)
dyd2x0 = (ymax - ymin) / ((x0max - x0min) * (x0max - x0min))
dyd2x1 = (ymax - ymin) / ((x1max - x1min) * (x1max - x1min))
dydx0dx1 = (ymax - ymin) / ((x0max - x0min) * (x1max - x1min))
c = {'val': (ymax + ymin) / 2., 'min': ylo, 'max': yhi}
cx1 = {'val': 0., 'min': -100 * dydx0, 'max': 100 * dydx0}
cy1 = {'val': 0., 'min': -100 * dydx1, 'max': 100 * dydx1}
cx2 = {'val': 0., 'min': -100 * dyd2x0, 'max': 100 * dyd2x0}
cy2 = {'val': 0., 'min': -100 * dyd2x1, 'max': 100 * dyd2x1}
cx1y1 = {'val': 0., 'min': -100 * dydx0dx1, 'max': 100 * dydx0dx1}
c22 = {'val': 0., 'min': ylo, 'max': yhi}
param_apply_limits(c, self.c, **kwargs)
param_apply_limits(cx1, self.cx1, **kwargs)
param_apply_limits(cy1, self.cy1, **kwargs)
param_apply_limits(cx2, self.cx2, **kwargs)
param_apply_limits(cy2, self.cy2, **kwargs)
param_apply_limits(cx1y1, self.cx1y1, **kwargs)
param_apply_limits(c22, self.cx1y2, **kwargs)
param_apply_limits(c22, self.cx2y1, **kwargs)
param_apply_limits(c22, self.cx2y2, **kwargs)
[docs]
def calc(self, p, *args, **kwargs):
kwargs = clean_kwargs2d(self, kwargs)
return _modelfcts.poly2d(p, *args, **kwargs)
[docs]
class TableModel(ArithmeticModel):
"""Tabulated values are linearly scaled and may be interpolated.
The `load` method is used to read in the tabular data. There are
two different modes:
- both independent (x) and dependent (y) axes are set, in which
case model evaluation will interpolate the requested grid onto
the data, with the interpolation scheme controlled by the
`method` attribute. The `fold` method does not need to be
used.
- the x axis is set to `None`, in which case the grid used to
evaluate the model must have the same size as the y values, and
the `fold` method must be sent the data object to be fit
before a fit is made. In this case the `method` attribute is
ignored.
The model values - the y argument to `load` - are multiplied by
the `ampl` parameter of the model.
Attributes
----------
ampl
The linear scaling factor for the table values
Notes
-----
The model has `ndim` set to `None` as it can be used for any
dimension of data. However, this only makes sense for the second
mode, that is when the x argument to `load` is `None`. For the
first mode, when x is not `None`, the model evaluation only uses
the first component of the independent axes - so the low values
for a Data1DInt object or the x0 values for a Data1D object.
The model ignores the integrate setting.
See Also
--------
Const1D, Scale1D
Examples
--------
If the x array is given to the `load` call then the model can
interpolate from the requested grid onto the load data (the
default is linear interpolation):
>>> tm = TableModel()
>>> tm.load([10, 20, 25, 30], [14, 12, 17, 18])
>>> tm.ampl = 10
>>> tm([15, 20, 27])
array([130., 120., 174.])
The `method` attribute can be changed to select a different
interpolation scheme: it requires a function that accepts (xout,
xin, yin) and returns yout which are the interpolated values of
xin,yin onto the grid xout:
>>> from sherpa.utils import neville
>>> tm.method = neville
>>> tm([15, 20, 27])
array([ 90. , 120. , 182.16])
The model can be used for data when the independent axis is either
not useful - such as for an image mask - or the data does not have
a meaningful independent axis, as in the the independent variable
being an index for a star and the dependent axis is a property of
each star. In this case the x argument to `load` is set to `None`,
which means that no interpolation is used and that the `fold`
method must be used if the data has been filtered in any way, but
it's safest to always use it:
>>> from sherpa.models.basic import TableModel
>>> from sherpa.data import Data1D
>>> from sherpa.fit import Fit
>>> d = Data1D('data', [1, 2, 3, 4, 5], [1.2, .4, 2.2, .3, 1.])
>>> d.staterror = [.2, .2, .2, .2, .2]
>>> tm1 = TableModel('tabmodel')
>>> tm1.load(None, [.6, .2, 1.1, .2, .5])
>>> tm1.ampl.val
1.0
>>> tm1.fold(d)
>>> fit1 = Fit(d, tm1)
>>> res1 = fit1.fit()
>>> tm1.ampl.val
1.9894736842102083
In this case the `fold` method is necessary, to ensure that the
fit only uses the valid data bins:
>>> import numpy as np
>>> y = np.ma.masked_invalid([np.nan, np.nan, 2.2, .3, 1.])
>>> d = Data1D('data', [1, 2, 3, 4, 5], y)
>>> d.staterror = [.2, .2, .2, .2, .2]
>>> tm2 = TableModel('tabmodel')
>>> tm2.load(None, [.6, .2, 1.1, .2, .5])
>>> tm2.fold(d)
>>> tm2.ampl.val
1.0
>>> fit2 = Fit(d, tm2)
>>> res2 = fit2.fit()
>>> tm2.ampl.val
1.9866666666663104
The masking also holds if the notice or ignore method has been
used on the dataset:
>>> d = Data1D('data', [1, 2, 3, 4, 5], [1.2, .4, 2.2, .3, 1])
>>> d.staterror = [.2, .2, .2, .2, .2]
>>> d.ignore(xhi=2)
>>> tm3 = TableModel('tabmodel')
>>> tm3.load(None, [.6, .2, 1.1, .2, .5])
>>> tm3.fold(d)
>>> tm3.ampl.val
1.0
>>> fit3 = Fit(d, tm3)
>>> res = fit3.fit()
>>> tm3.ampl.val
1.9866666666663104
"""
@property
def method(self):
"""The interpolation method, used when x is not None in the load call.
The method argument is a function that accepts arguments (xout,
xin, yin) and returns the yout values from interpolating xout onto
(xin, yin). The default is linear interpolation
(sherpa.utils.linear_interp).
"""
return self._method
@method.setter
def method(self, val):
self._method = val
# as the method affects the cache, clear it (we could skip
# this if the method has not changed but is it worth it?)
#
self.cache_clear()
def __init__(self, name='tablemodel'):
# these attributes should remain somewhat private
# as not to conflict with user defined parameter names
#
# NOTE added: actually, we can not have user-defined names
# here thanks to the NoNewAttributesAfterInit base class.
#
self.__x = None
self.__y = None
self.__filtered_y = None
self.filename = None
self._method = linear_interp
self.ampl = Parameter(name, 'ampl', 1)
ArithmeticModel.__init__(self, name, (self.ampl,))
[docs]
def load(self, x, y):
"""Set the model values.
Parameters
----------
x, y : None or sequence
The model values. It is expected that either both are
given, and have the same number of elements, or that only y
is set, although the model can be cleared by setting both
to None. If x is given then the data is saved after being
sorted into increasing order of x.
See Also
--------
get_x, get_y
"""
if x is not None and y is None:
raise ModelErr("y must be set if x is set")
# Clear the cache. We could avoid doing this if the
# data has not changed but this is not worth the
# complexity.
#
self.cache_clear()
# A simplified version of sherpa.data._check, which is not
# used here to avoid circular dependencies. Rather than raise
# DataErr, we raise ModelErr with a similar message.
#
def _check(val):
if val is None:
return None
val = numpy.asarray(val)
if val.ndim != 1:
raise ModelErr("Array must be 1D or None")
# Check we can add 0 to the values. This is to try and
# catch cases when a string column is passed to load. It
# will not catch all possible problem cases. An
# alternative would be to check
# isinstance(val[0], numbers.Number)
# but it's not clear which is best.
#
# We could also check whether values are np.isfinite() but
# users may want to read in bad values that we then always
# mask out.
#
try:
val + 0
except Exception:
raise ModelErr(f"Unable to treat array as numeric: {val}") from None
return val
self.__y = _check(y)
self.__x = _check(x)
# clear the filtered array
self.__filtered_y = None
if x is None:
return
nx = len(self.__x)
ny = len(self.__y)
if nx != ny:
raise ModelErr(f"size mismatch between x and y: {nx} vs {ny}")
# Ensure the data is sorted.
#
idx = self.__x.argsort()
self.__y = self.__y[idx]
self.__x = self.__x[idx]
[docs]
def get_x(self):
"""Return the independent axis or None.
See Also
--------
get_y, load
"""
return self.__x
[docs]
def get_y(self):
"""Return the dependent axis or None.
See Also
--------
get_x, load
"""
return self.__y
[docs]
def fold(self, data):
"""Ensure the model matches the data filter.
This should be called after load, to ensure that any existing
filter is applied correctly during a fit. It is only necessary
for models where x is None in the load call.
Parameters
----------
data : sherpa.data.Data instance
An object with a mask attribute.
"""
if self.__y is None:
raise ModelErr("The tablemodel's load method must be called first")
# Clear out the setting. If needed it will get reset.
#
self.__filtered_y = None
# If we are interpolating the data we do not care about the
# data mask.
#
if self.__x is not None:
return
# What should we do with data.mask = {True, False}?
#
mask = data.mask
if not numpy.iterable(mask):
return
# At this point we know mask is an iterable, so it should
# match the y data of the model.
#
if len(mask) != len(self.__y):
raise ModelErr("filtermismatch", 'table model',
f"data, ({len(self.__y)} vs {len(mask)})")
self.__filtered_y = self.__y[mask]
[docs]
@modelCacher1d
def calc(self, p, x0, x1=None, *args, **kwargs):
"""Evaluate the model.
The load method must have been called first. If both x and y
were given then the model is interpolated onto the x0 grid,
otherwise x0 is only checked to see if it has the right size
(fold should have been called if the data has been filtered).
"""
if self.__y is None:
raise ModelErr("The tablemodel's load method must be called first")
if self.__x is not None:
return p[0] * interpolate(x0, self.__x, self.__y, function=self.method)
if (self.__filtered_y is not None and
len(x0) == len(self.__filtered_y)):
return p[0] * self.__filtered_y
if len(x0) == len(self.__y):
return p[0] * self.__y
raise ModelErr("filtermismatch", 'table model',
f"data, ({len(self.__y)} vs {len(x0)})")
[docs]
class UserModel(ArithmeticModel):
"""Support for user-supplied models.
The class is used by sherpa.ui.load_user_model and
sherpa.astro.ui.load_user_model.
"""
def __init__(self, name='usermodel', pars=None):
# these attributes should remain somewhat private
# as not to conflict with user defined parameter names
#
# NOTE added: actually, we can not have user-defined names
# here thanks to the NoNewAttributesAfterInit base class.
#
self._y = []
self._file = None
if pars is None:
self.ampl = Parameter(name, 'ampl', 1)
pars = (self.ampl,)
else:
for par in pars:
setattr(self, par.name, par)
ArithmeticModel.__init__(self, name, pars)
class Integrator1D(CompositeModel, RegriddableModel1D):
"""Integrate a one-dimensional model.
It is expected that instances of this class are created by the
`Integrate1D` class and not directly.
Attributes
----------
model
The model to be evaluated.
otherargs
Currently unused.
otherkwargs
Used to pass extra parameters to the integrator (currently
`epsabs`, `epsrel`, `maxeval`, `errflag`, and `logger`).
Raises
------
sherpa.utils.err.ModelErr
The model was evaluated on a "point" grid.
See Also
--------
Integrate1D
Examples
--------
Numericaly integrate the Gauss1D model across the bins defined by
xlo and xhi, using an absolute tolerance of 1e-5. Note that you
would not use this in practice, since the Gauss1D model already
supports integrated bins, but it does allow us to compare the two
approaches (the `y` and `ytrue` arrays).
>>> import numpy as np
>>> from sherpa.models.basic import Gauss1D, Integrator1D
>>> gmdl = Gauss1D()
>>> imdl = Integrator1D(gmdl, epsabs=1e-5)
>>> xgrid = np.asarray([0, 0.5, 1, 2, 4])
>>> xlo = xgrid[:-1]
>>> xhi = xgrid[1:]
>>> y = imdl(xlo, xhi)
>>> ytrue = gmdl(xlo, xhi)
>>> (y - ytrue) / ytrue
array([-1.11278878e-16, 0.00000000e+00, 0.00000000e+00, -1.43150678e-16])
"""
@staticmethod
def wrapobj(obj):
if isinstance(obj, ArithmeticModel):
return obj
return ArithmeticFunctionModel(obj)
def __init__(self, model, *otherargs, **otherkwargs):
self.model = self.wrapobj(model)
self.otherargs = otherargs
self.otherkwargs = otherkwargs
self._errflag = 0
CompositeModel.__init__(self,
f'integrate1d({self.model.name})',
(self.model,))
def startup(self, cache=False):
self.model.startup(cache)
self._errflag = 1
CompositeModel.startup(self, cache)
def teardown(self):
self.model.teardown()
CompositeModel.teardown(self)
def calc(self, p, xlo, xhi=None, **kwargs):
if xhi is None:
raise ModelErr('needsint')
return _modelfcts.integrate1d(self.model.calc,
p, xlo, xhi, **self.otherkwargs)
# DOC note: we do not expose Integrator1D by default so it is not
# mentioned as a See Also link here.
#
[docs]
class Integrate1D(RegriddableModel1D):
"""Integrate a model across each bin (one dimensional).
Numerically integrate a one-dimensional model across each bin of a
"histogram" dataset - that is one with low and high edges for each
bin. The model to be integrated is supplied as an input to this
model and any attribute changes must be made before this.
Attributes
----------
epsabs
The maximum absolute difference allowed when integrating the
model. This parameter is always frozen.
epsrel
The maximum relative difference allowed when integrating the
model. This parameter is always frozen.
maxeval
The maximum number of iterations allowed per bin. This
parameter is always frozen.
Notes
-----
If `imdl` is an instance of `Integrate1D` and `omdl` the model to
integrate, then `imdl(omdl)` creates the integrated form. Note
that changes to the `Integrate1D` parameters - such as `epsabs` -
must be made before creating this integrated form. For example, to
change the absolute tolerance to a value appropriate for 32-bit
floats:
>>> import numpy as np
>>> omdl = Polynom1D() # note: this class already supports integration
>>> imdl = Integrate1D(name='imdl')
>>> imdl.epsabs = np.finfo(np.float32).eps
>>> mdl = imdl(omdl)
Examples
--------
The Integrate1D model lets you use a one-dimensional model which
can only be evaluated at a point in a case where the dataset has a
low and high edge for the independent axis - such as a Data1DInt
object. The Scale1D model is used as an example of a model which
only evaluates at a point (as it always returns the scale value),
but in actual code you should use the Const1D model rather than
apply Integrate1D to Scale1D. The evaluation of the `smdl`
instance returns the `c0` parameter value (set to 10) at each
point, whereas the `mdl` instance integrates this across each bin,
and so returning the bin width times `c0` for each bin. Note that
the tolerance for the integration:
- was changed from the default (tolerance for 64-bit float) to
the 32-bit float tolerance (to avoid a warning message when
evaluating mdl);
- and must be changed before being applied to the model to
integrate (smdl) in this case.
>>> import numpy as np
>>> from sherpa.models.basic import Scale1D, Integrate1D
>>> xlo, xhi = [2, 5, 8], [4, 8, 12]
>>> imdl = Integrate1D(name='imdl')
>>> imdl.epsabs = np.finfo(np.float32).eps
>>> smdl = Scale1D(name='smdl')
>>> mdl = imdl(smdl)
>>> smdl.c0 = 10
>>> print(mdl)
integrate1d(smdl)
Param Type Value Min Max Units
----- ---- ----- --- --- -----
smdl.c0 thawed 10 -3.40282e+38 3.40282e+38
>>> print(smdl(xlo, xhi))
[10. 10. 10.]
>>> print(mdl(xlo, xhi))
[20. 30. 40.]
"""
def __init__(self, name='integrate1d'):
tol = DBL_EPSILON
self.epsabs = Parameter(name, 'epsabs', tol, alwaysfrozen=True)
self.epsrel = Parameter(name, 'epsrel', 0, alwaysfrozen=True)
self.maxeval = Parameter(name, 'maxeval', 10000, alwaysfrozen=True)
ArithmeticModel.__init__(self, name, (self.epsabs, self.epsrel,
self.maxeval))
def __call__(self, model):
return Integrator1D(model,
epsabs=self.epsabs.val,
epsrel=self.epsrel.val,
maxeval=int(self.maxeval.val),
logger=warning)