def __init__(self, alpha, beta, K):
"""
Note alpha/beta = mean, alpha/beta^2 = variance
@param alpha:
shape parameter
@param beta:
inverse scale parameter
@param K:
number of bins
"""
if alpha <= 0:
raise Exception("alpha = %f <= 0" % alpha)
if beta <= 0:
raise Exception("beta = %f <= 0" % beta)
if K < 1:
raise Exception("Num bins = %d < 1" % K)
# find upper boundaries of each bin
max_prob = gammainc(alpha, self.MAX_RATE)
bin_prob = max_prob / K
targets = np.arange(bin_prob, max_prob+bin_prob/2, bin_prob)
#XXX: not sure why we don't divide targets / beta
bin_ubounds = gammaincinv(alpha, targets) / beta
bin_lbounds = np.zeros(K)
bin_lbounds[1:] = bin_ubounds[:-1]
tmp = gammainc(alpha+1, bin_ubounds * beta) - gammainc(alpha+1, bin_lbounds * beta)
bin_rates = tmp * alpha / beta * K
# Rate of middle of each bin
self.bin_rates = bin_rates
# Probability mass of each bin
self.bin_probs = np.zeros(K) + bin_prob
def truncated_gamma(shape=None, alpha=1., beta=1., x_min=None, x_max=None):
"""
Generate random variates from a lower-and upper-bounded gamma distribution.
@param shape: shape of the random sample
@param alpha: shape parameter (alpha > 0.)
@param beta: scale parameter (beta >= 0.)
@param x_min: lower bound of variate
@param x_max: upper bound of variate
@return: random variates of lower-bounded gamma distribution
"""
from scipy.special import gammainc, gammaincinv
from numpy.random import gamma
from numpy import inf
if x_min is None and x_max is None:
return gamma(alpha, 1 / beta, shape)
elif x_min is None:
x_min = 0.
elif x_max is None:
x_max = inf
x_min = max(0., x_min)
x_max = min(1e300, x_max)
a = gammainc(alpha, beta * x_min)
b = gammainc(alpha, beta * x_max)
return probability_transform(shape,
lambda x, alpha=alpha: gammaincinv(alpha, x),
a, b) / beta
def integrate(self,lower,upper,method=None,**kwargs):
"""
Analytically Compute Schechter integral using incomplete gamma
functions. If `method` is not None, numerical integration will be used.
The gamma functions break down for alpha<=-1, so numerical is used if
that is the case.
"""
if self.alpha<=-1:
method = True #use default numerical method, because gamma functions fail for alpha<=-1
if method is not None:
return FunctionModel1D.integrate(self,lower,upper,method,**kwargs)
from scipy.special import gamma,gammainc,gammaincc
s = self.alpha+1
u = upper/self.Lstar
l = lower/self.Lstar
if upper==np.inf and lower<=0:
I = gamma(s)
elif upper==np.inf:
I = gammaincc(s,l)*gamma(s)
elif lower==0:
I = gammainc(s,u)*gamma(s)
else:
I = (gammainc(s,u) - gammainc(s,l))*gamma(s)
return self.phistar*I
def dldtheta(self,dat):
"""
Evaluates the gradient of the Gamma function with respect to the primary parameters.
:param dat: Data on which the gradient should be evaluated.
:type dat: DataModule.Data
:returns: The gradient
:rtype: numpy.array
"""
m = dat.numex()
grad = zeros((len(self.primary),m))
s = self.param['s']
p = self.param['p']
u = array([self.param['a'],self.param['b']])
cab = .5 + 0.5*sign(u)*gammainc(1/p,abs(u)**p /s)
for ind,key in enumerate(self.primary):
if key == 's':
U = -.5*sign(u)*(abs(u) *exp(-abs(u)**p/s))/(s**(1/p+1)*gammafunc(1/p))
grad[ind,:] = -1.0/p/s + abs(squeeze(dat.X))**p/s**2 - (U[1]-U[0])/(cab[1]-cab[0])
if key == 'p':
f = lambda x: 1/x
df = lambda x: -1/x**2
g = lambda x,k: abs(u[k])**x/s
dg = lambda x,k: abs(u[k])**x*log(abs(u[k]))/s
dIncGamma = array([totalDerivativeOfIncGamma(p,f,lambda v: g(v,0),df,lambda v: dg(v,0)),\
totalDerivativeOfIncGamma(p,f,lambda v: g(v,1),df,lambda v: dg(v,1))])
U = .5*sign(u) * (dIncGamma/float(gammafunc(1/p)) + gammainc(1/p,abs(u)**p/s)*digamma(1/p)/p**2)
grad[ind,:] = 1/p + 1/p**2*log(s) + digamma(1/p)*1/p**2 - abs(squeeze(dat.X))**p*log(abs(squeeze(dat.X)))/s - (U[1]-U[0])/(cab[1]-cab[0])
return grad
def new_BD_ratio(P, zp): MB, ReB, nB, M1, M2, h1, h2, R_brk = P.values I01, I02, IeB = convert_I(M1, zp), convert_I(M2, zp), convert_I(MB, zp) disc1 = I01 * h1 * h1 * gammainc(2., R_brk / h1) disc2 = I02 * h2 * h2 * (1 - gammainc(2., R_brk / h2)) bn = get_b_n(nB) bulge = np.exp(bn) * IeB * ReB * ReB * gamma((2 * nB) + 1) / (bn**(2.*nB)) return bulge / (2* (disc2 + disc1))
def dndx_rvir(Lrng=(0.001, 10), nL=1000, beta=0.2, rvir_Lstar=250.*u.kpc,
phi_str_pref = 1.49, alpha = -1.05, Mstar = -20.44,
cosmo=None): # ; M* - 5 log h
""" Estimate dN/dX for a set of CGM assuming unit covering fraction
Following Prochaska+11
Use beta=0 and rvir_Lstar=300kpc for a constant CGM to 300kpc
Parameters
----------
Lrng : tuple, optional
Range to integrate luminosity in terms of L*
nL : int, optional
Number of evaluations of L in that interval
beta : float, optional
Parameterization of rvir with L
r_vir = 250 kpc * (L/L*)^beta
phi_str, alpha, Mstar : float, float, float
Blanton lum function
Phi(M) = 0.4 log(10) Phi* 10^(-0.4 [M-M*][alpha+1]) exp(-
; 10^(-0.4[M-M*] ) )
Phi(L) = Phi* (L/L*)^alpha exp(-L/L*)
Phi* has the funny units of 1e-2 h^3 Mpc^-3
cosmo : Cosmology, optional
Defaults to Planck15
Returns
-------
Lval : ndarray
Luminosities of evaluation
dNdX : float
Cumulative dNdX
"""
# Cosmology
if cosmo is None:
cosmo = Planck15
hubb = cosmo.H0.value / 100.
# Constants
phi_str_cgs = (phi_str_pref * 1e-2 * hubb**3) / u.Mpc**3
dndx_const = (const.c / cosmo.H0).cgs
# Cumulative
Lval = np.linspace(Lrng[0], Lrng[1], nL)
x = alpha + 1 + beta
# Integrate
if gflg:
igmma = np.zeros_like(Lval)
i0 = float(gammainc(x,Lrng[1]))
for kk,iLval in enumerate(Lval):
igmma[kk] = i0 - float(gammainc(x,iLval))
else:
igmma = gammainc(x,Lrng[1]) - gammainc(x,Lval)
dNdx_rvir = (dndx_const * phi_str_cgs * (np.pi * rvir_Lstar**2) * (
gamma(x) * igmma)).decompose().value
# Return
return Lval, dNdx_rvir
def getMass(self,sigCrit=1.):
import numpy
from math import pi
from scipy.special import gammainc,gamma
b,n,re = self.b,self.n,self.re
k = 2.*n-1./3+4./(405.*n)+46/(25515.*n**2)
amp = b*b*k**(2*n)/(2*n*re**2*(gamma(2*n)*gammainc(2*n,k*(b/re)**(1/n))))
return 2*pi*sigCrit*re**2*amp*gamma(2*n)*n/k**(2*n)
###
amp = b*b*k**(2*n)/(numpy.exp(k)*2*n*re**2*(gamma(2*n)*gammainc(2*n,k*(b/re)**(1/n))))
return 2*pi*sigCrit*re**2*amp*gamma(2*n)*n/(numpy.exp(k)*k**(2*n))
def cdf(self,dat):
'''
Evaluates the cumulative distribution function on the data points in dat.
:param dat: Data points for which the c.d.f. will be computed.
:type dat: natter.DataModule.Data
:returns: A numpy array containing the probabilities.
:rtype: numpy.array
'''
u = array([self.param['a'],self.param['b']])
u = .5 + 0.5*sign(u)*gammainc(1/self.param['p'],abs(u)**self.param['p'] / self.param['s'])
return squeeze(.5 + 0.5*sign(dat.X)*gammainc(1/self.param['p'],abs(dat.X)**self.param['p'] / self.param['s']) - u[0])/(u[1]-u[0])
def sfr_avg(self, tage=None, dt=0.1):
"""
The average SFR between time ``tage``-``dt`` and ``tage``, given the
SFH parameters, for ``sfh=1`` or ``sfh=4``. Like SFHSTAT in FSPS.
Requires scipy, as it uses gamma functions.
:param tage: (default, None)
The ages (in Gyr) at which the average SFR over the last ``dt`` is
desired. if ``None``, uses the current value of the ``tage`` in
the parameter set. Scalar or iterable.
:param dt: (default: 0.1)
The time interval over which the recent SFR is averaged, in Gyr.
defaults to 100 Myr (i.e. sfr8).
:returns sfr_avg:
The SFR at ``tage`` averaged over the last ``dt`` Gyr, in units of
solar masses per year.
"""
from scipy.special import gammainc
assert self.params['sf_trunc'] <= 0, \
"sfr_avg not supported for sf_trunc > 0"
if self.params['sfh'] == 1:
power = 1
elif self.params['sfh'] == 4:
power = 2
else:
raise ValueError("sfr_avg not supported for this SFH type.")
tau, sf_start = self.params['tau'], self.params['sf_start']
if tage is None:
tage = np.atleast_1d(self.params['tage'])
else:
tage = np.atleast_1d(tage)
if tage[0] <= 0:
tage = 10**(self.log_age - 9)
tb = (self.params['tburst'] - sf_start) / tau
tmax = (tage[-1] - sf_start) / tau
normalized_times = (np.array([tage, tage - dt]).T - sf_start) / tau
tau_mass_frac = gammainc(power, normalized_times) / gammainc(power, tmax)
burst_in_past = tb <= normalized_times
mass = (tau_mass_frac * (1 - self.params['const'] - self.params['fburst']) +
self.params['const'] * normalized_times / tmax +
burst_in_past * self.params['fburst'])
avsfr = (mass[..., 0] - mass[..., 1]) / dt / 1e9 # Msun/yr
return np.clip(avsfr, 0, np.inf)
def _getInterpProbs(self, n, log_s):
#These values come from:
#http://ned.ipac.caltech.edu/level5/March05/Graham/Graham2.html
b_n = 1.9992*n-0.3271
x = b_n*numpy.power(10,log_s)**(1./n)
probs = gammainc(2*n, x)
return interp1d(probs, log_s), numpy.min(probs), numpy.max(probs)
def det_tau_r(E_f, V_f, m_list, mu_xi, dp_tau, dp_r, tau_range, r_range, sV0):
w_arr = np.linspace(0, 20, 300)
tau_arr = np.linspace(tau_range[0], tau_range[1], dp_tau)
r_arr = np.linspace(r_range[0], r_range[1], dp_tau)
e_arr = orthogonalize([tau_arr, r_arr])
x_axis = e_arr[0]
y_axis = e_arr[1]
for mi in m_list:
res_array = np.zeros((dp_tau, dp_r))
for i_tau, taui in enumerate(tau_arr):
for i_r, ri in enumerate(r_arr):
s = get_scale(mu_xi, mi, taui, ri)
print s
#sV0 = float(s * (pi * mu_r ** 2) ** (1. / mi))
T = 2. * taui / ri
s0 = ((T * (mi + 1) * sV0 ** mi) / (2. * E_f * pi * ri ** 2)) ** (1. / (mi + 1))
print s0
k = np.sqrt(T / E_f)
ef0 = k * np.sqrt(w_arr)
G = 1 - np.exp(-(ef0 / s0) ** (mi + 1))
mu_int = ef0 * E_f * V_f * (1 - G)
I = s0 * gamma(1 + 1. / (mi + 1)) * gammainc(1 + 1. / (mi + 1), (ef0 / s0) ** (mi + 1))
mu_broken = E_f * V_f * I / (mi + 1)
result = mu_int + mu_broken
sigma_c = np.max(result)
if sigma_c == result[-1]:
print "w_arr too short"
pass
res_array[i_tau, i_r] = sigma_c
mlab.surf(x_axis, y_axis, res_array / 100.)
mlab.xlabel("det tau")
mlab.ylabel("det r")
mlab.zlabel("sigma")
mlab.show()
def _rhov2int(self, phi, r, ra):
"""Compute dimensionless pressure integral for phi, r """
# Isotropic case first (equation 9, GZ15)
rhov2r = exp(phi)*gammainc(self.g + 2.5, phi)
rhov2 = 3*rhov2r
rhov2t = 2*rhov2r
# Add anisotropy, add parts depending explicitly on r
# (see equations 12, 13, and 14 of GZ15)
if (ra < self.ramax) and (r>0) and (phi>0):
p, g = r/ra, self.g
p2 = p**2
p12 = 1+p2
g3, g5, g7, fp2 = g+1.5, g+2.5, g+3.5, phi*p2
P1 = p2*phi**g5/gamma(g7)
H1 = hyp1f1(1, g7, -fp2) if fp2 <self.max_arg_exp else g5/fp2
H2 = hyp1f1(2, g7, -fp2) if fp2 <self.max_arg_exp else g5*g3/fp2**2
rhov2r += P1*H1
rhov2r /= p12
rhov2t /= p12
rhov2t += 2*P1*(H1/p12 + H2)
rhov2t /= p12
rhov2 = rhov2r + rhov2t
return rhov2, rhov2r, rhov2t
def deterministic_r():
E_f, V_f, tau, m, sV0 = 200e3, 0.01, .1, 7.0, 3.e-3
for ri in [0.005, 0.01, 0.015]:
T = 2. * tau / ri
w_arr = np.linspace(0,1.2,1000)
# analytical solution
s0 = ((T * (m+1) * sV0**m)/(2. * E_f * pi * ri ** 2))**(1./(m+1))
k = np.sqrt(T/E_f)
ef0 = k*np.sqrt(w_arr)
G = 1 - np.exp(-(ef0/s0)**(m+1))
mu_int = ef0 * E_f * V_f * (1-G)
I = s0 * gamma(1 + 1./(m+1)) * gammainc(1 + 1./(m+1), (ef0/s0)**(m+1))
mu_broken = E_f * V_f * I / (m+1)
plt.plot(w_arr, mu_int + mu_broken, lw = 2, color = 'black')
plt.ylim(0, 16)
plt.xlim(0, 1.25)
plt.show()
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax2 = ax1.twinx()
for mi in [4.0, 5.0, 7.0]:
r_arr = np.linspace(0.001, 0.01, 100)
T = 2. * tau / r_arr
s0 = ((T * (mi+1) * sV0**mi)/(2. * E_f * pi * r_arr ** 2))**(1./(mi+1))
wstar = E_f / T * s0**2 * mi**(-2./(mi+1))
zeta = mi**(-1./(mi+1.)) * np.exp(-1./mi) + 1. / (mi + 1.) * gamma(1 + 1./(mi+1)) * gammainc(1 + 1./(mi+1), 1./mi)
strength = E_f * V_f * s0 * zeta
ax1.loglog(r_arr, strength, lw=2, color='black')
ax1.set_ylim(10,100)
ax2.loglog(r_arr, wstar, lw=2, color='black', ls='dashed')
ax2.set_ylim(0.1,1)
plt.xlim(0,0.01)
plt.show()
def int_sfh_2(t1, t2, tage, tau, sf_start, tburst=0, fburst=0):
"""Use Gamma functions
"""
normalized_times = (np.array([t1, t2, tage]) - sf_start) / tau
mass = gammainc(2, normalized_times)
intsfr = (mass[1] - mass[0]) / mass[2]
return intsfr
def deflections(self, xin, yin):
import numpy
from math import cos, sin, pi
from scipy.special import gammainc
x, y = self.align_coords(xin, yin)
q = self.q
if q == 1.:
q = 1. - 1e-7 # Avoid divide-by-zero errors
b, n, re = self.b, self.n, self.re
k = 2. * n - 1. / 3 + 4. / (405. * n) + 46 / (25515. * n**2)
amp = (b / re)**2 / gammainc(2 * n, k * (b / re)**(1 / n))
r0 = re / q**0.5
o = numpy.ones(x.size).astype(x.dtype)
eval = numpy.array(
[abs(x).ravel() / r0,
abs(y).ravel() / r0, n * o, q * o]).T
xout = numpy.sign(x) * (amp * self.xmod.eval(eval) * r0).reshape(
x.shape)
yout = numpy.sign(y) * (amp * self.ymod.eval(eval) * r0).reshape(
y.shape)
theta = -(self.theta - pi / 2.)
ctheta = cos(theta)
stheta = sin(theta)
x = xout * ctheta + yout * stheta
y = yout * ctheta - xout * stheta
return x, y
def deterministic_tau():
E_f, V_f, r, m, sV0 = 200e3, 0.01, 0.01, 7., 3.e-3
w_arr = np.linspace(0, 1.2, 300)
for taui in [0.05, .1, .2]:
T = 2. * taui / r
s0 = ((T * (m+1) * sV0**m)/(2. * E_f * pi * r ** 2))**(1./(m+1))
k = np.sqrt(T/E_f)
ef0 = k*np.sqrt(w_arr)
G = 1 - np.exp(-(ef0/s0)**(m+1))
mu_int = ef0 * E_f * V_f * (1-G)
I = s0 * gamma(1 + 1./(m+1)) * gammainc(1 + 1./(m+1), (ef0/s0)**(m+1))
mu_broken = E_f * V_f * I / (m+1)
plt.plot(w_arr, mu_int + mu_broken, lw = 2, color = 'black')
plt.ylim(0, 16)
plt.xlim(0, 1.25)
plt.show()
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax2 = ax1.twinx()
for mi in [4.0, 5.0, 7.0]:
tau_arr = np.linspace(0.05, .2, 5)
T = 2. * tau_arr / r
s0 = ((T * (mi+1) * sV0**mi)/(2. * E_f * pi * r ** 2))**(1./(mi+1))
wstar = E_f / T * s0**2 * mi**(-2./(mi+1))
zeta = mi**(-1./(mi+1.)) * np.exp(-1./mi) + 1. / (mi + 1.) * gamma(1 + 1./(mi+1)) * gammainc(1 + 1./(mi+1), 1./mi)
strength = E_f * V_f * s0 * zeta
ax1.loglog(tau_arr, strength, lw=2, color='black')
ax1.set_ylim(0,15)
ax2.loglog(tau_arr, wstar, lw=2, color='black', ls='dashed')
ax2.set_ylim(0.1,1.)
plt.xlim(0.05,.2)
plt.ylim(0)
plt.show()
def _draw_gamma_rates(self):
'''
Function to draw and assign rates from a discretized gamma distribution, if specified. By default, 4 categories are drawn.
'''
if self.rate_probs is not None:
print("\nThe provided value for the `rate_probs` argument will be ignored since gamma-distributed heterogeneity has been specified with the alpha parameter.")
if type(self.k_gamma) is not int:
raise TypeError("\nProvided argument `num_categories` must be an integer.")
#### Note that this code is adapted from gamma.c in PAML ####
rv = gamma(self.alpha, scale = 1./self.alpha)
freqK = np.zeros(self.k_gamma) ### probs
rK = np.zeros(self.k_gamma) ### rates
for i in range(self.k_gamma-1):
raw=rv.ppf( (i+1.)/self.k_gamma )
freqK[i] = gammainc(self.alpha + 1, raw*self.alpha)
rK[0] = freqK[0] * self.k_gamma
rK[self.k_gamma-1] = (1-freqK[self.k_gamma-2]) * self.k_gamma
for i in range(1,self.k_gamma-1):
rK[i] = self.k_gamma * (freqK[i] -freqK[i-1])
#############################################################
self.rate_probs = np.repeat(1./self.k_gamma, self.k_gamma)
self.rate_factors = deepcopy(rK)
if self.pinv > ZERO:
self.rate_probs = list(self.rate_probs - self.pinv/self.k_gamma) + [self.pinv]
self.rate_factors = list(self.rate_factors) + [0.0]
self.rate_probs = np.array( self.rate_probs )
self.rate_factors = np.array( self.rate_factors )
def cdf(self,x,nu=None,var_2=None,var_3=None,var_4=None): """cumulative distribution function """ lamb = 0.5 k = nu*.5 P = spec.gammainc(k,lamb*x) return P
def integrate_sfh_ben(t1, t2, tage, tau, sf_start, tburst=0, fburst=0):
normalized_times = (np.array([t1, t2, tage]) - sf_start) / tau
mass = gammainc(2, normalized_times)
intsfr = (mass[1] - mass[0]) / mass[2]
return intsfr
def _testCompareToExplicitDerivative(self, dtype):
"""Compare to the explicit reparameterization derivative.
Verifies that the computed derivative satisfies
dsample / dalpha = d igammainv(alpha, u) / dalpha,
where u = igamma(alpha, sample).
Args:
dtype: TensorFlow dtype to perform the computations in.
"""
delta = 1e-3
np_dtype = dtype.as_numpy_dtype
try:
from scipy import misc # pylint: disable=g-import-not-at-top
from scipy import special # pylint: disable=g-import-not-at-top
alpha_val = np.logspace(-2, 3, dtype=np_dtype)
alpha = constant_op.constant(alpha_val)
sample = random_ops.random_gamma([], alpha, np_dtype(1.0), dtype=dtype)
actual = gradients_impl.gradients(sample, alpha)[0]
(sample_val, actual_val) = self.evaluate((sample, actual))
u = special.gammainc(alpha_val, sample_val)
expected_val = misc.derivative(
lambda alpha_prime: special.gammaincinv(alpha_prime, u),
alpha_val, dx=delta * alpha_val)
self.assertAllClose(actual_val, expected_val, rtol=1e-3, atol=1e-3)
except ImportError as e:
tf_logging.warn("Cannot use special functions in a test: %s" % str(e))
def cdf(self, y):
a = self.a
b = self.b
p = self.p
z = (y/b)**a
cdf = gammainc(p, z)
return cdf
def F(nu,x):
"""
Boys function.
INPUT:
NU: Boys function index
X: Boys function variable
OUTPUT:
FF: Value of the Boys function for index NU evaluated at X
Source:
Evaluation of the Boys Function using Analytical Relations
I. I. Guseinov and B. A. Mamedov
Journal of Mathematical Chemistry
2006
"""
if x < 1e-8:
# Taylor expansion for argument close or equal to zero (avoid division by zero)
ff = 1 / (2 * nu + 1) - x / (2 * nu + 3)
else:
# Evaluate Boys function with incomplete and complete Gamma functions
ff = 0.5 / x**(nu+0.5) * spec.gamma(nu+0.5)*spec.gammainc(nu+0.5,x)
return ff
def df(self, *arg):
"""
Returns the normalised DF, can only be called after Poisson solver
Arguments can be:
- r, v (isotropic single-mass models)
- r, v, j (isotropic multi-mass models)
- r, v, theta, j (anisotropic models)
- x, y, z, vx, vy, vz, j (all models)
Here j specifies the mass bin, j=0 for single mass
Works with scalar and array input
"""
if (len(arg)<2)|(len(arg)==5)|(len(arg)==6)|(len(arg)>7):
raise ValueError("Error: df needs 2, 3, 4 or 7 arguments")
if len(arg) == 2:
r, v = (self._tonp(q) for q in arg)
j = 0
if len(arg) == 3:
r, v = (self._tonp(q) for q in arg[:-1])
j = arg[-1]
if len(arg) == 4:
r, v, theta = (self._tonp(q) for q in arg[:-1])
j = arg[-1]
if len(arg) < 7: r2, v2 = r**2, v**2
if len(arg) == 7:
x, y, z, vx, vy, vz = (self._tonp(q) for q in arg[:-1])
j = arg[-1]
r2 = x**2 + y**2 + z**2
v2 = vx**2 + vy**2 + vz**2
r, v = sqrt(r2), sqrt(v2)
phi = self.interp_phi(r)
vesc2 = 2.0*phi # Note: phi > 0
DF = numpy.zeros([max(r.size, v.size)])
c = (r<self.rt)&(v2<vesc2)
E = (phi-0.5*v2)/self.sig2[j] # Dimensionless positive energy
DF[c] = exp(E[c])
# Float truncation parameter
# Following Gomez-Leyton & Velazquez 2014, J. Stat. Mech. 4, 6
if (self.g>0): DF[c] *= gammainc(self.g, E[c])
if (self.raj[j] < self.ramax):
if (len(arg)==7): J2 = v2*r2 - (x*vx + y*vy + z*vz)**2
if (len(arg)==4): J2 = sin(theta)**2*v2*r2
DF[c] *= exp(-J2[c]/(2*self.raj[j]**2*self.sig2[j]))
DF[c] *= self.A[j]
return DF
def qtimedelay(set_par,deltim=1.):
'''
gamma-function based weight function to control the runoff delay
'''
alpha=3.0
print set_par['mut']
alamb = alpha/set_par['mut']
psave=0.0
set_par['frac_future']=np.zeros(500.) #Parameter added
ntdh = set_par['frac_future'].size
deltim=1.
print 'qtimedelay is calculated with a unit of',deltim,'hours'
for jtim in range(ntdh):
# print jtim
tfuture=jtim*deltim
# print alamb*tfuture
cumprob= special.gammainc(alpha, alamb*tfuture)# hoeft niet want verschil wordt genomen: /special.gamma(alpha)
# print cumprob
set_par['frac_future'][jtim]=max(0.,cumprob-psave)
if set_par['frac_future'][jtim] > 0.0001:
print set_par['frac_future'][jtim]
psave = cumprob
if cumprob < 0.99:
print 'not enough bins in the frac_future'
#make sure sum to one
set_par['frac_future'][:]=set_par['frac_future'][:]/set_par['frac_future'][:].sum()
return set_par
def deflections(self, xin, yin):
from numpy import ones, arctan as atan, arctanh as atanh
from math import cos, sin, pi, exp
from numpy import arcsin as asin, arcsinh as asinh
from scipy.special import gammainc, gamma
x, y = self.align_coords(xin, yin)
q = self.q
if q == 1.:
q = 1. - 1e-7 # Avoid divide-by-zero errors
from sersic import sersicdeflections as sd
b, n, re = self.b, self.n, self.re
k = 2. * n - 1. / 3 + 4. / (405. * n) + 46 / (25515. * n**2)
b = b / q**0.5
re = re / q**0.5
amp = b * b * k**(2 * n) / (
2 * n * re**2 * (gamma(2 * n) * gammainc(2 * n,
k * (b / re)**(1 / n))))
if x.ndim > 1:
yout, xout = sd(-1 * y.ravel(), x.ravel(), amp, re, n, q)
xout, yout = xout.reshape(x.shape), -1 * yout.reshape(y.shape)
else:
yout, xout = sd(-1 * y, x, amp, re, n, q)
yout = -1 * yout
theta = -(self.theta - pi / 2.)
ctheta = cos(theta)
stheta = sin(theta)
x = xout * ctheta + yout * stheta
y = yout * ctheta - xout * stheta
return x, y
def IRF3(self, parameters):
A1 = parameters['A1'].value
a1 = parameters['a1'].value
n1 = parameters['n1'].value
t = np.arange(1.0, 10000)
Fs = A1 * t**n1 * (t/a1)**-n1 * gammainc(n1, t/a1) # Step response function based on pearsonIII
return np.append(0, Fs[1:] - Fs[0:-1]) # block reponse function
def loglik(self,dat):
'''
Computes the loglikelihood of the data points in dat.
:param dat: Data points for which the loglikelihood will be computed.
:type dat: natter.DataModule.Data
:returns: An array containing the loglikelihoods.
:rtype: numpy.array
'''
sigma,kappa,delta,gamma,n,p,s = \
self.param['sigma'],self.param['kappa'],self.param['delta'],\
self.param['gamma'],self.param['n'],self.param['p'],self.param['s']
r = squeeze(dat.X)
if delta == 0.0:
addTerm = - log(gammainc(n/p,kappa**p/2.0/s))
else:
addTerm = 0
ll = log(p) + n*log(kappa) + (n*gamma+2.0*n*delta-2.0)/2.0 * log(r)\
+ log(2.0*delta*(r**gamma + sigma**2.0) + gamma*sigma**2.0)\
-gammaln(n/p) - n/p*log(s) - (n+p)/p * log(2.0) - (n+2.0)/2.0 *log(sigma**2 + r**gamma) \
- r**(p*gamma/2.0 + p*delta)*kappa**p/2.0/s/(sigma**2+r**gamma)**(p/2.0) + addTerm
return ll
def delta_chi2(ptarg, ndeg):
""" Find delta chi2 corresponding to the given probability for n
degrees of freedom (number of data points - number of free
parameters).
ptarg is the target probability (ptarg=0.9 correponds to 90%).
ptarg is the probability that a model fitted to data with the
given number of degrees of freedom will have a chisq val
delta_chi2 larger than the minimum chisq value for the
best-fitting parameters.
"""
assert 0 <= ptarg <= 1
d0, d1 = 0, 1e5
a = 0.5 * ndeg
while True:
d = 0.5*(d0 + d1)
p = gammainc(a, 0.5*d)
#print ptarg, p, d, d0, d1
if p > ptarg:
d1 = d
else:
d0 = d
if abs(p - ptarg) < 1e-4:
break
return d
def _rhov2int(self, phi, r, ra):
"""Compute product of density and mean square velocity """
# Isotropic case first
rhov2r = exp(phi)*gammainc(self.g + 2.5, phi)
rhov2 = 3*rhov2r
rhov2t = 2*rhov2r
# Add anisotropy
if (ra < self.ramax)&(r>0)&(phi>0):
p, g = r/ra, self.g
p2 = p**2
p12 = 1+p2
g3, g5, g7, fp2 = g+1.5, g+2.5, g+3.5, phi*p2
P1 = p2*phi**g5/gamma(g7)
H1, H2 = hyp1f1(1, g7, -fp2), hyp1f1(2, g7, -fp2)
rhov2r += P1*H1
rhov2r /= p12
rhov2t /= p12
rhov2t += 2*P1*(H1/p12 + H2)
rhov2t /= p12
rhov2 = rhov2r + rhov2t
return rhov2, rhov2r, rhov2t
def sersic_L_Rcut(Ieff,Reff,ndex,Rcut=100):
bn = 2.0*ndex-1/3.0+0.009876/ndex
xtmp = bn*(Rcut)**(1.0/ndex)
res = 2.0*np.pi*Ieff*Reff**2.0*ndex*(np.e**bn/(bn**(2.0*ndex)))*spf.gammainc(2.0*ndex,xtmp)*spf.gamma(2.0*ndex)
#res = np.pi*Ieff*Reff**2.0*(2.0*ndex)*(np.e**bn/(bn**(2.0*ndex)))*spf.gamma(2.0*ndex)
return res
def likelihood(p, f, nswitch):
if nswitch == 1:
ecc, pa, r_e, g_x, g_y = p #Galaxy position, eccentricity, position
x, y, N, area, max_rad, E_bg, n = f #angle, half-light radius, background density
else:
ecc, pa, r_e, n, g_x, g_y = p
x, y, N, area, max_rad, E_bg = f
# rand = np.random.choice(len(x),size=(int(round(area*E_bg))),replace=False)
#
# mask = np.zeros(len(x))
# mask[rand] = 1
# x = np.ma.masked_array(x,mask)
# y = np.ma.masked_array(y,mask)
# N = len(x) - int(round(area*E_bg))
# Eqn from Martin, De Jong, Rix 2008
ellip_r = np.sqrt(((1 - ecc)**-1 * ((x - g_x) * np.cos(pa) -
(y - g_y) * np.sin(pa)))**2 +
((x - g_x) * np.sin(pa) + (y - g_y) * np.cos(pa))**2)
b_n = 1.9992 * n - .3271
##Normalization term comes from Graham and Driver, 2005
norm_term = ((N - E_bg * area) * b_n**(2 * n) / (r_e**2 * 2 * np.pi * n *
(1 - ecc)) /
(gammainc(2 * n,
b_n * (max_rad / r_e)**(1. / n)) * gamma(2 * n)))
fcn = (np.sum(
np.log(norm_term * np.exp(-b_n * ((ellip_r / r_e)**(1. / n))) + E_bg)))
########David Sand's LL fcn
# N_ = N - area * E_bg
# S0 = N_ / (2 * np.pi * (r_e / b_n)**2 * (1 - ecc))
# fcn = np.sum(np.log(S0 * np.exp(-b_n* (ellip_r / r_e)) + E_bg))
# b_n = 1.9992*n - .3271
# # N = N - E_bg*area
# fcn = (-2 * N * np.log(r_e)
# + N * np.log(b_n**(2*n) / n
# / (gammainc(2*n,b_n * (max_rad/r_e)**(1./n)) * gamma(2*n))
# )
# - N * np.log(2 * np.pi)
# - N * np.log(1-ecc)
# - b_n * np.sum(ellip_r ** (1/n)) / r_e ** (1/n)
# # + np.log(E_bg)
# )
if (np.min(x) < g_x < np.max(x) and np.min(y) < g_y < np.max(y)
and 0 <= ecc < 1 and 0 <= pa < np.pi and # 0 < r_e < 10
.3 < n < 10 and 0 < r_e < 10
#and 0 <= E_bg < N/area
):
return (fcn)
else:
return (-np.inf)
def Iq(q, rg=60.0, porod_exp=3.0):
"""
:param q: Input q-value (float or [float, float])
:param rg: Radius of gyration
:param porod_exp: Porod exponent
:return: Calculated intensity
"""
usub = (q * rg)**2 * (2.0 / porod_exp + 1.0) * (2.0 / porod_exp +
2.0) / 6.0
with errstate(divide='ignore', invalid='ignore'):
upow = power(usub, -0.5 * porod_exp)
# Note: scipy gammainc is "regularized", being gamma(s,x)/Gamma(s),
# so need to scale by Gamma(s) to recover gamma(s, x).
result = (porod_exp * upow *
(gamma(0.5 * porod_exp) * gammainc(0.5 * porod_exp, usub) -
upow * gamma(porod_exp) * gammainc(porod_exp, usub)))
result[q <= 0] = 1.0
return result
def singinhomobremss(
freq, S_norm, alpha, p, peak_frequency
): # Single inhomogeneous free-free emission model
return (
S_norm
* (p + 1)
* ((freq / peak_frequency) ** (2.1 * (p + 1) - alpha))
* special.gammainc((p + 1), ((freq / peak_frequency) ** (-2.1)))
* special.gamma(p + 1)
)
def mass(self, r):
x = r / self._rs
gamma = spfn.gamma(3. / self._alpha)
return 4*np.pi*self._rhos*np.power(self._rs,3)/self._alpha* \
np.exp(2./self._alpha)* \
np.power(2./self._alpha,-3./self._alpha)* \
gamma*spfn.gammainc(3./self._alpha,
(2./self._alpha)*np.power(x,self._alpha))
def _rhoint_rot(self, phi, r, theta):
# Compute eq 3 for scalar phi, r and theta
g, om = self.g, self.omega
Q = 3 * sqrt(2) * r * sin(theta)
# Eq 3
integ = quad(
lambda x: exp(phi - x) * gammainc(g, phi - x) * numpy.sinh(
om * Q * x**0.5), 0, phi)[0]
return integ / (om * Q * gamma(1.5))
def _polymerexclvol(self, x):
sc = self.params['scale']
rg = self.params['rg']
mm = self.params['m']
bg = self.params['background']
nu = 1.0 / mm
Xx = x * x * rg * rg *(2.0 * nu + 1.0) * (2.0 * nu + 2.0) / 6.0
onu = 1.0 / nu
o2nu = 1.0 /(2.0 * nu)
Ps =(1.0 / (nu * pow(Xx,o2nu))) * (gamma(o2nu)*gammainc(o2nu,Xx) - \
1.0 / pow(Xx,o2nu) * gamma(onu)*gammainc(onu,Xx))
if x == 0:
Ps = 1.0
return (sc * Ps + bg);
def get_mass(self,r):
"""galpy power-law mass
be careful of the scipy gamma incomplete gamma function definition!
inputs
---------------
r : radius to compute enclosed mass for
"""
return 2.*np.pi*self.rho0*self.rcut**(3.-self.alpha)*gammainc(1.5-self.alpha/2.,(r/self.rcut)**2.)*gamma(1.5-self.alpha/2.)
def test_incomplete_gamma_p(self, N, percentile):
inverse = covid19.inv_incomplete_gamma_p(percentile, N)
forward = gammainc(N, inverse)
np.testing.assert_almost_equal(
percentile,
forward,
decimal=6,
err_msg=
"Forward and inverse calculations of incomplete gamma don't agree")
def calculate_r(z, t2, length, alpha, beta, N):
x1 = np.power(z, 2)
x1_ = np.cumsum(x1, axis=1)
x1 = np.outer(x1_, t2)
x2 = x1.transpose()
x3 = np.zeros(length)
x3 = np.c_[x3, x2[:, :-1]]
x3 = np.tril(x3)
x4 = (x1 - x3) / 2
# x5 = (1 - 1 / eta**2) * x4 - N * np.log(eta)
x5_1 = np.tril(np.log(gammainc(beta + x4, alpha + N / 2)))
x5_2 = np.tril(gammaln(alpha + N / 2))
x5_3 = alpha * np.log(beta) - gammaln(alpha) - np.log(gammainc(
beta, alpha))
x5_4 = -(alpha + N / 2) * np.tril(np.log(beta + x4))
x5 = x4 + x5_1 + x5_2 + x5_3 + x5_4
lmbda = np.tril(np.exp(x5))
r = np.sum(lmbda.T, axis=0)
return r
def posterior(D):
"""Remember gammainc is normalized to gamma"""
min_D = sigma2 / dt
if D <= min_D:
ln_res = -np.inf
else:
ln_res = (p * log(n * V * G3 / (4 * dt * D)) - n * V * G3 /
(4 * D * dt) - log(D) -
log(gammainc(p, rel_loc_error * G3)) - gammaln(p))
return np.exp(ln_res)
def getGamma(k, maxsbins):
# Space out bins according to shape k+1 so that each bin represents an equal
# (incomplete) expectation of X in the shape k distribution. Other bin choices are
# possible, but this is a good one from the point of view of achieving more accuracy
# using a smaller number of bins.
l = gammadist.ppf([i / maxsbins for i in range(maxsbins)], k + 1)
# Calculate:
# m0[i] = P[X < l_i]
# m1[i] = E[X; X < l_i]
# susc[i] = E[X | l_i <= X < l_{i+1}], the representative susceptibility for bin i
# q[i] = P(l_i <= X < l_{i+1})
m0 = np.append(gammainc(k, l), 1)
m1 = np.append(gammainc(k + 1, l), 1)
susc = np.array([(m1[i + 1] - m1[i]) / (m0[i + 1] - m0[i])
for i in range(maxsbins)])
q = np.array([m0[i + 1] - m0[i] for i in range(maxsbins)])
return susc, q
def getMass(self, sigCrit=1.):
from math import pi
from scipy.special import gammainc, gamma
b, n, re = self.b, self.n, self.re
k = 2. * n - 1. / 3 + 4. / (405. * n) + 46 / (25515. * n**2)
amp = b * b * k**(2 * n) / (
2 * n * re**2 * (gamma(2 * n) * gammainc(2 * n,
k * (b / re)**(1 / n))))
return 2 * pi * sigCrit * re**2 * amp * gamma(2 * n) * n / k**(2 * n)
def eta_esin(nu, r, tempi, tempe, v):
theta = boltzmann * tempe / (electronMass * cLight2)
ne = eDens(r, tempi, tempe, v)
h = height(r, tempi, tempe)
tau_es = 2.0 * ne * thomson * h
s_exp = tau_es * (tau_es + 1.0)
A = 1.0 + 4.0 * theta * (1.0 + 4.0 * theta)
etaMax = 3.0 * boltzmann * tempe / (planck * nu)
jm = np.log(etaMax) / np.log(A)
gamma1 = sp.gammainc(jm + 1.0, A * s_exp)
gamma2 = sp.gammainc(jm + 1.0, s_exp)
aux2 = s_exp * (A - 1.0)
result = np.where(aux2 < 200.0,
np.exp(aux2) * (1.0 - gamma1) + etaMax * gamma2,
etaMax * gamma2)
return result
def testIgammaMediumValues(self, dtype, rtol, atol):
# Test values near zero.
x = np.random.uniform(low=1., high=100., size=[NUM_SAMPLES]).astype(dtype)
a = np.random.uniform(low=1., high=100., size=[NUM_SAMPLES]).astype(dtype)
expected_values = sps.gammainc(a, x)
with self.session() as sess:
with self.test_scope():
actual = sess.run(math_ops.igamma(a, x))
self.assertAllClose(expected_values, actual, atol=atol, rtol=rtol)
def expected_grad(sample, concentration, rate):
u = sp_special.gammainc(
concentration, sample * rate)
delta = 1e-3
return sp_misc.derivative(
lambda concentration_prime: sp_special.
gammaincinv( # pylint:disable=g-long-lambda
concentration_prime, u),
concentration,
dx=delta * concentration) / rate
def sersic_mag_tot_to_Ieff(mag_tot, Reff, ndex, z, Rcut=100): # g band
bn = 2.0 * ndex - 1 / 3.0 + 0.009876 / ndex
xtmp = bn * (Rcut)**(1.0 / ndex)
ktmp = 2.0 * np.pi * ndex * (np.e**bn / (bn**(2.0 * ndex))) * spf.gammainc(
2.0 * ndex, xtmp) * spf.gamma(2.0 * ndex)
Dl_s = p13.luminosity_distance(z).value * 1e6
Ieff = 10.0**((5.12 - 2.5 * np.log10(ktmp) - 5.0 * np.log10(Reff) +
5.0 * np.log10(Dl_s / 10) - mag_tot) * 0.4) # L_sun/pc^2
return Ieff
def rho0(self, mass):
"""Central density for the Einasto profile in M_sun/Mpc^3 h^2"""
alpha = 0.18
R200 = self.R200(mass)
conc = self.concentration(mass)
gamma = special.gammainc(
3 / alpha, 2 / alpha * conc**alpha) * special.gamma(3 / alpha)
prefac = 4 * math.pi * np.exp(
2 / alpha) / alpha * (alpha / 2)**(3 / alpha)
return mass / gamma / prefac / (R200 / conc)**3
def cdf(self, dat):
'''
Evaluates the cumulative distribution function on the data points in dat.
:param dat: Data points for which the c.d.f. will be computed.
:type dat: natter.DataModule.Data
:returns: A numpy array containing the probabilities.
:rtype: numpy.array
'''
u = array([self.param['a'], self.param['b']])
u = .5 + 0.5 * sign(u) * gammainc(
1 / self.param['p'],
abs(u)**self.param['p'] / self.param['s'])
return squeeze(.5 + 0.5 * sign(dat.X) *
gammainc(1 / self.param['p'],
abs(dat.X)**self.param['p'] /
self.param['s']) - u[0]) / (u[1] - u[0])
def get_rho0(self):
"""solve the mass enclosed at 100rcut to recover the central density
returned in units of density, mass/(length^3)
"""
r = 100.*self.rscl
tmp_mass = 2.*np.pi*self.rcut**(3.-self.alpha)*gammainc(1.5-self.alpha/2.,(r/self.rcut)**2.)*gamma(1.5-self.alpha/2.)
return self.M/tmp_mass
def randsphere(n_points,ndim,radius,center = []):
if center == []:
center = np.array([0]*ndim)
r = radius
x = np.random.normal(size=(n_points, ndim))
ssq = np.sum(x**2,axis=1)
fr = r*gammainc(ndim/2,ssq/2)**(1/ndim)/np.sqrt(ssq)
frtiled = np.tile(fr.reshape(n_points,1),(1,ndim))
p = center + np.multiply(x,frtiled)
return p, center
def approximateEntropy(m, sequence):
n = len(sequence)
fi1 = get_fi(m, sequence)
fi2 = get_fi(m + 1, sequence)
print(fi1 - fi2)
print(math.log(2) - (fi1 - fi2))
obs = 2 * n * (math.log(2) - (fi1 - fi2))
print(obs)
p = 1 - special.gammainc(math.pow(2, m - 1), obs / 2)
return p, p >= 0.01
def v_esc_theory_flat_lambda(theta,z_c,alpha,rho_2_1e14,r_2,beta,little_h,omega_DE):
"""cosmology"""
H_z = H_z_function(z_c, [omega_DE,little_h],'flat_lambda')
q_z= q_z_function(z_c,omega_DE,'flat_lambda')
"""Einasto"""
### map betweenEinasto params and Miller et al params###
rho_2 = rho_2_1e14*1e14
n = 1/alpha
rho_0 = rho_2 * np.exp(2.*n)
h = r_2 / (2.*n)**n
d_n = (3*n) - (1./3.) + (8./1215.*n) + (184./229635.*n**2.) + (1048./31000725.*n**3.) - (17557576./1242974068875. * n**4.)
gamma_3n = 2 * ((ss.gammainc(3*n , d_n) ) * ss.gamma(3*n))
Mtot = 4. * np.pi * rho_0 * Msun * (h**3.) * n * gamma_3n #kg
G_newton = astroc.G.to( u.Mpc * u.km**2 / u.s**2 / u.kg).value #Mpc km2/s^2 kg
r_eq_cubed = -((G_newton*Mtot) / (q_z * H_z**2.)) #Mpc ^3
r_eq = (r_eq_cubed)**(1.0/3.0)# Mpc
"""if r_eq -> infinity then return vanilla Einasto"""
if q_z < 0.:
### phi orig ###
s_orig = r/h
part1_orig = 1. - ( ( ss.gammaincc(3.*n,s_orig**(1./n)) * ss.gamma(3.*n) ) / gamma_3n )
part2_orig = (s_orig * ss.gammaincc(2.*n,s_orig**(1./n)) * ss.gamma(2.*n) ) / gamma_3n
phi_ein_orig = -(G_newton* Mtot/(s_orig*h)) * (part1_orig+part2_orig)
### phi r_eq ###
s_req = r_eq/h
part1_req = 1. - ( ( ss.gammaincc(3.*n,s_req**(1./n)) * ss.gamma(3.*n) ) / gamma_3n )
part2_req = (s_req * ss.gammaincc(2.*n,s_req**(1./n)) * ss.gamma(2.*n) ) / gamma_3n
phi_ein_req = -(G_newton* Mtot/(s_req*h)) * (part1_req+part2_req)
v_esc = np.sqrt(-2.*( phi_ein_orig - phi_ein_req ) - q_z * (H_z**2.) * (r**2 - r_eq**2) )
elif q_z >= 0.:
### phi orig ###
s_orig = r/h
part1_orig = 1. - ( ( ss.gammaincc(3.*n,s_orig**(1./n)) * ss.gamma(3.*n) ) / gamma_3n )
part2_orig = (s_orig * ss.gammaincc(2.*n,s_orig**(1./n)) * ss.gamma(2.*n) ) / gamma_3n
phi_ein_orig = -(G_newton* Mtot/(s_orig*h)) * (part1_orig+part2_orig)
v_esc = np.sqrt(-2.*( phi_ein_orig ))
v_esc_projected = v_esc / np.sqrt(g_beta(beta))
return v_esc_projected
def v_esc_theory_non_flat_lil(theta,z_c,alpha,rho_2_1e14,r_2,beta,lil_omega_M,lil_omega_lambda):
"""cosmology"""
qHsquared = ( (0.5 * lil_omega_M * (1+z_c)**3.)- lil_omega_lambda) * 100.**2.
"""Einasto"""
### map betweenEinasto params and Miller et al params###
rho_2 = rho_2_1e14*1e14
n = 1/alpha
rho_0 = rho_2 * np.exp(2.*n)
h = r_2 / (2.*n)**n
d_n = (3*n) - (1./3.) + (8./1215.*n) + (184./229635.*n**2.) + (1048./31000725.*n**3.) - (17557576./1242974068875. * n**4.)
gamma_3n = 2 * ((ss.gammainc(3*n , d_n) ) * ss.gamma(3*n))
Mtot = 4. * np.pi * rho_0 * Msun * (h**3.) * n * gamma_3n #kg
G_newton = astroc.G.to( u.Mpc * u.km**2 / u.s**2 / u.kg).value #Mpc km2/s^2 kg
r_eq_cubed = -(G_newton*Mtot) / qHsquared #Mpc ^3
r_eq = (r_eq_cubed)**(1.0/3.0)# Mpc
"""if r_eq -> infinity then return vanilla Einasto"""
if math.isnan(r_eq) == False:
### phi orig ###
s_orig = r/h
part1_orig = 1. - ( ( ss.gammaincc(3.*n,s_orig**(1./n)) * ss.gamma(3.*n) ) / gamma_3n )
part2_orig = (s_orig * ss.gammaincc(2.*n,s_orig**(1./n)) * ss.gamma(2.*n) ) / gamma_3n
phi_ein_orig = -(G_newton* Mtot/(s_orig*h)) * (part1_orig+part2_orig)
### phi r_eq ###
s_req = r_eq/h
part1_req = 1. - ( ( ss.gammaincc(3.*n,s_req**(1./n)) * ss.gamma(3.*n) ) / gamma_3n )
part2_req = (s_req * ss.gammaincc(2.*n,s_req**(1./n)) * ss.gamma(2.*n) ) / gamma_3n
phi_ein_req = -(G_newton* Mtot/(s_req*h)) * (part1_req+part2_req)
v_esc = np.sqrt(-2.*( phi_ein_orig - phi_ein_req ) - qHsquared * (r**2 - r_eq**2) )
elif math.isnan(r_eq) == True:
### phi orig ###
s_orig = r/h
part1_orig = 1. - ( ( ss.gammaincc(3.*n,s_orig**(1./n)) * ss.gamma(3.*n) ) / gamma_3n )
part2_orig = (s_orig * ss.gammaincc(2.*n,s_orig**(1./n)) * ss.gamma(2.*n) ) / gamma_3n
phi_ein_orig = -(G_newton* Mtot/(s_orig*h)) * (part1_orig+part2_orig)
v_esc = np.sqrt(-2.* phi_ein_orig )
v_esc_projected = v_esc / np.sqrt(g_beta(beta))
return v_esc_projected
def mole_gammainc(m, z):
"""
Compute incomplete gamma function defined below
. F_m(z) = Int_0^1 t^{2m} Exp[-zt^2]dt
Impled with scipy.special.gammainc(a,x) defined below
. 1/Gamma[a] Int_0^x t^{a-1} Exp[-t] dt
Variable transformation
. t->(t^2)/x
. dt -> (2t/x) dt
. gammainc(a,x) = x^{1-a} /Gamma[a] Int_0^1 t^{2a-2} (2t/x) Exp[-(t^2)/x] dt
. = 2 (x^(-a))/gamma[a] Int_0^1 t^{2a-1} Exp[-(t^2)/x] dt
. F_m(z) = (1/2) gamma[a] x^a
Special case
. z = 0
. F_m(0) = Int_0^1 t^{2m} dt
. = (2m+1)^{-1} [t^{2m+1}]_0^1
. = (2m+1)^{-1}
"""
eps = 10.0**(-14.0)
if (abs(z) < eps):
return 1 / (2 * m + 1.0)
if (m == 0):
a = m + 0.5
res = gamma(a) / (2 * z**a) * gammainc(a, z)
# res = sqrt(pi/z)/2 * erf(sqrt(z))
else:
a = m + 0.5
res = gamma(a) / (2 * z**a) * gammainc(a, z)
if ((not res < 1) and (not res > -1)):
raise RuntimeError("""gammain failed
res: {2}
m: {0}
z: {1}
""".format(m, z, res))
return res
def qminus_esin_fast(r, tempi, tempe, v):
nuMin = 1.0e6
nuMax = 1.0e21
iter = 30
paso = np.power(nuMax / nuMin, 1.0 / float(iter - 1))
sum = 0.0
nu = nuMin
theta = boltzmann * tempe / (electronMass * cLight2)
ne = eDens(r, tempi, tempe, v)
h = height(r, tempi, tempe)
k_es = ne * thomson
tau_es = 2.0 * k_es * h
A = 1.0 + 4.0 * theta * (1.0 + 4.0 * theta)
for i in range(iter):
x = planck * nu / (electronMass * cLight2) / (3.0 * theta)
bnu = blackbody_nu(nu, tempe).value
bnu2 = 4.0 * np.pi * bnu
xi = xiBremss(nu, r, tempi, tempe, v) + xiSync(nu, r, tempi, tempe, v)
#k_nu = np.where(bnu2 > 1.0e-2*xi*h,xi/bnu2,50.0)
k_nu = xi / bnu2
#l_eff = 1.0/np.sqrt(k_nu*(k_nu+k_es))
#tau_es = 2.0*k_es*np.where(l_eff<h,l_eff,h)
s_exp = tau_es * (tau_es + 1.0)
etaMax = 3.0 * boltzmann * tempe / (planck * nu)
jm = np.log(etaMax) / np.log(A)
gamma1 = sp.gammainc(jm + 1.0, A * s_exp)
gamma2 = sp.gammainc(jm + 1.0, s_exp)
aux2 = s_exp * (A - 1.0)
#etaa = np.where(aux2<200.0,np.exp(aux2)*(1.0-gamma1)+etaMax*gamma2,etaMax*gamma2)
etaa = np.exp(aux2) * (1.0 - gamma1) + etaMax * gamma2
tau_nu = np.sqrt(np.pi) / 2.0 * k_nu * h
fluxx = 2.0*np.pi/np.sqrt(3.0) * bnu * \
(1.0-np.exp(-2.0*np.sqrt(3.0)*tau_nu))
dnu = nu * (np.sqrt(paso) - 1.0 / np.sqrt(paso))
sum += fluxx * etaa * dnu
nu = nu * paso
return sum * 2.0 / (np.sqrt(np.pi) * h)
def compute_deff_prime(alpha_ice, g_ice, sigma_a, delta_a1, delta_a2, sigma_v,
delta_v1, delta_v2, Dr, x):
alpha_a1_g_ice = (alpha_ice + delta_a1 + 1) / g_ice
alpha_a2_g_ice = (alpha_ice + delta_a2 + 1) / g_ice
t_k = (1 / x) * (Dr**g_ice)
#coordinate transform for limit of integration
gamma_fun_a1 = gm.gamma(alpha_a1_g_ice)
inc_gamma_fun_a1 = gm.gammainc(t_k, alpha_a1_g_ice)
gamma_fun_a2 = gm.gamma(alpha_a2_g_ice)
inc_gamma_fun_a2 = gm.gammainc(t_k, alpha_a2_g_ice)
alpha_v1_g_ice = (alpha_ice + 2 * delta_v1 + 1) / g_ice
alpha_v2_g_ice = (alpha_ice + 2 * delta_v2 + 1) / g_ice
gamma_fun_v1 = gm.gamma(alpha_v1_g_ice)
inc_gamma_fun_v1 = gm.gammainc(t_k, alpha_v1_g_ice)
gamma_fun_v2 = gm.gamma(alpha_v2_g_ice)
inc_gamma_fun_v2 = gm.gammainc(t_k, alpha_v2_g_ice)
A = Dr**(-2 * delta_v1) * inc_gamma_fun_v1 * gamma_fun_v1
B = Dr**(-2 * delta_v2) * (1 - inc_gamma_fun_v2) * gamma_fun_v2
C = Dr**(-delta_a1) * inc_gamma_fun_a1 * gamma_fun_a1
G = Dr**(-delta_a2) * (1 - inc_gamma_fun_a2) * gamma_fun_a2
area_factor = (C * x**(delta_a1 / g_ice) + G * x**(delta_a2 / g_ice))
deff_prime = (A * x**(2 * delta_v1 / g_ice) +
B * x**(2 * delta_v2 / g_ice)) / area_factor
deff_prime = Dr * deff_prime**0.25 * (sigma_v**2 / sigma_a)**0.25
#compute average area of particle in square microns.
ave_area = sigma_a * (np.pi / 4) * Dr**2 * area_factor / gm.gamma(
(alpha_ice + 1) / g_ice)
ave_area *= 1e-12
LOG.debug('ave_area')
LOG.debug(ave_area)
return deff_prime, ave_area
def boys(n, x):
import scipy.special as sp
# x -> 0: the denominator goes to 0, zero deviation error will occur.
# thus, have to return 0-limit value(x->0.)
if x < 0.: raise
radius = 0.0001
if x < radius: return 1. / (2.0 * n + 1.0)
else:
numerator = sp.gammainc(n+0.5, x) * sp.gamma(n + 0.5)
denominator = 2.0 * math.pow(x, n + 0.5)
return numerator /denominator
def Q(self, nu, chisq):
"""
Calculate the probability that a normally distributed random numbers produce a chi squared value greater than
measured. The distribution is described by an incomplete gamma function and should only be used for models that
are linear or close to linear in their parameters.
[1] pg. 778-780.
:param nu: Number of degrees of freedom.
:param chisq: Measured chi squared value.
:return: The probability that random noise will produce a chi squared greater than the measured value.
"""
return 1 - spsp.gammainc(nu / 2, chisq / 2)
def _cdf(self, x):
k = self.k
mu = self.mu
p = self.p
sumand = (k * (-p + 1) * gamma(k) * gammainc(k, mu * x)) / (
(factorial(k - 1) * gamma(k + 1)))
if mu == 0:
return sumand + mu * p * x
else:
return sumand + mu * p * (-(1 / mu) * np.exp(-mu * x))
def gamma_dist(k, theta, n=400):
assert k > 0 and theta > 0
assert type(n) == int and n > 2
dist = np.zeros((n, 2))
dist[:, 0] = np.linspace(0,
k * theta + 5 * np.sqrt(k) * theta,
n,
endpoint=False)
cumdist = np.concatenate((sp.gammainc(k, dist[:, 0] / theta), [1]))
dist[:, 1] = np.diff(cumdist)
return dist
