def make_xiell_fixedbias(self,
f,
bvec,
apar=1,
aperp=1,
ngauss=4,
kmin=1e-3,
kmax=0.8,
nk=100,
nmax=5):
kv, p0k, p2k, p4k = self.make_pell_fixedbias(f,
bvec,
apar=apar,
aperp=aperp,
ngauss=ngauss,
kmin=kmin,
kmax=kmax,
nk=nk,
nmax=nmax)
damping = np.exp(-(self.kint / 10)**2)
p0int = loginterp(kv, p0k)(self.kint) * damping
p2int = loginterp(kv, p2k)(self.kint) * damping
p4int = loginterp(kv, p4k)(self.kint) * damping
ss0, xi0 = self.sphr.sph(0, p0int)
ss2, xi2 = self.sphr.sph(2, p2int)
xi2 *= -1
ss4, xi4 = self.sphr.sph(4, p0int)
return (ss0, xi0), (ss2, xi2), (ss4, xi4)
def compute_cumulants(self, b1, b2, bs, b3, alpha, alpha_v, alpha_s0, alpha_s2, s2fog):
'''
Calculate velocity moments and turn into cumulants.
'''
# Compute each moment
self.pars = np.array([b1, b2, bs, b3, alpha, alpha_v, alpha_s0, alpha_s2, s2fog])
self.kv = self.pktable[:,0]
self.pzel = self.pktable[:,-1]
self.peft = self.pktable[:,1] + b1*self.pktable[:,2] + b1**2*self.pktable[:,3]\
+ b2*self.pktable[:,4] + b1*b2*self.pktable[:,5] + b2**2 * self.pktable[:,6]\
+ bs*self.pktable[:,7] + b1*bs*self.pktable[:,8] + b2*bs*self.pktable[:,9]\
+ bs**2*self.pktable[:,10] + b3*self.pktable[:,11] + b1*b3*self.pktable[:,12] + alpha* self.kv**2 * self.pzel
_integrand = loginterp(self.kv, self.peft)(self.kint) #at some point turn into theory extrapolation
_integrand = self.weight * _integrand + (1-self.weight) * ((1+b1)**2 * self.plin)
qint, xi = self.sph_gsm.sph(0,_integrand*self.window)
self.xieft = np.interp(self.rint, qint, xi)
self.vkeft = self.vktable[:,1] + b1*self.vktable[:,2] + b1**2*self.vktable[:,3]\
+ b2*self.vktable[:,4] + b1*b2*self.vktable[:,5] \
+ bs*self.vktable[:,7] + b1*bs*self.vktable[:,8] + b3 * self.vktable[:,11]\
+ alpha_v * self.kv * self.pzel
_integrand = loginterp(self.kv, self.vkeft)(self.kint)
_integrand = self.weight * _integrand + (1-self.weight) * (-2 * (1+b1) * self.plin/self.kint)
qint, xi = self.sph_gsm.sph(1,_integrand*self.window)
self.veft = np.interp(self.rint, qint, xi)
self.s0keft = self.s0[:,1] + b1*self.s0[:,2] + b1**2*self.s0[:,3]\
+ b2*self.s0[:,4] \
+ bs*self.s0[:,7] \
+ alpha_s0 * self.pzel
self.s2keft = self.s2[:,1] + b1*self.s2[:,2] + b1**2*self.s2[:,3]\
+ b2*self.s2[:,4] \
+ bs*self.s2[:,7] \
+ alpha_s2 * self.pzel
_integrand = loginterp(self.kv, self.s0keft)(self.kint)
_integrand = self.weight * _integrand + (1-self.weight) * (-2./3 * self.plin/self.kint**2)
qint, xi = self.sph_gsm.sph(0,_integrand * self.window)
self.s0eft = np.interp(self.rint, qint, xi)
_integrand = loginterp(self.kv, self.s2keft)(self.kint)
_integrand = self.weight * _integrand + (1-self.weight) * (-4./3 * self.plin/self.kint**2)
qint2, xi = self.sph_gsm.sph(2,_integrand * self.window); xi *=-1
self.s2eft = np.interp(self.rint, qint2, xi)
self.s0eft += (self.Xddot + self.Xloopddot + 2*b1*self.X10ddot + 2*bs*self.Xs2ddot)[-1] + s2fog #add in 0-lag term
def update_power_spectrum(self, k, p):
# Updates the power spectrum and various q functions. Can continually compute for new cosmologies without reloading FFTW
self.k = k
self.p = p
self.pint = loginterp(k, p)(
self.kint) * np.exp(-(self.kint / self.cutoff)**2)
self.setup_powerspectrum()
def xi_l_n(self,
l,
n,
_int=None,
IR_cut='all',
extrap=False,
qmin=1e-3,
qmax=1000,
side='both'):
'''
Calculates the generalized correlation function xi_l_n, which is xi when l = n = 0
If _int is None assume integrating the power spectrum.
'''
if _int is None:
integrand = self.p * self.k**n
else:
integrand = _int * self.k**n
if IR_cut is not 'all':
if IR_cut == 'gt':
integrand *= self.ir_greater
elif IR_cut == 'lt':
integrand *= self.ir_less
qs, xint = self.sph.sph(l, integrand)
if extrap:
qrange = (qs > qmin) * (qs < qmax)
return loginterp(qs[qrange], xint[qrange], side=side)(self.qv)
else:
return np.interp(self.qv, qs, xint)
def __init__(self, *args, kmin = 3e-3, kmax=0.5, nk = 100, kswitch=1e-2, jn = 5, cutoff=20, **kw):
'''
Same keywords and arguments as the other two classes for now.
'''
# Setup ffts etc.
VelocityMoments.__init__(self, *args, third_order=True, **kw)
self.kmin, self.kmax, self.nk = kmin, kmax, nk
self.kv = np.logspace(np.log10(kmin), np.log10(kmax), nk); self.nk = nk
self.kint = np.logspace(-5,3,4000)
self.plin = loginterp(self.k, self.p)(self.kint)
self.sph_gsm = SphericalBesselTransform(self.kint,L=3,fourier=True)
self.rint = np.logspace(-3,5,4000)
self.rint = self.rint[(self.rint>0.1)*(self.rint<600)] #actual range of integration
self.window = tukey(4000)
self.weight = 0.5 * (1 + np.tanh(3*np.log(self.kint/kswitch)))
self.weight[self.weight < 1e-3] = 0
self.pars = np.array([0,0,0,0, 0,0,0,0, 0])
self.peft = None
self.veft = None
self.s0eft = None
self.s2eft = None
self.setup_velocity_moments()
def __init__(self, *args, kmin=1e-2, kmax=0.5, nk=100, **kw):
# Initialize the velocity class
KEVelocityMoments.__init__(self, *args, **kw)
self.nk, self.kmin, self.kmax = nk, kmin, kmax
self.make_tables(kmin=kmin, kmax=kmax, nk=nk)
self.kv = self.pktable[:, 0]
self.plin = loginterp(self.k, self.p)(self.kv)
# Now convert to EPT table
self.convert_ptable()
self.convert_vtable()
self.convert_stable()
if self.beyond_gauss:
self.convert_gtable()
def __init__(self, k, p, third_order = True, shear=True, one_loop=True,\
kIR = None, cutoff=10, jn=5, N = 2000, threads=1, extrap_min = -5, extrap_max = 3):
self.N = N
self.extrap_max = extrap_max
self.extrap_min = extrap_min
self.kIR = kIR
self.cutoff = cutoff
self.kint = np.logspace(extrap_min, extrap_max, self.N)
self.qint = np.logspace(-extrap_max, -extrap_min, self.N)
self.third_order = third_order
self.shear = shear or third_order
self.one_loop = one_loop
self.k = k
self.p = p
self.pint = loginterp(k, p)(
self.kint) * np.exp(-(self.kint / self.cutoff)**2)
self.setup_powerspectrum()
self.pktables = {}
if self.third_order:
self.num_power_components = 13
elif self.shear:
self.num_power_components = 11
else:
self.num_power_components = 7
self.jn = jn
self.threads = threads
self.sph = SphericalBesselTransform(self.qint,
L=self.jn,
ncol=self.num_power_components,
threads=self.threads)
self.sph1 = SphericalBesselTransform(self.qint,
L=self.jn,
ncol=1,
threads=self.threads)
self.sphr = SphericalBesselTransformNP(self.kint, L=5, fourier=True)
def compute_xi_real(self, rr, b1, b2, bs, b3, alpha, alpha_v, alpha_s0, alpha_s2, s2fog):
'''
Compute the real-space correlation function at rr.
'''
# This is just the zeroth moment:
self.kv = self.pktable[:,0]
self.pzel = self.pktable[:,-1]
self.peft = self.pktable[:,1] + b1*self.pktable[:,2] + b1**2*self.pktable[:,3]\
+ b2*self.pktable[:,4] + b1*b2*self.pktable[:,5] + b2**2 * self.pktable[:,6]\
+ bs*self.pktable[:,7] + b1*bs*self.pktable[:,8] + b2*bs*self.pktable[:,9]\
+ bs**2*self.pktable[:,10] + b3*self.pktable[:,11] + b1*b3*self.pktable[:,12] + alpha* self.kv**2 * self.pzel
_integrand = loginterp(self.kv, self.peft)(self.kint) #at some point turn into theory extrapolation
_integrand = self.weight * _integrand + (1-self.weight) * ((1+b1)**2 * self.plin)
qint, xi = self.sph_gsm.sph(0,_integrand)
xir = Spline(qint,xi)(rr)
return xir
def combine_bias_terms_xiell(self,bvec,method='loginterp'):
'''
Same as above but further transform the pkells into xiells.
Again, the paramters f, AP are assumed to be what was input into p{ell}ktable.
'''
kv, p0, p2, p4 = self.combine_bias_terms_pkell(bvec)
if method == 'loginterp':
damping = np.exp(-(self.kint/10)**2)
p0int = loginterp(kv, p0)(self.kint) * damping
p2int = loginterp(kv, p2)(self.kint) * damping
p4int = loginterp(kv, p4)(self.kint) * damping
elif method == 'gauss_poly':
# Add a point at k = 0 to the spline in k taper nicely
frac = 1
p0int = gaussian_poly_extrap( self.kint,\
np.concatenate(([0], kv)),\
np.concatenate(([0], p0)), frac=frac)
p2int = gaussian_poly_extrap( self.kint,\
np.concatenate(([0], kv)),\
np.concatenate(([0], p2)), frac=frac )
p4int = gaussian_poly_extrap( self.kint,\
np.concatenate(([0], kv)),\
np.concatenate(([0], p4)), frac=frac )
elif method == 'min_cut':
# Start log extrapolating when p_ell is below a threshold value:
ftol = 1e-4
damping = np.exp(-(self.kint/10)**2)
pints = [np.zeros_like(self.kint), np.zeros_like(self.kint), np.zeros_like(self.kint),]
for ii, pp in enumerate([p0,p2,p4]):
iis = np.arange(len(kv))
pval = np.max(pp)
try:
zero_crossing = np.where(np.diff(np.sign(pp)))[0][0]
except:
zero_crossing = len(pp)
cross_min = pp > (ftol * pval)
# union is where we interpolate
where_int = (iis < zero_crossing) * cross_min
ktemp, ptemp = kv[where_int], pp[where_int]
pints[ii] += loginterp(ktemp, ptemp)(self.kint) * damping
p0int, p2int, p4int = pints
ss0, xi0 = self.sphr.sph(0,p0int)
ss2, xi2 = self.sphr.sph(2,p2int); xi2 *= -1
ss4, xi4 = self.sphr.sph(4,p4int)
return (ss0, xi0), (ss2, xi2), (ss4, xi4)
def __init__(self,
k,
p,
pnw=None,
*args,
rbao=110,
kmin=1e-2,
kmax=0.5,
nk=100,
**kw):
self.nk, self.kmin, self.kmax = nk, kmin, kmax
self.rbao = rbao
self.ept = EPT(k, p, kmin=kmin, kmax=kmax, nk=nk, **kw)
if pnw is None:
knw = self.ept.kint
Nfilter = np.ceil(
np.log(7) / np.log(knw[-1] / knw[-2])
) // 2 * 2 + 1 # filter length ~ log span of one oscillation from k = 0.01
#print(Nfilter)
pnw = savgol_filter(self.ept.pint, int(Nfilter), 4)
else:
knw, pnw = k, pnw
self.ept_nw = EPT(knw, pnw, kmin=kmin, kmax=kmax, nk=nk, **kw)
self.beyond_gauss = self.ept.beyond_gauss
self.kv = self.ept.kv
self.plin = loginterp(k, p)(self.kv)
self.plin_nw = loginterp(knw, pnw)(self.kv)
self.plin_w = self.plin - self.plin_nw
self.sigma_squared_bao = np.interp(
self.rbao, self.ept_nw.qint,
self.ept_nw.Xlin + self.ept_nw.Ylin / 3.)
self.damp_exp = -0.5 * self.kv**2 * self.sigma_squared_bao
self.damp_fac = np.exp(self.damp_exp)
self.plin_ir = self.plin_nw + self.plin_w * self.damp_exp
self.pktable_nw = self.ept_nw.pktable_ept
self.pktable_w = self.damp_fac[:, None] * (
(self.ept.pktable_ept - self.pktable_nw) - self.damp_exp[:, None] *
(self.ept.pktable_ept_linear - self.ept_nw.pktable_ept_linear))
self.pktable_w[:, 0] = self.kv
self.pktable = self.pktable_nw + self.pktable_w
self.pktable[:, 0] = self.kv
self.vktable_nw = self.ept_nw.vktable_ept
self.vktable_w = self.damp_fac[:, None] * (
(self.ept.vktable_ept - self.vktable_nw) - self.damp_exp[:, None] *
(self.ept.vktable_ept_linear - self.ept_nw.vktable_ept_linear))
self.vktable_w[:, 0] = self.kv
self.vktable = self.vktable_nw + self.vktable_w
self.vktable[:, 0] = self.kv
self.s0ktable_nw = self.ept_nw.s0ktable_ept
self.s0ktable_w = self.damp_fac[:, None] * (
(self.ept.s0ktable_ept - self.s0ktable_nw) -
self.damp_exp[:, None] *
(self.ept.s0ktable_ept_linear - self.ept_nw.s0ktable_ept_linear))
self.s0ktable_w[:, 0] = self.kv
self.s0ktable = self.s0ktable_nw + self.s0ktable_w
self.s0ktable[:, 0] = self.kv
self.s2ktable_nw = self.ept_nw.s2ktable_ept
self.s2ktable_w = self.damp_fac[:, None] * (
(self.ept.s2ktable_ept - self.s2ktable_nw) -
self.damp_exp[:, None] *
(self.ept.s2ktable_ept_linear - self.ept_nw.s2ktable_ept_linear))
self.s2ktable_w[:, 0] = self.kv
self.s2ktable = self.s2ktable_nw + self.s2ktable_w
self.s2ktable[:, 0] = self.kv
if self.beyond_gauss:
self.g1ktable_nw = self.ept_nw.g1ktable_ept
self.g1ktable_w = self.damp_fac[:, None] * (
self.ept.g1ktable_ept - self.ept_nw.g1ktable_ept)
self.g1ktable_w[:, 0] = self.kv
self.g1ktable = self.g1ktable_nw + self.g1ktable_w
self.g1ktable[:, 0] = self.kv
self.g3ktable_nw = self.ept_nw.g3ktable_ept
self.g3ktable_w = self.damp_fac[:, None] * (
self.ept.g3ktable_ept - self.ept_nw.g3ktable_ept)
self.g3ktable_w[:, 0] = self.kv
self.g3ktable = self.g3ktable_nw + self.g3ktable_w
self.g3ktable[:, 0] = self.kv
self.k0_nw, self.k2_nw, self.k4_nw = self.ept_nw.k0, self.ept_nw.k2, self.ept_nw.k4
self.k0_w = self.damp_fac * (self.ept.k0 - self.ept_nw.k0)
self.k2_w = self.damp_fac * (self.ept.k2 - self.ept_nw.k2)
self.k4_w = self.damp_fac * (self.ept.k4 - self.ept_nw.k4)
self.k0 = self.k0_nw + self.k0_w
self.k2 = self.k2_nw + self.k2_w
self.k4 = self.k4_nw + self.k4_w
