def run():
def f2(x,y):
return numx.sin(x) / x
sys2 = integrate.gsl_function(f2, None)
def f1(x,y):
return 1 / x
sys1 = integrate.gsl_function(f1, None)
def f3(x,y):
return 1 / -x
sys3 = integrate.gsl_function(f3, None)
w = integrate.workspace(1000000)
cyclew = integrate.workspace(1000000)
table1 = integrate.qawo_table(1, 100, integrate.SINE, 100)
table2 = integrate.qawo_table(-1, 100, integrate.SINE, 100)
# Borders and singualrity for gagp
pts = numx.array((-numx.pi, 0, numx.pi))
flag, result1, error = integrate.qagp(sys2, pts, 1e-8, 1e-8, 100000, w)
flag, result2, error = integrate.qawf(sys1, numx.pi, 1e-8, 100, w,
cyclew, table1)
flag, result3, error = integrate.qawf(sys3, numx.pi, 1e-8, 100, w,
cyclew, table2)
print "Result of integration is :", result1 + result2 + result3
def run():
def f2(x, y):
return numx.sin(x) / x
sys2 = integrate.gsl_function(f2, None)
def f1(x, y):
return 1 / x
sys1 = integrate.gsl_function(f1, None)
def f3(x, y):
return 1 / -x
sys3 = integrate.gsl_function(f3, None)
w = integrate.workspace(1000000)
cyclew = integrate.workspace(1000000)
table1 = integrate.qawo_table(1, 100, integrate.SINE, 100)
table2 = integrate.qawo_table(-1, 100, integrate.SINE, 100)
# Borders and singualrity for gagp
pts = numx.array((-numx.pi, 0, numx.pi))
flag, result1, error = integrate.qagp(sys2, pts, 1e-8, 1e-8, 100000, w)
flag, result2, error = integrate.qawf(sys1, numx.pi, 1e-8, 100, w, cyclew,
table1)
flag, result3, error = integrate.qawf(sys3, numx.pi, 1e-8, 100, w, cyclew,
table2)
print "Result of integration is :", result1 + result2 + result3
def test_qawf(self):
def f2(x, y):
return Numeric.sin(x) / x
sys2 = integrate.gsl_function(f2, None)
def f1(x, y):
return 1 / x
sys1 = integrate.gsl_function(f1, None)
def f3(x, y):
return 1 / -x
sys3 = integrate.gsl_function(f3, None)
pts = Numeric.array((-Numeric.pi, 0, Numeric.pi))
flag, result1, error = integrate.qagp(sys2, pts, 1e-8, 1e-8, 100000,
self.w)
table1 = integrate.qawo_table(1, 100, integrate.SINE, 100)
cyclew = integrate.workspace(1000000)
flag, result2, error = integrate.qawf(sys1, Numeric.pi, 1e-8, 100,
self.w, cyclew, table1)
table2 = integrate.qawo_table(-1, 100, integrate.SINE, 100)
flag, result3, error = integrate.qawf(sys3, Numeric.pi, 1e-8, 100,
self.w, cyclew, table2)
assert (Numeric.absolute(result1 + result2 + result3 - Numeric.pi) <
1e-8)
def test_qawf(self):
def f2(x,y):
return Numeric.sin(x) / x
sys2 = integrate.gsl_function(f2, None)
def f1(x,y):
return 1 / x
sys1 = integrate.gsl_function(f1, None)
def f3(x,y):
return 1 / -x
sys3 = integrate.gsl_function(f3, None)
pts = Numeric.array((-Numeric.pi, 0, Numeric.pi))
flag, result1, error = integrate.qagp(sys2, pts, 1e-8, 1e-8, 100000, self.w)
table1 = integrate.qawo_table(1, 100, integrate.SINE, 100)
cyclew = integrate.workspace(1000000)
flag, result2, error = integrate.qawf(sys1, Numeric.pi, 1e-8, 100, self.w, cyclew, table1)
table2 = integrate.qawo_table(-1, 100, integrate.SINE, 100)
flag, result3, error = integrate.qawf(sys3, Numeric.pi, 1e-8, 100, self.w, cyclew, table2)
assert(Numeric.absolute(result1+result2+result3 - Numeric.pi) < 1e-8)
def getage(self,z):
"""Look-back time as a function of redhshift. Returns time in tunit.
Hogg paper Equation 30.
"""
def agefunc(z2,y):
return 1./ (self.epeebles(z2)*(1.+z2))
#gsl baggage
gsl_agefunc=integ.gsl_function(agefunc,None) #extra pygsl baggage
w=integ.workspace(1000000)
try: #check if z is list or array
output=[]
for zz in z:
flag,result,error=integ.qagiu(gsl_agefunc,zz,1e-8,1e-8,100000,w)
output.append(self.thubble()*result)
output=numpy.array(output)
return output
except TypeError: #z is single value
flag,result,error=integ.qagiu(gsl_agefunc,z,1e-8,1e-8,100000,w)
output=self.thubble()*result
return output
def setUp(self):
self.w = integrate.workspace(1000000)
def main():
parser = argparse.ArgumentParser(description='Simulation of selected and neutral site frequency spectra using the Poisson Random Field Model (PRF)')
parser.add_argument('-n', '--nsam', type=int, required=True, dest='sample', help='The sample size')
parser.add_argument('-t', '--theta', type=float, required=True, dest='theta', help='The scaled mutation parameter per site')
parser.add_argument('-g', '--gamma', type=float, required=True, dest='gamma', help='The scaled selection coefficient (Ns). Can be positive or negative')
parser.add_argument('-N', '--Nsites', required=True, type=int, dest='length_S', help = 'Size of region for neutral sites in base pairs to simulate using the PRF [default: 500] ')
parser.add_argument('-S', '--Ssites', required=True, type=int, dest='length_N', help = 'Size of region for selected sites in base pairs to simulate using the PRF [default: 500] ')
parser.add_argument('-o', '--out', required=True, dest='outfile', help = 'Output file for simulated Neutral and Selected sites site frequency spectra')
parser.add_argument('-s', '--sfs', default='folded', required=False, dest='fold', help = 'Specify folded or unfolded [ default: folded]')
args = parser.parse_args()
n = args.sample
Ls_N = args.length_N
theta_ls_N = args.theta * Ls_N
Ls_S = args.length_S
theta_ls_S = args.theta * Ls_S # specified theta per site at input, so multiply by length
gamma = args.gamma # gamma in Kai's program is twice that in the Bustamante paper so multiply by a half
if args.fold == 'folded':
fold = True
elif args.fold == 'unfolded':
fold = False
else:
print('Invalid option for -s, it should be folded or unfolded')
sys.exit(1)
with open(args.outfile, 'w') as outfile:
nl = 1 # consider adding option to have multiple replicate of varying size output
sfs = []
workspace = integrate.workspace(1000000)
if fold:
print('S', file=outfile)
print(nl, file=outfile)
for i in xrange(1, (n + 2) / 2):
if gamma == 0: # still print "Selected" sites in output when gamma is zero
if i < (n - i):
nci = n_choose_i(n, i)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, i])
F_i_gamma = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
nci = n_choose_i(n, n - i)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, n - i])
F_n_minus_i_gamma = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
G_i_gamma = F_i_gamma + F_n_minus_i_gamma # equation 6 condition i < n - i
xN = get_sfs(G_i_gamma, theta_ls_N)
print(i, G_i_gamma*theta_ls_N)
sfs.append(str(xN))
elif i == n / 2:
nci = n_choose_i(n, n / 2)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, n / 2])
F_n_over_two_gamma = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
xN = get_sfs(F_n_over_two_gamma, theta_ls_N)
print(i, F_n_over_two_gamma * theta_ls_N)
sfs.append(str(xN))
elif gamma < 0:
if i < (n - i):
nci = n_choose_i(n, i)
integrand_eq2_func = integrate.gsl_function(integrand_eq2_neg, [gamma, nci, n, i])
F_i_gamma = integrate.qag(integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
nci = n_choose_i(n, n - i)
integrand_eq2_func = integrate.gsl_function(integrand_eq2_neg, [gamma, nci, n, n - i])
F_n_minus_i_gamma = integrate.qag(integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
G_i_gamma = F_i_gamma + F_n_minus_i_gamma # equation 6 condition i < n - i
#print(G_i_gamma, F_i_gamma, F_n_minus_i_gamma)
xS = get_sfs(G_i_gamma, theta_ls_S)
print(i, G_i_gamma * theta_ls_S)
sfs.append(str(xS))
elif i == n / 2: # floor division in python 2
nci = n_choose_i(n, n / 2)
integrand_eq2_func = integrate.gsl_function(integrand_eq2_neg, [gamma, nci, n, n / 2])
F_n_over_two_gamma = integrate.qag(integrand_eq2_func, 0, 1, 1e-10, 1e-10, 1000, integrate.GAUSS61, workspace)[1]
xS = get_sfs(F_n_over_two_gamma , theta_ls_S)
#print(F_n_over_two_gamma)
print(i, F_n_over_two_gamma * theta_ls_S)
sfs.append(str(xS))
else:
if i < (n - i):
nci = n_choose_i(n, i)
integrand_eq2_func = integrate.gsl_function(integrand_eq2, [gamma, nci, n, i])
F_i_gamma = integrate.qag(integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
nci = n_choose_i(n, n - i)
integrand_eq2_func = integrate.gsl_function(integrand_eq2, [gamma, nci, n, n - i])
F_n_minus_i_gamma = integrate.qag(integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
G_i_gamma = F_i_gamma + F_n_minus_i_gamma # equation 6 condition i < n - i
xS = get_sfs(G_i_gamma, theta_ls_S)
print(i, G_i_gamma * theta_ls_S)
sfs.append(str(xS))
elif i == n / 2: # floor division in python 2
nci = n_choose_i(n, n / 2)
integrand_eq2_func = integrate.gsl_function(integrand_eq2, [gamma, nci, n, n / 2])
F_n_over_two_gamma = integrate.qag(integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
xS = get_sfs(F_n_over_two_gamma , theta_ls_S)
print(i, F_n_over_two_gamma * theta_ls_S)
sfs.append(str(xS))
sfs_string = ', '.join(sfs)
print(n, Ls_S, sfs_string, sep=', ', file=outfile)
del sfs[:]
# Neutral sites
print('N', file=outfile)
print(nl, file=outfile)
for i in xrange(1, (n + 2) / 2):
if i < (n - i):
nci = n_choose_i(n, i)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, i])
F_i_gamma = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
nci = n_choose_i(n, n - i)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, n - i])
F_n_minus_i_gamma = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
G_i_gamma = F_i_gamma + F_n_minus_i_gamma # equation 6 condition i < n - i
xN = get_sfs(G_i_gamma, theta_ls_N)
#print(i, G_i_gamma * theta_ls_N)
sfs.append(str(xN))
elif i == n / 2:
nci = n_choose_i(n, n / 2)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, n / 2])
F_n_over_two_gamma = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
xN = get_sfs(F_n_over_two_gamma, theta_ls_N)
#print(i, F_n_over_two_gamma * theta_ls_N)
sfs.append(str(xN))
sfs_string = ', '.join(sfs)
print(n, Ls_N, sfs_string, sep=', ', file=outfile)
del sfs[:]
else: # produce unfolded sfs
print('S', file=outfile)
print(nl, file=outfile)
for i in xrange(1, n):
if gamma == 0:
nci = n_choose_i(n , i)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, i])
eq2_eval_neut = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
xN = get_sfs(eq2_eval_neut, theta_ls_N)
# print(i, xN)
sfs.append(str(xN))
elif gamma < 0:
nci = n_choose_i(n, i)
integrand_eq2_func = integrate.gsl_function(integrand_eq2_neg, [gamma, nci, n, i])
eq2_eval_sel = integrate.qag(integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
xS = get_sfs(eq2_eval_sel, theta_ls_S)
# print(i, eq2_eval_sel)
sfs.append(str(xS))
else:
nci = n_choose_i(n, i)
integrand_eq2_func = integrate.gsl_function(integrand_eq2, [gamma, nci, n, i])
eq2_eval_sel = integrate.qag(integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
xS = get_sfs(eq2_eval_sel, theta_ls_S)
# print(i, eq2_eval_sel)
sfs.append(str(xS))
sfs_string = ', '.join(sfs)
print(n, Ls_S, sfs_string, sep=', ', file=outfile)
del sfs[:]
print('N', file=outfile)
print(nl, file=outfile)
for i in xrange(1, n):
nci = n_choose_i(n , i)
neutral_integrand_eq2_func = integrate.gsl_function(neutral_integrand_eq2, [nci, n, i])
eq2_eval_neut = integrate.qag(neutral_integrand_eq2_func, 0, 1, 1e-8, 1e-8, 1000, integrate.GAUSS61, workspace)[1]
xN = get_sfs(eq2_eval_neut, theta_ls_N)
# print(i, xN)
sfs.append(str(xN))
sfs_string = ', '.join(sfs)
print(n, Ls_N, sfs_string, sep=', ', file=outfile)
del sfs[:]
from PMFpack import * from numpy import * from pylab import * from pygsl import integrate from scipy.integrate import quad, dblquad, Inf from scipy.special import erf, erfinv w1 = integrate.workspace(10000) w2 = integrate.workspace(10000) #fmean = 0.639756248702123 #sigma = 0.16674922174125 #pmf = PMF(fmean, sigma, 0, 0, 0, 2) #pmf.set_fmean_gradient(-0.4178939524532, 0, 0) # Last two are zero because the simulation is 1D #pmf.set_sigma_gradient(-0.5095137299633, 0, 0) #pmf.chi = 0.38430391476813 # Mean scalar dissipation rate (defined with the factor 2) #fmean = 0.32734785270559791 #sigma = 0.036258388156940768 #pmf = PMF(fmean, sigma, 0, 0, 0, 2) #pmf.set_fmean_gradient(-0.40312611148071925, 0, 0) # Last two are zero because the simulation is 1D #pmf.set_sigma_gradient(-0.29357760186462173, 0, 0) #pmf.chi = 0.36258388156940158 # Mean scalar dissipation rate (defined with the factor 2) #fmean = 0.67265214729440204 #sigma = 0.036258388156940768 #pmf = PMF(fmean, sigma, 0, 0, 0, 2) #pmf.set_fmean_gradient(-0.40312611148071925, 0, 0) # Last two are zero because the simulation is 1D #pmf.set_sigma_gradient(-0.51267462109681816, 0, 0) #pmf.chi = 0.36258388156940158 # Mean scalar dissipation rate (defined with the factor 2)
def GetSelfEnergyRPADummyMomFrequency(n, omega, kappa, L, asymptoteParams, nbrRTraining=40, orderInterpolate=3, nbrRTest=200, workspaceSize=100, maxBisections=100, epsilonQuad=1.0e-07,
epsilonRelQuad=1.0e-04, useInterpolator=True, verbose=False, findEnergy=False):
"""n: vector of integers characterizing external momentum"""
"""omega: external frequency OR energy"""
"""kappa: hopping (eV)"""
"""L: vector of integers characterizing size of lattice"""
"""asymptoteParams: tensor of fit parameter characterizing asymptote of \delta \Pi at large frequency"""
"""nbrRTaining: number of points for training interpolator for integrand"""
"""nbrRTest: number of points to test integrand"""
"""workspaceSize: GSL integration workspace size"""
"""maxBisections: GSL integration number of subdivisions"""
"""epsilonQuad: absolute error for integration"""
"""epsilonRelQuad: relative error for integration"""
"""useInterpolator: use the interpolation routine or explicit evalaution of \delta \Pi"""
"""verbose: output integrand?"""
"""findEnergy: searching for energy? i.e. omega \to i\omega (Wick rotation)"""
assert(len(n)==2)
assert(len(L)==2)
assert(len(asymptoteParams.shape)==2)
assert(asymptoteParams.shape[0]==L[0])
assert(asymptoteParams.shape[1]==L[1])
prefactor = -1/(4*np.pi*L[0]*L[1])
resultIntegration = np.zeros((int(L[0]),int(L[1]),4))
work = integrate.workspace(workspaceSize) #create workspace for integration
bareSigma = GetSelfEnergyBareDummyMom(n,kappa,L)
result = 0.0
for n1 in range(int(L[0])): # sum over spatial loop momentum \vec{q}
for n2 in range(int(L[1])):
# integrate over q_0 for each loop momentum \vec{q}
nLoop = np.copy([n1,n2])
print 'GetSelfEnergyRPADummyMomFrequency nLoop:',nLoop
pPlusLoop = n+nLoop[:]
pMinusLoop = n-nLoop[:]
factorPlusLoop = ComputeStructureFactorHexagonalNN(pPlusLoop,L)
factorMinusLoop = ComputeStructureFactorHexagonalNN(pMinusLoop,L)
w = asymptoteParams[n1,n2] # get parameter which defines change of variables
assert(w>0)
rMax = 0.5*np.pi/w
if useInterpolator:
rIntegrandTrain = np.linspace(-rMax,rMax,num=nbrRTraining,endpoint=True)
integrandTrain = np.zeros((nbrRTraining,2))
if verbose:
rIntegrandTest = np.linspace(-rMax,rMax,num=nbrRTest,endpoint=True)
integrandData = np.zeros((nbrRTest,2))
if np.abs(factorPlusLoop) > 1.0e-10:
if findEnergy:
numReal = lambda r: GetSigmaRPAIntegrandWeight(r,w)*((np.abs(factorPlusLoop)**2-omega**2+w*np.tan(r*w))*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).real
+ 2*omega*w*np.tan(r*w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).imag)
denom = lambda r: ((1j*omega+w*np.tan(r*w))**2+np.abs(factorPlusLoop)**2)*((-1j*omega+w*np.tan(r*w))**2+np.abs(factorPlusLoop)**2)
realIntegrand1 = lambda r: numReal(r)/denom(r)
numImag = lambda r: GetSigmaRPAIntegrandWeight(r,w)*((np.abs(factorPlusLoop)**2-omega**2+w*np.tan(r*w))*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).imag
- 2*omega*w*np.tan(r*w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).real)
imagIntegrand1 = lambda r: numImag(r)/denom(r)
else:
# \delta \Pi_{1,2}(q_0, \vec{q}) G_{(0);2,1}(p_0+q_0,\vec{p}+\vec{q})
realIntegrand1 = lambda r: GetSigmaRPAIntegrandWeight(r,w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).real/((omega+w*np.tan(r*w))**2 + (kappa*np.abs(factorPlusLoop))**2)
imagIntegrand1 = lambda r: GetSigmaRPAIntegrandWeight(r,w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).imag/((omega+w*np.tan(r*w))**2 + (kappa*np.abs(factorPlusLoop))**2)
if useInterpolator:
for i in range(nbrRTraining):
integrandTrain[i,0] = realIntegrand1(rIntegrandTrain[i])
integrandTrain[i,1] = imagIntegrand1(rIntegrandTrain[i])
realIntegrand1Interp = InterpolationWrappers.MyUnivariateSpline(x=rIntegrandTrain,y=integrandTrain[:,0],name="realIntegrand1",order=orderInterpolate,boundsError=True,verbose=True)
imagIntegrand1Interp = InterpolationWrappers.MyUnivariateSpline(x=rIntegrandTrain,y=integrandTrain[:,1],name="imagIntegrand1",order=orderInterpolate,boundsError=True,verbose=True)
realIntFuc = integrate.gsl_function(MakeIntFunction(realIntegrand1Interp.EvaluateUnivariateInterpolationPoint), None) #set function to be integrated (real)
imagIntFuc = integrate.gsl_function(MakeIntFunction(imagIntegrand1Interp.EvaluateUnivariateInterpolationPoint), None) #set function to be integrated (imag)
else:
realIntFuc = integrate.gsl_function(MakeIntFunction(realIntegrand1), None) #set function to be integrated (real)
imagIntFuc = integrate.gsl_function(MakeIntFunction(imagIntegrand1), None) #set function to be integrated (imag)
flag, resultIntegration[n1,n2,0], error = integrate.qag(realIntFuc, -rMax, rMax, epsilonQuad, epsilonRelQuad, maxBisections, integrate.GAUSS61, work)
flag, resultIntegration[n1,n2,1], error = integrate.qag(imagIntFuc, -rMax, rMax, epsilonQuad, epsilonRelQuad, maxBisections, integrate.GAUSS61, work)
if verbose:
for i in range(nbrRTest):
if useInterpolator:
integrandData[i,0] = realIntegrand1Interp.EvaluateUnivariateInterpolationPoint(rIntegrandTest[i])
integrandData[i,1] = imagIntegrand1Interp.EvaluateUnivariateInterpolationPoint(rIntegrandTest[i])
else:
integrandData[i,0] = realIntegrand1(rIntegrandTest[i])
integrandData[i,1] = imagIntegrand1(rIntegrandTest[i])
Output1DGrid("Integrand1_kappa_%g_alpha_%g_P_%d_%d_Q_%d_%d_omega_%g_L_%d_nbrR_%d"%(kappa,alpha,n[0],n[1],nLoop[0],nLoop[1],omega,L[0],nbrRTest),
rIntegrandTest,integrandData)
result += -kappa*np.conj(factorPlusLoop)*(resultIntegration[n1,n2,0]+1j*resultIntegration[n1,n2,1])
filename = "RPA_INTEGRAL1_P_%d_%d_Q_%d_%d_omega_%e"%(n[0],n[1],nLoop[0],nLoop[1],omega)
checkExistingFile(filename)
if verbose:
with open(filename, "w") as text_file:
if useInterpolator:
text_file.write("interpolator_result: %.16e %.16e\n"%(resultIntegration[n1,n2,0],resultIntegration[n1,n2,1]))
else:
text_file.write("exact_result: %.16e %.16e\n"%(resultIntegration[n1,n2,0],resultIntegration[n1,n2,1]))
text_file.close()
if np.abs(factorMinusLoop) > 1.0e-10:
if findEnergy:
numReal = lambda r: GetSigmaRPAIntegrandWeight(r,w)*((np.abs(factorMinusLoop)**2-omega**2+w*np.tan(r*w))*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).real
- 2*omega*w*np.tan(r*w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).imag)
denom = lambda r: ((1j*omega-w*np.tan(r*w))**2+np.abs(factorMinusLoop)**2)*((-1j*omega+w*np.tan(r*w))**2-np.abs(factorMinusLoop)**2)
realIntegrand2 = lambda r: numReal(r)/denom(r)
numImag = lambda r: GetSigmaRPAIntegrandWeight(r,w)*((np.abs(factorMinusLoop)**2-omega**2+w*np.tan(r*w))*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).imag
+ 2*omega*w*np.tan(r*w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).real)
imagIntegrand2 = lambda r: numImag(r)/denom(r)
else:
# \delta \Pi_{2,1}(q_0, \vec{q}) G_{(0);2,1}(p_0-q_0,\vec{p}+\vec{q})
realIntegrand2 = lambda r: GetSigmaRPAIntegrandWeight(r,w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).real/((omega-w*np.tan(r*w))**2 + (kappa*np.abs(factorMinusLoop))**2)
imagIntegrand2 = lambda r: -GetSigmaRPAIntegrandWeight(r,w)*GetDifferenceDummyBosonPropagator(nLoop,w*np.tan(r*w),kappa,alpha,L).imag/((omega-w*np.tan(r*w))**2 + (kappa*np.abs(factorMinusLoop))**2)
if useInterpolator:
for i in range(nbrRTraining):
integrandTrain[i,0] = realIntegrand2(rIntegrandTrain[i])
integrandTrain[i,1] = imagIntegrand2(rIntegrandTrain[i])
realIntegrand2Interp = InterpolationWrappers.MyUnivariateSpline(x=rIntegrandTrain,y=integrandTrain[:,0],name="realIntegrand2",order=orderInterpolate,boundsError=True,verbose=True)
imagIntegrand2Interp = InterpolationWrappers.MyUnivariateSpline(x=rIntegrandTrain,y=integrandTrain[:,1],name="imagIntegrand2",order=orderInterpolate,boundsError=True,verbose=True)
realIntFuc = integrate.gsl_function(MakeIntFunction(realIntegrand2Interp.EvaluateUnivariateInterpolationPoint), None) #set function to be integrated (real)
imagIntFuc = integrate.gsl_function(MakeIntFunction(imagIntegrand2Interp.EvaluateUnivariateInterpolationPoint), None) #set function to be integrated (imag)
else:
realIntFuc = integrate.gsl_function(MakeIntFunction(realIntegrand2), None) #set function to be integrated (real)
imagIntFuc = integrate.gsl_function(MakeIntFunction(imagIntegrand2), None) #set function to be integrated (imag)
flag, resultIntegration[n1,n2,2], error = integrate.qag(realIntFuc, -rMax, rMax, epsilonQuad, epsilonRelQuad, maxBisections, integrate.GAUSS61, work)
flag, resultIntegration[n1,n2,3], error = integrate.qag(imagIntFuc, -rMax, rMax, epsilonQuad, epsilonRelQuad, maxBisections, integrate.GAUSS61, work)
if verbose:
for i in range(nbrRTest):
if useInterpolator:
integrandData[i,0] = realIntegrand2Interp.EvaluateUnivariateInterpolationPoint(rIntegrandTest[i])
integrandData[i,1] = imagIntegrand2Interp.EvaluateUnivariateInterpolationPoint(rIntegrandTest[i])
else:
integrandData[i,0] = realIntegrand2(rIntegrandTest[i])
integrandData[i,1] = imagIntegrand2(rIntegrandTest[i])
Output1DGrid("Integrand2_kappa_%g_alpha_%g_P_%d_%d_Q_%d_%d_omega_%g_L_%d_nbrR_%d"%(kappa,alpha,n[0],n[1],nLoop[0],nLoop[1],omega,L[0],nbrRTest),
rIntegrandTest,integrandData)
result += -kappa*np.conj(factorMinusLoop)*(resultIntegration[n1,n2,2]+1j*resultIntegration[n1,n2,3])
filename = "RPA_INTEGRAL2_P_%d_%d_Q_%d_%d_omega_%e"%(n[0],n[1],nLoop[0],nLoop[1],omega)
checkExistingFile(filename)
if verbose:
with open(filename, "w") as text_file:
if useInterpolator:
text_file.write("interpolator_result: %.16e %.16e\n"%(resultIntegration[n1,n2,2],resultIntegration[n1,n2,3]))
else:
text_file.write("exact_result: %.16e %.16e\n"%(resultIntegration[n1,n2,2],resultIntegration[n1,n2,3]))
text_file.close()
return prefactor*result+bareSigma
from PMFpack import * from numpy import * from pylab import * from pygsl import integrate from scipy.integrate import quad, dblquad, Inf from scipy.special import erf, erfinv w1 = integrate.workspace(10000) w2 = integrate.workspace(10000) #fmean = 0.639756248702123 #sigma = 0.16674922174125 #pmf = PMF(fmean, sigma, 0, 0, 0, 2) #pmf.set_fmean_gradient(-0.4178939524532, 0, 0) # Last two are zero because the simulation is 1D #pmf.set_sigma_gradient(-0.5095137299633, 0, 0) #pmf.chi = 0.38430391476813 # Mean scalar dissipation rate (defined with the factor 2) #fmean = 0.32734785270559791 #sigma = 0.036258388156940768 #pmf = PMF(fmean, sigma, 0, 0, 0, 2) #pmf.set_fmean_gradient(-0.40312611148071925, 0, 0) # Last two are zero because the simulation is 1D #pmf.set_sigma_gradient(-0.29357760186462173, 0, 0) #pmf.chi = 0.36258388156940158 # Mean scalar dissipation rate (defined with the factor 2) #fmean = 0.67265214729440204 #sigma = 0.036258388156940768 #pmf = PMF(fmean, sigma, 0, 0, 0, 2) #pmf.set_fmean_gradient(-0.40312611148071925, 0, 0) # Last two are zero because the simulation is 1D #pmf.set_sigma_gradient(-0.51267462109681816, 0, 0) #pmf.chi = 0.36258388156940158 # Mean scalar dissipation rate (defined with the factor 2)
def setUp(self):
self.w = integrate.workspace(1000000)
