def wrap_func(
sigma,
dec_inv_frq,
dec_interp,
bounds,
nbins,
bragg,
ird,
nfrq):
"""
Wrap ex_consistency and return x, y
y = weight sigma - dsa sigma
"""
from RS.rs_ex import ex_consistency as func
myrs, flow_weight, flow_dsa = func(
sigma=sigma,
dec_inv_frq=dec_inv_frq,
sin2psimx=bounds[1],
psi_nbin=nbins,
dec_interp=dec_interp,
theta_b=bragg,
ird=ird,
nfrq=nfrq,
iscatter=True,
ig_sub=True,
ilog=True,
iplot=False, # iplot=True
)
## My objects that quantify the propagated
## error to stress.
if type(nfrq).__name__=='NoneType':
epsilon_vm = flow_weight.epsilon_vm[::]
sigma_vm = flow_weight.sigma_vm[::]
elif type(nfrq).__name__=='int':
epsilon_vm = flow_weight.epsilon_vm[::nfrq]
sigma_vm = flow_weight.sigma_vm[::nfrq]
if len(epsilon_vm) != len(sigma_vm):
raise IOError, 'epsilon_vm and '+\
'sigma_vm should have the same dimension'
if len(epsilon_vm) != len(flow_dsa.sigma_vm):
raise IOError, 'epsilon_vm and '+\
'flow_dsa.sigma_vm should have the same dimension'
x = epsilon_vm
y = (sigma_vm - flow_dsa.sigma_vm)/\
sigma_vm
return x, y
def wrap_func(
sigma,
dec_inv_frq,
dec_interp,
bounds,
nbins,
bragg,
ird):
"""
Wrap ex_consistency and return x, y
y = weight sigma - dsa sigma
"""
from RS.rs_ex import ex_consistency as func
myrs, flow_weight, flow_dsa = func(
sigma=sigma,
dec_inv_frq=dec_inv_frq,
sin2psimx=bounds[1],
psi_nbin=nbins,
dec_interp=dec_interp,
theta_b=bragg,
ird=ird,
iscatter=True,
ig_sub=True,
iplot=False, # iplot=True
)
## My objects that quantify the propagated
## error to stress.
x = flow_weight.epsilon_vm
y = (flow_weight.sigma_vm - flow_dsa.sigma_vm)/\
flow_weight.sigma_vm
return x, y
def influence_of_cnts_stats(
## characteristics of an ensemble for stress data
sigmas=[1e-5, 2e-5, 5e-5, 1e-4],
bounds=[0.,0.5],
ss=3,
nbins=10,
iwgt=False,
nsample=4,
intp_opt=0,
iplot=False,
## diffraction condition
bragg=78.2, ## Fe {211} using Cr K-alpha
ird=0.182, ## Intensity of random distribution
## for {211} using window of 10 degree.
DEC_freq_sym=True,
NCPU=0):
"""
Influence of counting statistics uncertainty
With fixing other variables, investigates the
propagation of counting stat error on to the final
diffraction stress by examining a number of
statistical ensembles (nsample) characterized
by given arguments
Arguments
=========
sigmas : Pure CSE
bounds : Bounds of tilting range
ss : Sampling frequency in DECs
nbins : How many tilting angles in diffraction
iwgt : Do we weight (E-e) by peak intensity?
nsample : The number of ensembles.
intp_opt : DEC interpolation method
iplot : Flag for plotting
bragg : Bragg's angle
ird : Intensity of random distribution
DEC_freq_sym: Plotting option for where DECs are sampled.
NCPU : The number of CPU cores allowed to run
"""
from RS.rs_ex import ex_consistency as func
if iplot:
import matplotlib.pyplot as plt
from MP.lib import mpl_lib,axes_label
fancy_legend = mpl_lib.fancy_legend
wide_fig = mpl_lib.wide_fig
deco = axes_label.__deco__
fig = wide_fig(nw=2,nh=1)
ax1,ax2 = fig.axes[0], fig.axes[1]
print '\n\n****************'
print 'test run started'
print '****************\n\n'
myrs,flow1,flow2 =func(
sin2psimx=bounds[1],
iscatter=True,
sigma=sigmas[0],
psi_nbin=nbins,
ig_sub=True,
iplot=False, # iplot=True
dec_inv_frq=ss,
dec_interp=intp_opt,
)
print '\n\n**************'
print 'test completed'
print '**************\n\n'
# return ## debugging
Y_all = np.zeros((len(sigmas), nsample, flow1.nstp))
M = []
ls=['-+','-s','-o','-d','-t']
import multiprocessing
from multiprocessing import Pool
if NCPU==0: NCPU = multiprocessing.cpu_count()
print 'NCPU: %i'%NCPU
pool = Pool(processes = NCPU)
## function is now wrap_func
results = []
for i in range(len(sigmas)):
results.append([])
for j in range(nsample):
results[i].append(
pool.apply_async(
wrap_func,
args=(
sigmas[i],
ss,
intp_opt,
bounds,
nbins
)
,))
## close/join
pool.close()
pool.join()
## terminate
pool.terminate()
## below is to post-process the results
for i in range(len(sigmas)):
for j in range(nsample):
x,y = results[i][j].get()
Y_all[i][j][:] = y[::]
M = np.zeros((len(sigmas),len(x)))
S = np.zeros((len(sigmas),len(x)))
for i in range(len(sigmas)):
y = Y_all[i][:][:]
y = y.T[::] ## len(x), nsample
for k in range(len(x)):
M[i][k] = y[k].mean()
S[i][k] = y[k].std()
nbins = 10
H = np.zeros((len(sigmas), len(x), nbins))
BE = np.zeros((len(sigmas), len(x), nbins+1))
for i in range(len(sigmas)):
y = Y_all[i][:][:]
y = y.T[::] ## len(x), nsample
for k in range(len(x)):
hist,bin_edges = np.histogram(y[k],bins=10)
H[i,k,:] = hist[::]
BE[i,k,:] = bin_edges[::]
if iplot:
for i in range(len(sigmas)):
ax1.plot(x,M[i],ls[i],mfc='None',color='k',label='%6.0e'%sigmas[i])
ax2.errorbar(x,M[i],yerr=S[i],color='k',ls=ls[i])
if type(ss).__name__=='int' and DEC_freq_sym:
ax1.plot(x[::ss],np.zeros((len(x[::ss]),)),'o',mec='r',mfc='None',ms=8)
ax2.plot(x[::ss],np.zeros((len(x[::ss]),)),'o',mec='r',mfc='None',ms=8)
elif type(ss).__name__=='list' and DEC_freq_sym:
ax1.plot(x[ss],np.zeros((len(x[ss]),)),'o',mec='r',mfc='None',ms=8)
ax2.plot(x[ss],np.zeros((len(x[ss]),)),'o',mec='r',mfc='None',ms=8)
ax1.set_ylim(0.,); ax1.set_xlim(0.,ax1.get_xlim()[1]*1.05)
deco(iopt=9,ft=15,ax=ax1)
fancy_legend(ax=ax1, size=7,ncol=2,nscat=1)
ax2.set_ylim(0.,); ax2.set_xlim(0.,ax2.get_xlim()[1]*1.05)
deco(iopt=9,ft=15,ax=ax2)
fig.savefig('ee.pdf')
return x, M, S, H, BE
def influence_of_cnts_stats(
## characteristics of an ensemble for
## stress data
sigmas=[1e-5, 2e-5, 5e-5, 1e-4],
bounds=[0.,0.5],
ss=3,
nbins=10,
iwgt=False,
nsample=4,
intp_opt=0,
iplot=False,
## diffraction condition
bragg=78.2*np.pi/180., ## Fe {211} using Cr K-alpha
#bragg=162.2*np.pi/180., ## Fe {310} using Cobalt Radiation.
ird=0.182, ## Intensity of random distribution
## for {211} using window of 10 degree.
## sample a portion of data.
nfrq=None,
DEC_freq_sym=True,
NCPU=0):
"""
Propagate errors due to
1) counting statistics error (CSE)
2) incomplete DECs (by utilizing only
a portion of model DECs)
3) Infinite number of tilting angles in use
4) Incomplete range of psi angle ranges in use
5) Geometric distortion (Bragg angle / psi)
6) Peak intensity: I(phi,psi) / Ird
With fixing other variables, investigates the
propagation of counting stat error to final
diffraction stress by examining a number of
statistical ensembles <nsample> characterized
by given arguments
## diffraction condition
#bragg=78.2, ## Fe {211} using Cr K-alpha
#bragg=78.2, ## Fe {310} using Cobalt
#ird=0.182, ## Intensity of random distribution
Arguments
=========
sigmas : Counting Stat Error
bounds : Bounds of tilting range
(low psi to high psi)
ss : Sampling frequency in DECs
nbins : How many tilting angles in
diffraction
iwgt : Do we weight (E-e) by peak intensity?
nsample : The number of ensembles
intp_opt : DEC interpolation method
iplot : Flag for plotting
bragg : Bragg's angle
ird : Intensity of random distribution
nfrq : if not (None) but an integer,
use it as denominator to 'continue' to skip
--------------------------
if istp % nfrq != 0 :
continue ! skip
else:
run stress analysis
--------------------------
DEC_freq_sym: Plotting option for where DECs are sampled.
NCPU : The number of CPU cores allowed to run
"""
from RS.rs_ex import ex_consistency as func
if iplot:
import matplotlib.pyplot as plt
from MP.lib import mpl_lib,axes_label
fancy_legend = mpl_lib.fancy_legend
wide_fig = mpl_lib.wide_fig
deco = axes_label.__deco__
fig = wide_fig(nw=2,nh=1)
ax1,ax2 = fig.axes[0], fig.axes[1]
print '\n\n****************'
print 'test run started'
print '****************\n\n'
myrs,flow_raw,flow2 =func(
sin2psimx=bounds[1],
iscatter=False,
sigma=sigmas[0],
psi_nbin=nbins,
ig_sub=True,
iplot=False, # iplot=True
dec_inv_frq=ss,
dec_interp=intp_opt,
nfrq=nfrq)
print '\n\n**************'
print 'test completed'
print '**************\n\n'
Y_all = np.zeros((len(sigmas), nsample, flow2.nstp))
M = []
ls=['-+','-s','-o','-d','-t']
import multiprocessing
from multiprocessing import Pool
if NCPU==0: NCPU = multiprocessing.cpu_count()
print 'NCPU: %i'%NCPU
pool = Pool(processes = NCPU)
## function is now wrap_func
results = []
for i in xrange(len(sigmas)):
results.append([])
for j in xrange(nsample):
results[i].append(
pool.apply_async(
wrap_func,
args=(
sigmas[i],
ss,
intp_opt,
bounds,
nbins,
bragg,
ird,nfrq),))
## close/join
pool.close(); pool.join()
## terminate
pool.terminate()
## below is to post-process the results
for i in xrange(len(sigmas)):
for j in xrange(nsample):
x,y = results[i][j].get()
Y_all[i][j][:] = y[::]
M = np.zeros((len(sigmas),len(x)))
S = np.zeros((len(sigmas),len(x)))
for i in xrange(len(sigmas)):
y = Y_all[i][:][:]
y = y.T[::] ## len(x), nsample
for k in xrange(len(x)):
M[i][k] = y[k].mean()
S[i][k] = y[k].std()
nbins = 10
H = np.zeros((len(sigmas), len(x), nbins))
BE = np.zeros((len(sigmas), len(x), nbins+1))
for i in xrange(len(sigmas)):
y = Y_all[i][:][:]
y = y.T[::] ## len(x), nsample
for k in xrange(len(x)):
hist,bin_edges = np.histogram(y[k],bins=10)
H[i,k,:] = hist[::]
BE[i,k,:] = bin_edges[::]
if iplot:
for i in xrange(len(sigmas)):
ax1.plot(x,M[i],ls[i],mfc='None',color='k',label='%6.0e'%sigmas[i])
ax2.errorbar(x,M[i],yerr=S[i],color='k',ls=ls[i])
if type(ss).__name__=='int' and DEC_freq_sym:
ax1.plot(x[::ss],np.zeros((len(x[::ss]),)),'o',mec='r',mfc='None',ms=8)
ax2.plot(x[::ss],np.zeros((len(x[::ss]),)),'o',mec='r',mfc='None',ms=8)
elif type(ss).__name__=='list' and DEC_freq_sym:
ax1.plot(x[ss],np.zeros((len(x[ss]),)),'o',mec='r',mfc='None',ms=8)
ax2.plot(x[ss],np.zeros((len(x[ss]),)),'o',mec='r',mfc='None',ms=8)
# ax1.set_ylim(0.,); ax1.set_xlim(0.,ax1.get_xlim()[1]*1.05)
deco(iopt=9,ft=15,ax=ax1)
fancy_legend(ax=ax1, size=7,ncol=2,nscat=1)
ax2.set_ylim(0.,); ax2.set_xlim(0.,ax2.get_xlim()[1]*1.05)
deco(iopt=9,ft=15,ax=ax2)
fig.savefig('ee.pdf')
return x, M, S, H, BE, flow_raw