def run():
# Quantities for the response function
# and randomization
# construct a default response function for a single filament
dict_var = {'fu':1200.0e6,
'qf':1500.,
'L':0.02,
'A':5.30929158457e-10,
'E_mod':70.0e9,
'z':0.0,
'phi':0.,
'f':0.0}
outfile = open( 'pullout_comb_stat.dat', 'w' )
comb = 0
while_cond = 1
while while_cond == 1: #for comb in range( 1, n_com + 1 ):
comb += 1
for i in range( 0, 2 ):
# construct a default response function for a single filament
rf = ConstantFrictionFiniteFiber()
rf.set( **dict_var )
# construct the integrator and provide it with the response function.
s = SPIRRID( rf=rf,
min_eps=0.00, max_eps=0.05, n_eps=80 )
# construct the random variables
n_int = 10
dict_rv = {
'fu':{'variable':'fu', 'distribution':'weibull_min', 'loc':1200.0e6, 'scale':200., 'n_int':n_int },
'qf':{'variable':'qf', 'distribution':'uniform', 'loc':1500., 'scale':100., 'n_int':n_int },
'L':{'variable':'L', 'distribution':'uniform', 'loc':0.02, 'scale':rf.L / 2., 'n_int':n_int},
'A':{'variable':'A', 'distribution':'uniform', 'loc':5.30929158457e-10, 'scale':.03 * 5.30929158457e-10, 'n_int':n_int },
'E_mod':{'variable':'E_mod', 'distribution':'uniform', 'loc':70.e9, 'scale':250.e9, 'n_int':n_int},
'z':{'variable':'z', 'distribution':'uniform', 'loc':0., 'scale':.03, 'n_int':n_int },
'phi':{'variable':'phi', 'distribution':'cos_distr', 'loc':0., 'scale':1., 'n_int':n_int},
'f':{'variable':'f', 'distribution':'uniform', 'loc':0., 'scale':.03, 'n_int':n_int },
}
n_var = len( dict_rv )
n_com = 0
for k in range( 6, 7 ):
n_com += choose( n_var, k )
# end of the while loop
if comb == n_com:
while_cond = 0
list_rv_comb = list( powerset( dict_rv, par=6 ) )
for rv in list_rv_comb[comb]:
s.add_rv( **dict_rv[rv] )
# define a tables with the run configurations to start in a batch
run_list = [
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'bx-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) $ - %4.2f sec'
),
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : False },
'r-2',
'$\mathrm{P}_{e} ( \mathrm{C}_{\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) ) $ - %4.2f sec',
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'go-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot G[\\theta] ) $ - %4.2f sec',
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : False,
'compiled_eps_loop' : False },
'b--',
'$\mathrm{P}_{e} ( \mathrm{N}_{\\theta} ( q(e,\\theta) \cdot G[\\theta] ) ) $ - %4.2f sec'
),
]
print 'Combination', list_rv_comb[comb], ' -- ', i
outfile.write( "Combination %s -- %i\n" % ( list_rv_comb[comb], i ) )
outfile.write( "%s \t %s\n" % ( "Run", "Time" ) )
for idx, run in enumerate( run_list ):
run_options, plot_options, legend_string = run
print 'run', idx,
s.set( **run_options )
plt.figure( comb )
s.mean_curve.plot( plt, plot_options, linewidth=2, label=legend_string % s.exec_time )
print 'execution time', s.exec_time, s.mu_q_peak
outfile.write( "%s \t %s\n" % ( idx, s.exec_time ) )
outfile.write( "=================\n" )
plt.xlabel( 'strain [-]' )
plt.ylabel( 'stress' )
plt.legend( loc='lower right' )
plt.title( s.rf.title )
del s
plt.show()
def run():
# Quantities for the response function
# and randomization
# construct a default response function for a single filament
rf = ConstantFrictionFiniteFiber( fu=1200e15, qf=1200,
L=0.02, A=0.00000002,
E_mod=210.e9, z=0.004,
phi=0.5, f=0.01 )
# construct the integrator and provide it with the response function.
s = SPIRRID( rf=rf,
min_eps=0.00, max_eps=0.0008, n_eps=80 )
# construct the random variables
n_int = 15
s.add_rv( 'E_mod', distribution='uniform', loc=170.e9, scale=250.e9, n_int=n_int )
s.add_rv( 'L', distribution='uniform', loc=0.02, scale=0.03, n_int=n_int )
s.add_rv( 'phi', distribution='cos_distr', loc=0., scale=1., n_int=n_int )
s.add_rv( 'z', distribution='uniform', loc=0, scale=rf.L / 2., n_int=n_int )
# define a tables with the run configurations to start in a batch
run_list = [
(
{'cached_qg' : False,
'compiled_qg_loop' : True,
'compiled_eps_loop' : True },
'bx-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) $ - %4.2f sec'
),
(
{'cached_qg' : False,
'compiled_qg_loop' : True,
'compiled_eps_loop' : False },
'r-2',
'$\mathrm{P}_{e} ( \mathrm{C}_{\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) ) $ - %4.2f sec',
),
(
{'cached_qg' : True,
'compiled_qg_loop' : True,
'compiled_eps_loop' : True },
'go-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot G[\\theta] ) $ - %4.2f sec',
),
(
{'cached_qg' : True,
'compiled_qg_loop' : False,
'compiled_eps_loop' : False },
'b--',
'$\mathrm{P}_{e} ( \mathrm{N}_{\\theta} ( q(e,\\theta) \cdot G[\\theta] ) ) $ - %4.2f sec'
),
]
for idx, run in enumerate( run_list ):
run_options, plot_options, legend_string = run
print 'run', idx,
s.set( **run_options )
s.mean_curve.plot( plt, plot_options, linewidth=2, label=legend_string % s.exec_time )
print 'execution time', s.exec_time, s.mu_q_peak
plt.xlabel( 'strain [-]' )
plt.ylabel( 'stress' )
plt.legend( loc='lower right' )
plt.title( s.rf.title )
plt.show()
def run():
# Quantities for the response function
# and randomization
#
dict_var = {'x' : 10,
'y' : 20,
'z' : 30,
'a' : 40}
n_var = len( dict_var )
n_com = 0
for k in range( 0, n_var ):
n_com += choose( n_var, k )
outfile = open( 'simple_comb_stat.dat', 'w' )
for comb in range( 1, n_com + 1 ):
for i in range( 0, 2 ):
# construct a default response function for a single filament
rf = Filament()
rf.set( **dict_var ) # (x=10, y=20, ...)
# construct the integrator and provide it with the response function.
s = SPIRRID( rf=rf,
min_eps=0.00, max_eps=0.05, n_eps=1 )
# construct the random variables
n_int = 35
dict_rv = {
'x':{'variable':'x', 'distribution':'norm', 'loc':10., 'scale':2., 'n_int':n_int},
'y':{'variable':'y', 'distribution':'norm', 'loc':10., 'scale':2., 'n_int':n_int},
'z':{'variable':'z', 'distribution':'norm', 'loc':10., 'scale':2., 'n_int':n_int},
'a':{'variable':'a', 'distribution':'norm', 'loc':10., 'scale':2., 'n_int':n_int},
}
list_rv_comb = list( powerset( dict_rv ) )
for rv in list_rv_comb[comb]:
s.add_rv( **dict_rv[rv] )
# define a tables with the run configurations to start in a batch
run_list = [
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'bx-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) $ - %4.2f sec'
),
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : False },
'r-2',
'$\mathrm{P}_{e} ( \mathrm{C}_{\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) ) $ - %4.2f sec',
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'go-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot G[\\theta] ) $ - %4.2f sec',
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : False,
'compiled_eps_loop' : False },
'y--',
'$\mathrm{P}_{e} ( \mathrm{N}_{\\theta} ( q(e,\\theta) \cdot G[\\theta] ) ) $ - %4.2f sec'
),
]
legend = []
print 'Combination', list_rv_comb[comb], ' -- ', i
outfile.write( "Combination %s -- %i\n" % ( list_rv_comb[comb], i ) )
outfile.write( "%s \t %s\n" % ( "Run", "Time" ) )
for idx, run in enumerate( run_list ):
run_options, plot_options, legend_string = run
print 'run', idx,
s.set( **run_options )
s.mean_curve.plot( plt, plot_options )
print 'execution time', s.exec_time, s.mu_q_peak
outfile.write( "%s \t %s\n" % ( idx, s.exec_time ) )
legend.append( legend_string % s.exec_time )
outfile.write( "=================\n" )
plt.xlabel( 'strain [-]' )
plt.ylabel( 'stress' )
plt.legend( legend )
plt.title( s.rf.title )
del s
plt.show()
outfile.close()
erf(0.5 * math.sqrt(2) * (e - m_xi) / std_xi)) * m_la
#===========================================================================
# Lab
#===========================================================================
slab = SPIRRIDLAB(s = s, save_output = True, show_output = True,
exact_arr = mu_q_ex(e_arr, m_xi, std_xi, m_la))
#===========================================================================
# Compare efficiency of sampling types
#===========================================================================
powers = np.linspace(1, math.log(1000, 10), 80)
n_int_range = np.array(np.power(10, powers), dtype = int)
# slab.sampling_efficiency(n_int_range = n_int_range)
#===========================================================================
# Compare the structure of sampling
#===========================================================================
# slab.sampling_structure(ylim = 18.0, xlim = 1.2,)
#===========================================================================
# Compare the code efficiency
#==============================8=============================================
s.set(e_arr = np.linspace(0, 2.0, 40), n_int = 400)
#slab.codegen_language_efficiency(extra_compiler_args = False)
slab.codegen_language_efficiency(extra_compiler_args = True)
def run():
# Quantities for the response function
# and randomization
#
dict_var = {"x": 10, "y": 20, "z": 30, "a": 40}
n_var = len(dict_var)
n_com = 0
for k in range(0, n_var):
n_com += choose(n_var, k)
outfile = open("simple_comb_loop_stat.dat", "w")
for comb in range(1, n_com + 1):
for i in range(0, 2):
# construct a default response function for a single filament
rf = Filament()
rf.set(**dict_var) # (x=10, y=20, ...)
# construct the integrator and provide it with the response function.
s = SPIRRID(rf=rf, min_eps=0.00, max_eps=0.05, n_eps=1)
# construct the random variables
n_int = 35
dict_rv = {
"x": {"variable": "x", "distribution": "norm", "loc": 10.0, "scale": 2.0, "n_int": n_int},
"y": {"variable": "y", "distribution": "norm", "loc": 10.0, "scale": 2.0, "n_int": n_int},
"z": {"variable": "z", "distribution": "norm", "loc": 10.0, "scale": 2.0, "n_int": n_int},
"a": {"variable": "a", "distribution": "norm", "loc": 10.0, "scale": 2.0, "n_int": n_int},
}
list_rv_comb = list(powerset(dict_rv))
for rv in list_rv_comb[comb]:
s.add_rv(**dict_rv[rv])
# define a tables with the run configurations to start in a batch
run_list = [
(
{"cached_qg": False, "compiled_qg_loop": True, "compiled_eps_loop": True},
"bx-",
"$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) $ - %4.2f sec",
),
(
{"cached_qg": False, "compiled_qg_loop": True, "compiled_eps_loop": False},
"r-2",
"$\mathrm{P}_{e} ( \mathrm{C}_{\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) ) $ - %4.2f sec",
),
(
{"cached_qg": True, "compiled_qg_loop": True, "compiled_eps_loop": True},
"go-",
"$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot G[\\theta] ) $ - %4.2f sec",
),
(
{"cached_qg": True, "compiled_qg_loop": False, "compiled_eps_loop": False},
"y--",
"$\mathrm{P}_{e} ( \mathrm{N}_{\\theta} ( q(e,\\theta) \cdot G[\\theta] ) ) $ - %4.2f sec",
),
]
legend = []
print "Combination", list_rv_comb[comb], " -- ", i
outfile.write("Combination %s -- %i\n" % (list_rv_comb[comb], i))
outfile.write("%s \t %s\n" % ("Run", "Time"))
for idx, run in enumerate(run_list):
run_options, plot_options, legend_string = run
print "run", idx,
s.set(**run_options)
s.mean_curve.plot(plt, plot_options)
print "execution time", s.exec_time, s.mu_q_peak
outfile.write("%s \t %s\n" % (idx, s.exec_time))
legend.append(legend_string % s.exec_time)
outfile.write("=================\n")
plt.xlabel("strain [-]")
plt.ylabel("stress")
plt.legend(legend)
plt.title(s.rf.title)
del s
plt.show()
outfile.close()
def run():
# Quantities for the response function
# and randomization
#
E_mod = 70 * 1e+9 # Pa
sig_u = 1.25 * 1e+9 # Pa
D = 26 * 1.0e-6 # m
A = ( D / 2.0 ) ** 2 * pi
xi_u = sig_u / E_mod
dict_var = {'E_mod':70.e9,
'xi':0.02,
'A':A,
'theta':0,
'lambd':0}
outfile = open( 'filament_comb_stat.dat', 'w' )
comb = 0
while_cond = 1
while while_cond == 1: #for comb in range( 1, n_com + 1 ):
comb += 1
for i in range( 0, 2 ):
# construct a default response function for a single filament
rf = Filament()
rf.set( **dict_var )
# construct the integrator and provide it with the response function.
s = SPIRRID( rf=rf,
min_eps=0.00, max_eps=0.05, n_eps=80 )
# construct the random variables
n_int = 32
dict_rv = {
'xi':{'variable':'xi', 'distribution':'weibull_min', 'scale':0.02, 'shape':10., 'n_int':n_int},
'E_mod':{'variable':'E_mod', 'distribution':'uniform', 'loc':70e+9, 'scale':15e+9, 'n_int':n_int},
'theta':{'variable':'theta', 'distribution':'uniform', 'loc':0.0, 'scale':0.01, 'n_int':n_int},
'lambd':{'variable':'lambd', 'distribution':'uniform', 'loc':0.0, 'scale':.2, 'n_int':n_int},
'A':{'variable':'A', 'distribution':'uniform', 'loc':A * 0.3, 'scale':0.7 * A, 'n_int':n_int},
}
n_var = len( dict_rv )
n_com = 0
for k in range( 0, n_var ):
n_com += choose( n_var, k )
print 'The only last combination'
comb = n_com
# end of the while loop
if comb == n_com:
while_cond = 0
list_rv_comb = list( powerset( dict_rv ) )
for rv in list_rv_comb[comb]:
s.add_rv( **dict_rv[rv] )
# define a tables with the run configurations to start in a batch
run_list = [
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'bx-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) $ - %4.2f sec'
),
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : False },
'r-2',
'$\mathrm{P}_{e} ( \mathrm{C}_{\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) ) $ - %4.2f sec',
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : False,
'compiled_eps_loop' : False },
'y--',
'$\mathrm{P}_{e} ( \mathrm{N}_{\\theta} ( q(e,\\theta) \cdot G[\\theta] ) ) $ - %4.2f sec'
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'go-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot G[\\theta] ) $ - %4.2f sec',
),
]
legend = []
print 'Combination', list_rv_comb[comb], ' -- ', i
outfile.write( "Combination %s -- %i\n" % ( list_rv_comb[comb], i ) )
outfile.write( "%s \t %s\n" % ( "Run", "Time" ) )
for idx, run in enumerate( run_list ):
run_options, plot_options, legend_string = run
print 'run', idx,
s.set( **run_options )
plt.figure( comb )
s.mean_curve.plot( plt, plot_options )
print 'execution time', s.exec_time
outfile.write( "%s \t %s\n" % ( idx, s.exec_time ) )
legend.append( legend_string % s.exec_time )
outfile.write( "=================\n" )
plt.xlabel( 'strain [-]' )
plt.ylabel( 'stress' )
plt.legend( legend )
plt.title( s.rf.title )
del s
plt.show()
def run():
# Quantities for the response function
# and randomization
#
xx = 10 # Pa
yy = 20 # Pa
zz = 30
aa = 40
# construct a default response function for a single filament
rf = Filament( x=xx, y=yy, z=zz, a=aa )
# construct the integrator and provide it with the response function.
s = SPIRRID( rf=rf,
min_eps=0.00, max_eps=0.05, n_eps=1 )
# construct the random variables
n_int = 80
s.add_rv( 'x', distribution='norm', scale=10., shape=2., n_int=n_int )
s.add_rv( 'y', distribution='norm', loc=10., scale=2., n_int=n_int )
s.add_rv( 'z', distribution='norm', loc=10., scale=2., n_int=n_int )
s.add_rv( 'a', distribution='norm', loc=10., scale=2., n_int=n_int )
# define a tables with the run configurations to start in a batch
run_list = [
(
{'cached_qg' : False,
'compiled_qg_loop' : True,
'compiled_eps_loop' : True },
'bx-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) $ - %4.2f sec'
),
(
{'cached_qg' : False,
'compiled_qg_loop' : True,
'compiled_eps_loop' : False },
'r-2',
'$\mathrm{P}_{e} ( \mathrm{C}_{\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) ) $ - %4.2f sec',
),
(
{'cached_qg' : True,
'compiled_qg_loop' : True,
'compiled_eps_loop' : True },
'go-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot G[\\theta] ) $ - %4.2f sec',
),
(
{'cached_qg' : True,
'compiled_qg_loop' : False,
'compiled_eps_loop' : False },
'y--',
'$\mathrm{P}_{e} ( \mathrm{N}_{\\theta} ( q(e,\\theta) \cdot G[\\theta] ) ) $ - %4.2f sec'
),
]
legend = []
for idx, run in enumerate( run_list ):
run_options, plot_options, legend_string = run
print 'run', idx,
s.set( **run_options )
s.mean_curve.plot( plt, plot_options )
print 'execution time', s.exec_time
legend.append( legend_string % s.exec_time )
plt.xlabel( 'strain [-]' )
plt.ylabel( 'stress' )
plt.legend( legend )
plt.title( s.rf.title )
plt.show()
exact_arr = mu_q_ex(e_arr))
#===========================================================================
# Compare efficiency of sampling types
#===========================================================================
powers = np.linspace(1, math.log(20, 10), 15)
n_int_range = np.array(np.power(10, powers), dtype = int)
#slab.sampling_efficiency(n_int_range = n_int_range)
#===========================================================================
# Compare the structure of sampling
#===========================================================================
#slab.sampling_structure( ylim = 1.1, xlim = 0.04 )
#===========================================================================
# Compare the code efficiency
#===========================================================================
s.set(e_arr = np.linspace(0, 0.04, 20), n_int = 40)
s.sampling_type = 'PGrid'
slab.codegen_efficiency()
#===========================================================================
# Compare the efficiency wrt. order and number of RV (combinations)
#===========================================================================
#slab.combination_efficiency(tvars_det = dict(lambd = la_mean, xi = xi_mean, E_mod = E_mean, theta = th_mean, A = A_mean),
# tvars_rand = dict(lambd = g_la, xi = g_xi, E_mod = g_E, theta = g_th, A = g_A))
def run():
# Quantities for the response function
# and randomization
#
E_mod = 70 * 1e+9 # Pa
sig_u = 1.25 * 1e+9 # Pa
D = 26 * 1.0e-6 # m
A = ( D / 2.0 ) ** 2 * pi
xi_u = sig_u / E_mod
# construct a default response function for a single filament
rf = Filament( E_mod=70.e9, xi=0.02, A=A, theta=0, lambd=0 )
# construct the integrator and provide it with the response function.
s = SPIRRID( rf=rf,
min_eps=0.00, max_eps=0.05, n_eps=80 )
# construct the random variables
n_int = 70
s.add_rv( 'xi', distribution='weibull_min', scale=0.02, shape=10., n_int=n_int )
# s.add_rv( 'E_mod', distribution = 'uniform', loc = 70e+9, scale = 15e+9, n_int = n_int )
s.add_rv( 'theta', distribution='uniform', loc=0.0, scale=0.01, n_int=n_int )
s.add_rv( 'lambd', distribution='uniform', loc=0.0, scale=.2, n_int=n_int )
s.add_rv( 'A', distribution='uniform', loc=A * 0.3, scale=0.7 * A, n_int=n_int )
# define a tables with the run configurations to start in a batch
run_list = [
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'bx-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) $ - %4.2f sec'
),
(
{'cached_dG' : False,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : False },
'r-2',
'$\mathrm{P}_{e} ( \mathrm{C}_{\\theta} ( q(e,\\theta) \cdot g[\\theta_1] g[\\theta_2] \dots g[\\theta_n] ) ) $ - %4.2f sec',
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : True,
'compiled_eps_loop' : True },
'go-',
'$\mathrm{C}_{e,\\theta} ( q(e,\\theta) \cdot G[\\theta] ) $ - %4.2f sec',
),
(
{'cached_dG' : True,
'compiled_QdG_loop' : False,
'compiled_eps_loop' : False },
'y--',
'$\mathrm{P}_{e} ( \mathrm{N}_{\\theta} ( q(e,\\theta) \cdot G[\\theta] ) ) $ - %4.2f sec'
),
]
legend = []
for idx, run in enumerate( run_list ):
run_options, plot_options, legend_string = run
print 'run', idx,
s.set( **run_options )
s.mean_curve.plot( plt, plot_options )
print 'execution time', s.exec_time
legend.append( legend_string % s.exec_time )
plt.xlabel( 'strain [-]' )
plt.ylabel( 'stress' )
plt.legend( legend )
plt.title( s.rf.title )
plt.show()