def fve_gun(exp, nom, plt):
from fit import Opt
from gun import magic
from gun import Gun
fig = plt.figure('fve_gun',figsize=fig_y_size(9))
p2f = magic.newton2dyne*magic.area/1e11
opt = Opt(
nom.eos,
{'gun':nom.gun,'stick':nom.stick},
{'gun':exp.vt,'stick':exp.stick_data})
cs,costs = opt.fit(max_iter=5)
print('costs={0}'.format(costs))
opt_gun = Gun(opt.eos)
nom.gun.eos = nom.eos # Restore nominal eos after optimization
t2vs = [Gun(eos).fit_t2v() for eos in [opt.eos.new_c(c) for c in cs]]
e = [exp.v - t2v(exp.t) for t2v in t2vs[1:]]
data = {'nominal':(
(exp.x, nom.eos(exp.x)*p2f, 'f'),
(exp.x, nom.gun.x_dot(exp.x)/magic.cm2km, 'v'))}
data['experimental']=(
(exp.x, exp.eos(exp.x)*p2f, 'f'),
(exp.x, exp.gun.x_dot(exp.x)/magic.cm2km, 'v'))
data['fit']=(
(exp.x, opt.eos(exp.x)*p2f, 'f'),
(exp.x, opt_gun.x_dot(exp.x)/magic.cm2km, 'v'))
cm = r'$x/(\rm{cm})$'
mu_sec = r'$t/(\mu\rm{sec})$'
f_key = r'$f/({\rm dyn}\cdot 10^{11})$'
v_key = r'$v/(\rm{km/s})$'
e_key = r'$\epsilon_k/(\rm{m/s})$'
ax_d = {
'f':{'ax':fig.add_subplot(3,1,1), 'l_x':cm,
'l_y':f_key, 'loc':'upper right'},
'v':{'ax':fig.add_subplot(3,1,2), 'l_x':cm,
'l_y':v_key, 'loc':'lower right'},
'e':{'ax':fig.add_subplot(3,1,3), 'l_x':mu_sec,
'l_y':e_key, 'loc':'upper right'}
}
for mod,xyn in data.items():
for x,y,name in xyn:
if mod == 'experimental':
ax_d[name]['ax'].plot(x,y,'r-.',label=r'$\rm %s$'%mod)
else:
ax_d[name]['ax'].plot(x,y,label=r'$\rm %s$'%mod)
for i in range(len(e)):
ax_d['e']['ax'].plot(exp.t*1e6,e[i]/100,label=r'$\epsilon_%d$'%(i+1))
for name,d in ax_d.items():
d['ax'].legend(loc=ax_d[name]['loc'])
d['ax'].set_xlabel(d['l_x'])
d['ax'].set_ylabel(r'$%s$'%name)
if 'l_y' in d:
d['ax'].set_ylabel(d['l_y'])
return fig
def work():
''' This code for debugging stuff will change often
'''
import matplotlib.pyplot as plt
from eos import Nominal, Spline_eos
from fit import Opt
vt = data()
nom = Spline_eos(Nominal(),precondition=True)
stick = Stick(nom)
opt = Opt(
nom,
{'stick':stick},
{'stick':vt})
cs,costs = opt.fit(max_iter=10)
D, ep, Sigma_inv = stick.compare(vt, cs[-1])
info = np.dot(D.T, np.dot(Sigma_inv, D))
_vals,_vecs = np.linalg.eigh(info)
_vals = np.maximum(_vals, 0)
i = np.argsort(_vals)[-1::-1]
vals = _vals[i]
vecs = _vecs[i]
n_vals = len(np.where(vals > vals[0]*1e-20)[0])
n_vecs = len(np.where(vals > vals[0]*1e-2)[0])
fig = plt.figure()
ax = fig.add_subplot(2,2,1)
costs = np.array(costs)
x = range(len(costs))
c_min = costs.min()
c_max = costs.max()
if c_min > 0:
y = costs
ylabel = r'$\log_{e}(p(y|c_i))$'
else:
offset = (c_max-c_min)*1e-5 -c_min
y = costs + offset
ylabel = r'$\log_{e}(p(y|\hat c_i)) + {0:.3e}$'.format(offset)
ax.semilogy(x,y)
ax.semilogy(x,y, 'kx')
ax.set_ylabel(ylabel)
ax.set_xlabel(r'$i$')
ax.set_xticks(x)
v = np.linspace(.2,50,500)
ax = fig.add_subplot(2,2,2)
ax.loglog(v,nom(v))
ax = fig.add_subplot(2,2,3)
ax.semilogy(range(n_vals), vals[:n_vals])
ax = fig.add_subplot(2,2,4)
for vec in vecs[:n_vecs]:
f = nom.new_c(vec)
ax.semilogx(v,f(v))
plt.show()
vecs = _vecs[i]
# for j,v in enumerate(vals):
# print('v[{0:2d}] = {1:.3e}'.format(j,v))
return 0
def work():
''' This code for debugging stuff will change often
'''
import matplotlib.pyplot as plt
from eos import Nominal, Spline_eos
from fit import Opt
vt = data()
nom = Spline_eos(Nominal(), precondition=True)
stick = Stick(nom)
opt = Opt(nom, {'stick': stick}, {'stick': vt})
cs, costs = opt.fit(max_iter=10)
D, ep, Sigma_inv = stick.compare(vt, cs[-1])
info = np.dot(D.T, np.dot(Sigma_inv, D))
_vals, _vecs = np.linalg.eigh(info)
_vals = np.maximum(_vals, 0)
i = np.argsort(_vals)[-1::-1]
vals = _vals[i]
vecs = _vecs[i]
n_vals = len(np.where(vals > vals[0] * 1e-20)[0])
n_vecs = len(np.where(vals > vals[0] * 1e-2)[0])
fig = plt.figure()
ax = fig.add_subplot(2, 2, 1)
costs = np.array(costs)
x = range(len(costs))
c_min = costs.min()
c_max = costs.max()
if c_min > 0:
y = costs
ylabel = r'$\log_{e}(p(y|c_i))$'
else:
offset = (c_max - c_min) * 1e-5 - c_min
y = costs + offset
ylabel = r'$\log_{e}(p(y|\hat c_i)) + {0:.3e}$'.format(offset)
ax.semilogy(x, y)
ax.semilogy(x, y, 'kx')
ax.set_ylabel(ylabel)
ax.set_xlabel(r'$i$')
ax.set_xticks(x)
v = np.linspace(.2, 50, 500)
ax = fig.add_subplot(2, 2, 2)
ax.loglog(v, nom(v))
ax = fig.add_subplot(2, 2, 3)
ax.semilogy(range(n_vals), vals[:n_vals])
ax = fig.add_subplot(2, 2, 4)
for vec in vecs[:n_vecs]:
f = nom.new_c(vec)
ax.semilogx(v, f(v))
plt.show()
vecs = _vecs[i]
# for j,v in enumerate(vals):
# print('v[{0:2d}] = {1:.3e}'.format(j,v))
return 0
def tx_stick(exp, nom, plt):
import stick
from fit import Opt
fig = plt.figure('tx_stick', figsize=fig_y_size(5.0))
opt = Opt(
nom.eos,
{'stick':nom.stick},
{'stick':exp.stick_data})
cs,costs = opt.fit(max_iter=10)
print('stick costs={0}'.format(costs))
opt_t = stick.data(opt.eos)
x = nom.stick.x / 10 #cm
ax = fig.add_subplot(1,1,1)
ax.plot(x, nom.stick_data*1e6, label='nom')
ax.plot(x, exp.stick_data*1e6, 'r+', label='exp')
ax.plot(x, opt_t*1e6, label='fit')
ax.set_xlabel(r'$x/{\rm cm}$')
ax.set_ylabel(r'$t/\mu{\rm sec}$')
ax.legend(loc='lower right')
fig.subplots_adjust(bottom=0.15) # Make more space for label
return fig
def opt_result(exp, nom, plt):
from fit import Opt
from gun import magic
fig = plt.figure('opt_result', figsize=fig_y_size(6))
p2f = magic.newton2dyne*magic.area/1e11
opt = Opt(
nom.eos,
{'gun':nom.gun,'stick':nom.stick},
{'gun':exp.vt,'stick':exp.stick_data})
eos_0 = opt.eos
opt.fit(max_iter=1)
eos_1 = opt.eos
nom.gun.eos = nom.eos # Restore after optimization
ax = fig.add_subplot(1,1,1)
ax.plot(exp.x,eos_0(exp.x)*p2f,label=r'$f_0$')
ax.plot(exp.x,eos_1(exp.x)*p2f,label=r'$f_1$')
ax.legend()
ax.set_xlabel(r'$x/{\rm cm}$')
ax.set_ylabel(r'$f/({\rm dyn}\cdot 10^{11})$')
return fig
