def triangle_contour(x_center, y_center, values, smoothing, ckwargs={}):
"""A helper funciton used to draw contorus with pl.tricontour"""
# make Triangulation object using the centers of each of the hexbins
triag = Triangulation(x_center, y_center)
refiner = UniformTriRefiner(triag) # refines the mesh of triangle
# returns refines triangle field of triangles and interpolated
# contour values by dividing each triangle into 4**subdiv triangles
tri_refi, c_refi = refiner.refine_field(values, subdiv=smoothing)
T = pl.tricontour(tri_refi, c_refi, **ckwargs)
return T
def extract_contour(x_list, y_list, z_list, z_value):
import matplotlib
matplotlib.use('Agg')
import pylab
import numpy
temp = pylab.tricontour(x_list, y_list, z_list, levels=[z_value])
pylab.clf()
return numpy.concatenate([itm.vertices for itm
in temp.collections[0].get_paths()])
def plot_mesh(pts, tri, z=None):
if z is not None:
x = pts[:,0]
y = pts[:,1]
tripcolor(x, y, tri, z)
colorbar()
cs = tricontour(x, y, tri, z, colors="black", linestyles="dashed")
clabel(cs)
else:
if len(tri[0]) == 3:
plot_tri_mesh(pts, tri)
else:
plot_quad_mesh(pts, tri)
axis('tight')
xticks([])
yticks([])
axes().set_aspect('equal')
def plot_mesh(pts, tri, z=None):
if z is not None:
x = pts[:, 0]
y = pts[:, 1]
tripcolor(x, y, tri, z)
colorbar()
cs = tricontour(x, y, tri, z, colors="black", linestyles="dashed")
clabel(cs)
else:
if len(tri[0]) == 3:
plot_tri_mesh(pts, tri)
else:
plot_quad_mesh(pts, tri)
axis('tight')
xticks([])
yticks([])
axes().set_aspect('equal')
import numpy as np
import matplotlib as mpl
mpl.use("Agg")
import pylab as plt
plt.figure()
data = np.load("lossgrad_lensing2.npz")
c = data['loss']-data['loss'].min()
#c = np.minimum(25, c)
plt.tricontour(data['entropy/a'], data['entropy/b'], c, levels = [1, 4, 9, 25, 36, 49, 64, 81, 100, 200, 400])
plt.xlabel("a")
plt.ylabel("b")
plt.scatter(data['entropy/a'], data['entropy/b'], c=c)
plt.colorbar()
plt.title(data['loss'].min())
plt.savefig("loss.pdf")
#plt.figure(figsize = (20, 10))
#data = np.load("mock_lensing.npz")
#data2 = np.load("ppd_lensing.npz")
##mu = data['entropy/mu']
#mu2 = data2['entropy/mu']
#obs = data['entropy/obs']
##plt.plot(mu)
#plt.plot(mu2)
#plt.plot(obs, marker='.', ls='')
#plt.savefig("mu.pdf")
def main():
"""
D(R) = <(phi(r) - phi(r+R))^2>
= <K^2(TEC(r) - TEC(r + R))^2>
= K^2<DTEC(r, -R)^2>
:return:
"""
fig, ax = plt.subplots(1, 1, figsize=(6, 6))
baselines = 10**np.linspace(-1., 2, 100)
k_obs = []
for _ in range(500):
ds = np.random.uniform(5., 30.)
beta = np.random.normal(1.89, 0.1)
k_obs.append((baselines / ds)**(beta))
ax.plot(baselines,
np.mean(k_obs, axis=0),
lw=2.,
color='black',
label='Mevius+ 2016')
ax.fill_between(baselines,
np.percentile(k_obs, 5, axis=0),
np.percentile(k_obs, 95, axis=0),
color='yellow',
alpha=0.5)
k_turb = (baselines / 10.)**(5. / 3.)
ax.plot(baselines, k_turb, color='green', lw=2., label=r'Kolmogorov $5/3$')
with tf.Session() as sess:
x = tf.placeholder(float_type, shape=(3, ))
k = tf.placeholder(float_type, shape=(3, ))
khat = k / tf.linalg.norm(k)
X = tf.concat([khat[None, :], x[None, :]], axis=1) #1,6
theta = tf.constant([6., 14., 1. / 3.], float_type)
int_kern = M32(theta) # + RBF(theta/200.)
kern = TrapezoidKernel(int_kern,
20,
tf.constant(250., float_type),
tf.constant(100., float_type),
obs_type='DTEC')
K = kern.K(X, X)
xy = []
z = []
for b in baselines:
xy.append([np.concatenate([[b, 0., 0.]], axis=0), [0., 0., 1.]])
# np.concatenate([0.0 * np.random.normal(size=2), [1.]],axis=0)
xy[-1][1] /= np.linalg.norm(xy[-1][1])
np_K = sess.run(K, {x: xy[-1][0], k: xy[-1][1]})
z.append(np.sqrt(np_K[0, 0]))
xy = np.array(xy)
z = 8.448e6 * np.array(z) / 150e6
ax.plot(xy[:, 0, 0],
z**2,
ls='dotted',
lw=2.,
color='blue',
label='dawn')
with tf.Session() as sess:
x = tf.placeholder(float_type, shape=(3, ))
k = tf.placeholder(float_type, shape=(3, ))
khat = k / tf.linalg.norm(k)
X = tf.concat([khat[None, :], x[None, :]], axis=1) # 1,6
theta = tf.constant([3., 17., 1. / 3.], float_type)
int_kern = RBF(theta) # + RBF(theta/200.)
kern = TrapezoidKernel(int_kern,
20,
tf.constant(350., float_type),
tf.constant(200., float_type),
obs_type='DTEC')
K = kern.K(X, X)
xy = []
z = []
for b in baselines:
xy.append(
[np.concatenate([[b, 0., 0.]], axis=0), [0., 0., 1.]])
# np.concatenate([0.0 * np.random.normal(size=2), [1.]],axis=0)
xy[-1][1] /= np.linalg.norm(xy[-1][1])
np_K = sess.run(K, {x: xy[-1][0], k: xy[-1][1]})
z.append(np.sqrt(np_K[0, 0]))
xy = np.array(xy)
z = 8.448e6 * np.array(z) / 150e6
ax.plot(xy[:, 0, 0],
z**2,
ls='dashed',
lw=2.,
color='pink',
label='dusk')
print(z)
ax.grid()
ax.set_yscale('log')
ax.set_xscale('log')
ax.set_xlim(0.1, 100.)
ax.set_ylim(1e-4, 1e2)
ax.set_ylabel(r"$\mathrm{Var}[\phi_{\rm 150}]$ [rad$^2$]")
ax.set_xlabel("Baseline [km]")
# plt.tricontour(xy[:,0,0],xy[:,1,0], z,levels=10)
# plt.title("Over x")
ax.legend()
plt.savefig("/home/albert/Documents/structure_function.pdf")
plt.show()
plt.subplot(projection='3d')
plt.scatter(xy[:, 1, 0], xy[:, 1, 1], c=z, alpha=0.5, marker='+')
plt.tricontour(xy[:, 1, 0], xy[:, 1, 1], z, levels=10)
plt.title("Over k")
plt.show()
diff_scale = []
thetas = np.linspace(0., np.pi / 6., 100)
with tf.Session() as sess:
x = tf.placeholder(float_type, shape=(3, ))
k = tf.placeholder(float_type, shape=(3, ))
khat = k / tf.linalg.norm(k)
X = tf.concat([khat[None, :], x[None, :]], axis=1) #1,6
theta = tf.constant([10., 14., 1. / 3.], float_type)
int_kern = M32(theta) # + RBF(theta/200.)
kern = TrapezoidKernel(int_kern,
20,
tf.constant(250., float_type),
tf.constant(100., float_type),
obs_type='DTEC')
K = kern.K(X, X)
for theta in thetas:
xy = []
z = []
for b in baselines:
xy.append([
np.concatenate([[b, 0., 0.]], axis=0),
[0., np.sin(theta), np.cos(theta)]
])
# np.concatenate([0.0 * np.random.normal(size=2), [1.]],axis=0)
xy[-1][1] /= np.linalg.norm(xy[-1][1])
np_K = sess.run(K, {x: xy[-1][0], k: xy[-1][1]})
z.append(np.sqrt(np_K[0, 0]))
xy = np.array(xy)
z = 8.448e6 * np.array(z) / 150e6
ds = np.interp(1., z, xy[:, 0, 0])
diff_scale.append(ds)
fig, ax = plt.subplots(1, 1, figsize=(6, 4))
ax.plot(np.array(thetas) * 180 / np.pi,
diff_scale,
lw=2.,
color='blue')
ax.grid()
ax.set_ylabel(r'$r_{\rm scale}$ [km]')
ax.set_xlabel('Zenith angle [degrees]')
plt.savefig('/home/albert/Documents/diffractive_scale.pdf')
plt.show()
n_steps=n_steps,
lr=1e-3,
n_particles=n_particles)
losses += swyft.train(network_2d,
xz,
n_steps=n_steps,
lr=1e-4,
n_particles=n_particles)
# Plot results 1-dim
z_lnL = swyft.estimate_lnL(network, x0, swyft.get_z(xz))
plt.clf()
plt.plot(z_lnL[0]['z'], np.exp(z_lnL[0]['lnL']))
plt.axvline(0.1)
plt.axvline(0.9)
plt.savefig("figs/testrun_09a.png")
# plot 2-dim results
z_lnL = swyft.estimate_lnL_2d(network_2d, x0, swyft.get_z(xz))
lnL = z_lnL['lnL']
z_ij = z_lnL['z']
plt.clf()
plt.tricontour([zzz[0] for zzz in z_ij], [zzz[1] for zzz in z_ij],
lnL * 2,
levels=[-36, -25, -16, -9, -4, -1, 0])
t = np.linspace(0, 6.5, 1000)
xc = np.sin(t) * 0.4 + 0.5
yc = np.cos(t) * 0.4 + 0.5
plt.scatter(xc, yc)
plt.savefig('figs/testrun_09b.png')
MN = [ 0.1001, 0.20, 0.30, 0.40, 0.50, 0.60, 0.70, 0.80, 0.8999 ]
pylab.figure( figsize=(12,12), facecolor='lightgrey' )
axes = [ 0.00, 1.2001, 0.0, 1.0 ]
pylab.axis( axes )
pylab.yticks([0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0])
pylab.tick_params(axis='both', which='major', labelsize=16)
pylab.ylabel('Area Ratio (Bypass/Tube)', fontsize=18)
pylab.xlabel('Bypass Mach Number', fontsize=18)
pylab.grid(b=True, which='major', color='grey', linestyle='--')
#pylab.title( 'hyperloop' )
pylab.legend(loc="best")
CS = pylab.tricontour( MNbyp, AR, MNpod, MN, colors = ['darkblue','red','darkgreen','purple', 'grey','darkorange', 'black', 'lightblue'] )
fmt = {} #tricontour labels
strs = [ 'Pod M=0.1', '0.2', '0.3', '0.4', '0.5', '0.6', '0.7', '0.8', '0.9' ]
for l,s in zip( CS.levels, strs ):
fmt[l] = s
xx = [0.928, 0.97, 0.97, 0.97, 0.97, 0.97, 0.97, 0.97, 0.97]
yy = [0.18, 0.34, 0.49, 0.62, 0.75, 0.83, 0.92, 0.96, 0.985]
labelpts = zip(xx,yy)
pylab.clabel( CS, inline=1,fontsize=14, fmt = fmt , manual=labelpts )
pylab.annotate(" Available pod area \ndecreasing relative to tube", fontsize=16, xy=(1.09, 0.486), xycoords='data', xytext=(0., 0),
rotation=90, textcoords='offset points', bbox=dict(boxstyle="square", edgecolor='lightgrey',facecolor='lightgrey') )
def main(num_train, num_test, test_fraction, time_limit, output_folder,
num_processes):
if test_fraction is None:
if num_test is None:
raise ValueError(
"num_test must be int > 0 if test_fraction is None.")
test_fraction = num_test / (num_train + num_test)
total_desired = num_train / (1. - test_fraction)
logging.info("Total desired examples: {}".format(total_desired))
logging.info("Test fraction: {}".format(test_fraction))
t0 = default_timer()
if time_limit is None:
time_limit = np.inf
logging.info("Simulating with timelimit {} seconds".format(time_limit))
output_folder = os.path.abspath(output_folder)
train_folder = os.path.join(output_folder, 'train_examples')
test_folder = os.path.join(output_folder, 'test_examples')
os.makedirs(train_folder, exist_ok=True)
os.makedirs(test_folder, exist_ok=True)
logging.info("Using output folder {}".format(output_folder))
num_done = 0
while default_timer() - t0 < time_limit and num_done < total_desired:
logging.info("Working on example: {:04d}".format(num_done))
# Create a cluster
star_cluster = create_cluster_with_IMF(N=200)
star_cluster.position += [-6.5, 0, 0] | units.kpc
star_cluster.velocity += [0, 50, 0] | units.kms
# Construct the galactic model and make the evolution
evolution = RealisticEvolutionCluster(simulation_time=150 | units.Myr)
res_dict = evolution.evolution_of_the_cluster(star_cluster)
gaia_dict = stars_to_gaia_observables(**res_dict)
save_location = os.path.join(train_folder,
"example_{:05d}".format(num_done))
# plots
plot_location = os.path.join(save_location, 'data_plots')
os.makedirs(plot_location, exist_ok=True)
for i in range(len(res_dict.keys())):
key1 = res_dict.keys()[i]
for j in range(i + 1, len(res_dict.keys())):
key2 = res_dict.keys()[j]
plt.figure(figsize=(6, 6))
plt.scatter(res_dict[key1],
res_dict[key2],
c='black',
marker='+',
alpha=0.5)
points = np.stack([res_dict[key1], res_dict[key2]], axis=0)
kernel = st.gaussian_kde(points)
z = kernel(points)
plt.tricontour(res_dict[key1],
res_dict[key2],
z,
levels=7,
linewidths=0.5,
colors='k')
plt.xlabel(key1)
plt.ylabel(key2)
plt.grid()
plt.tight_layout()
plt.savefig(
os.path.join(plot_location, '{}_{}.png'.format(key1,
key2)))
plt.close('all')
with h5py.File(os.path.join(save_location, "starcluster_data.h5"),
'w') as f:
for key, value in res_dict.items():
f[key] = np.array(value)
with h5py.File(os.path.join(save_location, "gaia_data.h5"), 'w') as f:
for key, value in gaia_dict.items():
f[key] = np.array(value)
import pylab
class MyDesign(LHSDesign):
spec = (('x', (0, 6.28, 'uniform', 10)), ('y', (0, 6.28, 'uniform', 10)))
samples = 50
class MyModel:
# model can be any class, but it is required to have 'solve' method
# which takes sample `d` as argument and returns dict of results `res`
def solve(self, d):
x, y = d.x, d.y
res = {'F': sin(x) * cos(y), 'G': x * y}
return res
# Evaluate experiments and dump everything to 'test' directory:
ex = Experiment(MyDesign(), MyModel(), dirname='test')
ex.run()
# Plot 'F'
x = ex.design.x
y = ex.design.y
F = ex.result.F
pylab.tricontour(x, y, F, levels=14, linewidths=0.5, colors='k')
cntr = pylab.tricontourf(x, y, F, levels=14, cmap="RdBu_r")
pylab.colorbar(cntr)
pylab.plot(x, y, 'ko', ms=3)
pylab.title('$\sin(x)\cos(y)$\n (%d LHS samples)' % ex.design.samples)
#slope = np.array([-0.02,-0.01,0.0,0.01,0.02,0.03,0.04])
#flux = np.array([0,2,4,6])
plt.close('all')
width = 3.487 * 2
height = width / 1.618 / 2
fig = plt.figure(num=1, figsize=(width, height), facecolor='w', edgecolor='k')
fig.subplots_adjust(left=0.073, right=1.05, top=.725, bottom=0.2)
#fig.set_size_inches(width, height)
#plt.draw()
#fig, ax = plt.subplots(constrained_layout=True)
dLdt = np.array(dLdt)
F = np.array(F)
plt.tricontour(S, F / 1e3, dLdt / 1e3, [0], colors='k', linestyles='--')
c = plt.tricontourf(S,
F / 1e3,
dLdt / 1e3,
v,
cmap='bwr_r',
norm=norm,
extend='both')
#plt.scatter([-0.03],[4],s=50,c=[351.5094769326309 /1e3],cmap='bwr_r',norm=norm,marker='o',edgecolor='k')
#plt.scatter([-0.01],[2],s=50,c=[-2137.5210508252835/1e3],cmap='bwr_r',norm=norm,marker='o',edgecolor='k')
#plt.scatter([0.0],[5],s=50,c=[487.56420334626046/1e3],cmap='bwr_r',norm=norm,marker='o',edgecolor='k')
cbar = plt.colorbar(extend='both', ticks=[vmin, 0, vmax])
#plt.clim([vmin,vmax])
cbar.ax.set_yticklabels([vmin, 0, vmax])
cbar.set_label(r'$dL/dt$ (km/a)')
