def __init__(self, net, logger=None):
super(pol_net, self).__init__()
if logger is None:
self.logger = get_logger('.log.txt')
else:
self.logger = logger
self.net = net
self.gamma = 1.0
self.lr = 0.001
### HOW LONG ARE WE TRAINING FOR !!!!! CHANGE HERE DON'T HARDCODE ###
self.num_trainings = 10
self.num_epochs = 15
self.num_batches = 10000
self.initialize_lstm()
self.linear = nn.Linear(self.lstm_hidden_size, 1, bias=False)
with torch.no_grad():
self.linear.weight.data.fill_(-0.1)
self.log_std = nn.Parameter(torch.empty(1).fill_(-2))
self.saved_log_probs = []
self.optimizer = torch.optim.Adam(self.parameters(), lr=self.lr)
self.color = cm.cool(
np.linspace(
0, 1,
self.num_batches * self.num_trainings // self.net.plot_freq +
2))
plt.plot(data_N200mln["B"],
data_N200mln["exactPrice"],
color="orchid",
label="Exact price",
marker="x",
linewidth=1)
plt.xlabel("$B$")
plt.ylabel("Estimated Monte Carlo exact price [EUR]")
plt.grid()
plt.savefig(outputDirectory + "OptionPriceVsB_PriceVsB_N200mln.pdf",
bbox_inches='tight')
plt.close()
## Plot exact errors vs. N in log-log scale
colorlist = iter(cm.cool(np.linspace(0, 1, len(BValues))))
for selectedB in BValues:
mask = (data["B"] == selectedB)
maskeddata = data[mask]
maskeddata.sort("N")
currentcolor = next(colorlist)
if ((selectedB > 0.25) & (selectedB < 5)):
plt.plot(maskeddata["N"],
maskeddata["exactError"],
color=currentcolor,
marker="",
label="B = " + str(selectedB),
linewidth=1)
plt.xscale("log")
def hro(Imap, Qmap, Umap, steps=10, hsize=15, minI=0., outh=[0, 4, 9]):
# Calculates the relative orientation angle between the density structures and the magnetic field.
# INPUTS
# Imap - Intensity or column density map
# Qmap - Stokes Q map
# Upam - Stokes U map
sz = np.shape(Imap)
phi = roangles(Imap, Qmap, Umap)
hist, bin_edges = np.histogram(Imap[(Imap > minI).nonzero()],
bins=100 * sz[0])
bin_centre = 0.5 * (bin_edges[0:np.size(bin_edges) - 1] +
bin_edges[1:np.size(bin_edges)])
chist = np.cumsum(hist)
pitch = np.max(chist) / float(steps)
hsteps = pitch * np.arange(0, steps + 1, 1)
Isteps = np.zeros(steps + 1)
for i in range(0, np.size(Isteps) - 1):
good = np.logical_and(chist > hsteps[i],
chist < hsteps[i + 1]).nonzero()
Isteps[i] = np.min(bin_centre[good])
Isteps[np.size(Isteps) - 1] = np.max(Imap)
hros = np.zeros([steps, hsize])
Smap = 0. * Imap
xi = np.zeros(steps)
prs = np.zeros(steps)
s_prs = np.zeros(steps)
cdens = np.zeros(steps)
for i in range(0, np.size(Isteps) - 1):
good = np.logical_and(Imap > Isteps[i], Imap < Isteps[i + 1]).nonzero()
hist, bin_edges = np.histogram((180 / np.pi) * phi[good],
bins=hsize,
range=(-90., 90.))
bin_centre = 0.5 * (bin_edges[0:np.size(bin_edges) - 1] +
bin_edges[1:np.size(bin_edges)])
hros[i, :] = hist
cdens[i] = np.mean([Isteps[i], Isteps[i + 1]])
Smap[good] = i
xi[i] = roparameter(bin_centre, hist)
TEMPprs, TEMPs_prs = projRS(phi[good])
prs[i] = TEMPprs
s_prs[i] = TEMPs_prs
print(i)
outsteps = np.size(outh)
color = iter(cm.cool(np.linspace(0, 1, outsteps)))
fig = plt.figure()
for i in range(0, outsteps):
c = next(color)
labeltext = "%.2f" % Isteps[outh[
i]] + r' < $N_{\rm H}/$cm$^{-2}$ < ' + "%.2f" % Isteps[outh[i] + 1]
plt.plot(bin_centre,
hros[outh[i], :],
'-',
linewidth=2,
c=c,
label=labeltext) #drawstyle
plt.xlabel(r'cos($\phi$)')
plt.legend()
plt.show()
# --------------------------------------------------------------
fig = plt.figure()
plt.plot(cdens, xi, '-', linewidth=2, c='red')
plt.axhline(y=0., c='k', ls='--')
plt.xlabel(r'log$_{10}$ ($N_{\rm H}/$cm$^{-2}$)')
plt.ylabel(r'$\zeta$')
plt.show()
# --------------------------------------------------------------
fig = plt.figure()
plt.plot(cdens, prs, '-', linewidth=2, c='blue')
plt.errorbar(cdens, prs, yerr=s_prs, c='blue', fmt='o')
plt.axhline(y=0., c='k', ls='--')
plt.xlabel(r'log$_{10}$ ($N_{\rm H}/$cm$^{-2}$)')
plt.ylabel(r'$Z_{x}$')
plt.show()
return Isteps, xi
def hro3D(dens, Bx, By, Bz, steps=10, hsize=21, mind=0, outh=[0,4,9]):
"""
Calculate the histogram of relative orientations (HRO) in three-dimensional data.
Parameters
----------
dens : ...
Bx,y,z : ...
steps : ...
hsize : ...
mind : ...
outh : ...
Returns
-------
hro : ...
cdens : ...
zeta : ...
"""
cosphi=roangles3D(dens, Bx, By, Bz)
dsteps=equibins(dens, steps=steps, mind=mind)
hros = np.zeros([steps,hsize])
cdens = np.zeros(steps)
xi = np.zeros(steps)
scube = 0.*dens
for i in range(0, np.size(dsteps)-1):
good = np.logical_and(dens>dsteps[i],dens<dsteps[i+1]).nonzero()
hist, bin_edges = np.histogram(cosphi[good], bins=hsize, range=(-1.,1.))
bin_centre = 0.5*(bin_edges[0:np.size(bin_edges)-1]+bin_edges[1:np.size(bin_edges)])
hros[i,:] = hist
scube[good] = i
cdens[i] = np.mean([dsteps[i],dsteps[i+1]])
xi[i] = roparameter(bin_centre, hist)
outsteps = np.size(outh)
color = iter(cm.cool(np.linspace(0, 1, outsteps)))
fig = plt.figure()
for i in range(0, outsteps):
c = next(color)
labeltext = str(dsteps[outh[i]])+' < n < '+str(dsteps[outh[i]+1])
plt.plot(bin_centre, hros[outh[i],:], '-', linewidth=2, c=c, label=labeltext) #drawstyle
plt.xlabel(r'cos($\phi$)')
plt.legend()
plt.show()
fig=plt.figure()
plt.plot(cdens, xi, '-', linewidth=2, c=c)
plt.axhline(y=0., c='k', ls='--')
plt.xlabel(r'log$_{10}$ ($n_{\rm H}/$cm$^{-3}$)')
plt.ylabel(r'$\zeta$')
#plt.savefig(prefix + '-' + 'ROvsLogNH' + 'Thres' + "%d" % (thr) + '.png')
plt.show()
# Do you mean xi or zeta?
#return hros, cdens, zeta
return hros, cdens, xi
dec = np.arange(-80, 90, 20)
dec = np.append(dec, np.floor(lat / 5.) * 5)
dec = np.append(dec, np.floor(lat / 5.) * 5 - 5)
dec = np.append(dec, np.floor(lat / 5.) * 5 - 10)
dec = np.append(dec, np.floor(lat / 5.) * 5 - 15)
dec = np.append(dec, np.ceil(lat / 5.) * 5)
dec = np.append(dec, np.ceil(lat / 5.) * 5 + 5)
dec = np.append(dec, np.ceil(lat / 5.) * 5 + 10)
dec = np.append(dec, np.ceil(lat / 5.) * 5 + 15)
dec = np.unique(dec)
dec = dec[np.abs(dec) <= 90]
# http://matplotlib.org/examples/color/
# colormaps_reference.html
color = iter(cm.cool(np.linspace(0, 1, len(dec))))
'''
equations:
http://www.gb.nrao.edu/~rcreager/GBTMetrology/
140ft/l0058/gbtmemo52/memo52.html
'''
# parallactic angle
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
for d in range(dec.size):
p = np.arctan2(np.sin(ha*15.*d2r) , \
(np.tan(lat*d2r)*np.cos(dec[d]*d2r) - \
np.sin(dec[d]*d2r)*np.cos(ha*15.*d2r)))/d2r
el = np.arcsin(np.sin(dec[d]*d2r)*\