def add_cbar(fig, ax, im, **kwargs):
# deal with kwargs
kw = update_dict(add_cbar_kwargs_default, kwargs)
find_bad_keys(add_cbar_kwargs_default, kwargs, 'add_cbar')
# get fig dimensions
fig_width_inches, fig_height_inches = fig.get_size_inches()
fig_aspect = fig_height_inches / fig_width_inches
# get ax dimensions
ax_left, ax_right, ax_bottom, ax_top = axis_range(ax)
ax_width = ax_right - ax_left
ax_height = ax_top - ax_bottom
if kw.cbar_pos == 'bottom':
orientation = 'horizontal'
cbar_height = kw.cbar_thick / fig_height_inches
cbar_width = cbar_height / kw.cbar_aspect * fig_aspect
if cbar_width > kw.tol * ax_width: # don't let cbar be thicker than plot!
cbar_width = kw.tol * ax_width
# centrally position colorbar underneath the axes
label_buff = 3 / 8 / fig_height_inches # needs to contain
# the colorbar ticklabels and little buffer space
lilbit = 1 / 16 / fig_height_inches
cbar_left = ax_left + 0.5 * ax_width - 0.5 * cbar_width
cbar_bottom = ax_bottom - lilbit - cbar_height -\
(kw.cbar_no - 1)*(label_buff + cbar_height)
elif kw.cbar_pos == 'right':
orientation = 'vertical'
cbar_width = kw.cbar_thick / fig_width_inches
cbar_height = cbar_width / kw.cbar_aspect / fig_aspect
if cbar_height > kw.tol * ax_height:
cbar_height = kw.tol * ax_height
# centrally position colorbar to right of axes
label_buff = 3 / 4 / fig_width_inches # needs to contain
# the colorbar ticklabels and little buffer space
lilbit = 1 / 16 / fig_width_inches
cbar_bottom = ax_bottom + 0.5 * ax_height - 0.5 * cbar_height
cbar_left = ax_right + lilbit + (kw.cbar_no - 1) * (label_buff +
cbar_width)
cax = fig.add_axes((cbar_left, cbar_bottom, cbar_width, cbar_height))
cbar = plt.colorbar(im, cax=cax, orientation=orientation)
# deal with labels
if kw.nosci or kw.logscale:
cbar_label = kw.units
else:
cbar_label = (r'$\times10^{%i}\ $' % kw.exp) + kw.units
# font size for the tick labels
cax.tick_params(labelsize=kw.cbar_fs)
#cbar.ax.tick_params(labelsize=fontsize)
# ticklabel format
if not kw.logscale:
if kw.posdef:
tickvals = [kw.minmax[0], kw.minmax[1]]
else:
tickvals = [kw.minmax[0], 0, kw.minmax[1]]
if kw.cbar_labels is None:
fmt = '%.' + str(kw.cbar_prec) + 'f'
ticklabels = []
for tickval in tickvals:
ticklabels.append(fmt % tickval)
else:
ticklabels = kw.cbar_labels
if not kw.ticklabels is None:
ticklabels = kw.ticklabels
if not kw.tickvals is None:
tickvals = kw.tickvals
cbar.set_ticks(tickvals)
cbar.set_ticklabels(ticklabels)
else:
locator = ticker.LogLocator(subs='all')
cbar.set_ticks(locator)
if kw.cbar_pos == 'bottom':
fig.text(cbar_left + cbar_width + lilbit*fig_aspect,\
cbar_bottom + 0.5*cbar_height, cbar_label,\
ha='left', va='center', fontsize=kw.cbar_fs)
elif kw.cbar_pos == 'right':
#fig.text(cbar_left + cbar_width + lilbit/fig_aspect,\
# cbar_bottom + 0.5*cbar_height, cbar_label,\
# ha='left', va='center', fontsize=kw.fontsize)
cax.set_title(cbar_label, ha='left', fontsize=kw.cbar_fs)
def axis_label_and_ticks(self, axis, data_array, name, dim_to_shape):
"""
Get dimensions and label (if present) from requested axis.
Also retun axes tick formatters and locators.
"""
# Create some default axis tick formatter, depending on whether log
# for that axis will be True or False
formatter = {
False: ticker.ScalarFormatter(),
True: ticker.LogFormatterSciNotation()
}
locator = {False: ticker.AutoLocator(), True: ticker.LogLocator()}
dim = axis
# Convert to Dim object?
if isinstance(dim, str):
dim = sc.Dim(dim)
underlying_dim = dim
non_dimension_coord = False
var = None
if dim in data_array.coords:
dim_coord_dim = data_array.coords[dim].dims[-1]
tp = data_array.coords[dim].dtype
if tp == sc.dtype.vector_3_float64:
var = make_fake_coord(dim_coord_dim,
dim_to_shape[dim_coord_dim],
unit=data_array.coords[dim].unit)
form = ticker.FuncFormatter(lambda val, pos: "(" + ",".join([
value_to_string(item, precision=2) for item in self.
scipp_obj_dict[name].coords[dim].values[int(val)]
]) + ")" if (int(val) >= 0 and int(val) < dim_to_shape[
dim_coord_dim]) else "")
formatter.update({False: form, True: form})
locator[False] = ticker.MaxNLocator(integer=True)
if dim != dim_coord_dim:
underlying_dim = dim_coord_dim
elif tp == sc.dtype.string:
var = make_fake_coord(dim_coord_dim,
dim_to_shape[dim_coord_dim],
unit=data_array.coords[dim].unit)
form = ticker.FuncFormatter(
lambda val, pos: self.scipp_obj_dict[name].coords[
dim].values[int(val)] if (int(val) >= 0 and int(
val) < dim_to_shape[dim_coord_dim]) else "")
formatter.update({False: form, True: form})
locator[False] = ticker.MaxNLocator(integer=True)
if dim != dim_coord_dim:
underlying_dim = dim_coord_dim
elif dim != dim_coord_dim:
# non-dimension coordinate
non_dimension_coord = True
if dim_coord_dim in data_array.coords:
var = data_array.coords[dim_coord_dim]
else:
var = make_fake_coord(dim_coord_dim,
dim_to_shape[dim_coord_dim])
underlying_dim = dim_coord_dim
form = ticker.FuncFormatter(
lambda val, pos: value_to_string(data_array.coords[
dim].values[np.abs(var.values - val).argmin()]))
formatter.update({False: form, True: form})
else:
var = data_array.coords[dim]
else:
# dim not found in data_array.coords
var = make_fake_coord(dim, dim_to_shape[dim])
label = var
if non_dimension_coord:
label = data_array.coords[dim]
return underlying_dim, var, label, formatter, locator
def gen_snap_all(rarr, dir_name, base_name, snap_num, size, title_name=''):
rc('font', size=15)
fig = plt.figure()
fig.set_figwidth(10)
fig.set_figheight(8.05)
fig.subplots_adjust(left=0.08, right=0.96, top=0.95, bottom=0.06)
plt.figtext(0.5, 0.97, title_name, ha='center', size='large')
majorLocator = ticker.MultipleLocator(10)
minorLocator = ticker.MultipleLocator(5)
################### snap_den ####################################
ax = fig.add_subplot(221)
arr = np.sort(rarr, order='den')
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
den = arr['den'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
width, height = rcParams['figure.figsize']
sz = min(width, height)
# plt.xlabel('X [kpc]')
plt.ylabel('Z [kpc]')
cmap = mpl.cm.jet
norm = mpl.colors.LogNorm()
plt.scatter(x[idw], z[idw], s = 1, c = den[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = 1e-7, vmax = 10, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
cb = plt.colorbar(ticks=ticker.LogLocator(base=10.0))
cb.set_label('Density [cm$^{-3}$]')
plt.xlim(-size, size)
plt.ylim(-size, size)
######################## gen_snap_T ####################################
ax = fig.add_subplot(222)
ax.apply_aspect()
pos = ax.get_position().get_points()
x4 = pos[0][0]
width = pos[1][0] - pos[0][0]
height = pos[1][1] - pos[0][1]
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
T = arr['T'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
# plt.xlabel('X [kpc]')
# plt.ylabel('Z [kpc]')
cmap = mpl.cm.jet
norm = mpl.colors.LogNorm()
plt.scatter(x[idw], z[idw], s = 1, c = T[idw], marker = 'o', cmap = cmap, \
norm = norm, vmin = 1e3, vmax = 1e7, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
cb = plt.colorbar(ticks=ticker.LogLocator(base=10.0))
cb.set_label('Temperature [K]')
plt.xlim(-size, size)
plt.ylim(-size, size)
################### gen_snap_cold_mass ####################################
ax = fig.add_subplot(223)
ax.apply_aspect()
pos = ax.get_position().get_points()
y4 = pos[0][1]
arr = np.sort(rarr, order='mass')
# arr = rarr
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
mass = arr['mass'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
if np.size(idc) == 0:
return
plt.xlabel('X [kpc]')
plt.ylabel('Z [kpc]')
# plt.subplots_adjust(left = 0.13, right = 0.93, top = 0.93, bottom = 0.13)
cmap = mpl.cm.jet
norm = mpl.colors.LogNorm()
plt.scatter(x[idc], z[idc], s = 2, c = mass[idc], marker = 'o', \
cmap = cmap, norm = norm, vmin = 1e3, vmax = 1e7, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
cb = plt.colorbar(ticks=ticker.LogLocator(base=10.0))
cb.set_label('Cold Cloud Mass [M$_\odot$]')
plt.xlim(-size, size)
plt.ylim(-size, size)
################### dm_star ####################################
ax = fig.add_subplot(224, aspect='equal')
ax.set_anchor('W')
# ax = fig.add_axes([x4, y4, width, height])
arr = np.sort(rarr, order='mass')
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
mass = arr['mass'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
idm = np.where(arr['type'] == mp.TYPE_DM_PARAL)
# if np.size(idc) == 0:
# return
plt.xlabel('X [kpc]')
# plt.ylabel('Z [kpc]')
# plt.subplots_adjust(left = 0.13, right = 0.93, top = 0.93, bottom = 0.13)
plt.scatter(x[idm], z[idm], s = 1, c = 'k', marker = 'o', \
edgecolors = 'none')
if (np.size(ids) > 0):
plt.scatter(x[ids], z[ids], s = 2, c = 'r', marker = 'o', \
edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
plt.xlim(-size, size)
plt.ylim(-size, size)
########################## save ######################################
print 'writing xz...'
figure_name = 'Panel-xz-' + dir_name + '-' + base_name + '-' \
+ snap_num + '.png'
plt.savefig(figure_name)
######################################################################
################################## X Y ###############################
######################################################################
plt.clf()
fig = plt.figure()
fig.set_figwidth(10)
fig.set_figheight(8.05)
fig.subplots_adjust(left=0.08, right=0.96, top=0.95, bottom=0.06)
plt.figtext(0.5, 0.97, title_name, ha='center', size='large')
################### snap_den ####################################
ax = fig.add_subplot(221)
arr = np.sort(rarr, order='den')
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
den = arr['den'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
width, height = rcParams['figure.figsize']
sz = min(width, height)
# plt.xlabel('X [kpc]')
plt.ylabel('Y [kpc]')
cmap = mpl.cm.jet
norm = mpl.colors.LogNorm()
plt.scatter(x[idw], y[idw], s = 1, c = den[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = 1e-7, vmax = 10, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
cb = plt.colorbar(ticks=ticker.LogLocator(base=10.0))
cb.set_label('Density [cm$^{-3}$]')
plt.xlim(-size, size)
plt.ylim(-size, size)
######################## gen_snap_T ####################################
ax = fig.add_subplot(222)
ax.apply_aspect()
pos = ax.get_position().get_points()
x4 = pos[0][0]
width = pos[1][0] - pos[0][0]
height = pos[1][1] - pos[0][1]
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
T = arr['T'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
# plt.xlabel('X [kpc]')
# plt.ylabel('Z [kpc]')
cmap = mpl.cm.jet
norm = mpl.colors.LogNorm()
plt.scatter(x[idw], y[idw], s = 1, c = T[idw], marker = 'o', cmap = cmap, \
norm = norm, vmin = 1e3, vmax = 1e7, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
cb = plt.colorbar(ticks=ticker.LogLocator(base=10.0))
cb.set_label('Temperature [K]')
plt.xlim(-size, size)
plt.ylim(-size, size)
################### gen_snap_cold_mass ####################################
ax = fig.add_subplot(223)
ax.apply_aspect()
pos = ax.get_position().get_points()
y4 = pos[0][1]
arr = np.sort(rarr, order='mass')
# arr = rarr
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
mass = arr['mass'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
if np.size(idc) == 0:
return
plt.xlabel('X [kpc]')
plt.ylabel('Y [kpc]')
# plt.subplots_adjust(left = 0.13, right = 0.93, top = 0.93, bottom = 0.13)
cmap = mpl.cm.jet
norm = mpl.colors.LogNorm()
plt.scatter(x[idc], y[idc], s = 2, c = mass[idc], marker = 'o', \
cmap = cmap, norm = norm, vmin = 1e3, vmax = 1e7, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
cb = plt.colorbar(ticks=ticker.LogLocator(base=10.0))
cb.set_label('Cold Cloud Mass [M$_\odot$]')
plt.xlim(-size, size)
plt.ylim(-size, size)
################### dm_star ####################################
ax = fig.add_subplot(224, aspect='equal')
ax.set_anchor('W')
# ax = fig.add_axes([x4, y4, width, height])
arr = np.sort(rarr, order='mass')
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
mass = arr['mass'][:]
idw = np.where(arr['type'] == mp.TYPE_WARM)
idc = np.where(arr['type'] == mp.TYPE_COLD)
ids = np.where(arr['type'] == mp.TYPE_STAR_PARAL)
idm = np.where(arr['type'] == mp.TYPE_DM_PARAL)
# if np.size(idc) == 0:
# return
plt.xlabel('X [kpc]')
# plt.ylabel('Z [kpc]')
# plt.subplots_adjust(left = 0.13, right = 0.93, top = 0.93, bottom = 0.13)
plt.scatter(x[idm], y[idm], s = 1, c = 'k', marker = 'o', \
edgecolors = 'none')
if (np.size(ids) > 0):
plt.scatter(x[ids], y[ids], s = 2, c = 'r', marker = 'o', \
edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
plt.xlim(-size, size)
plt.ylim(-size, size)
########################## save ######################################
figure_name = 'Panel-xy-' + dir_name + '-' + base_name + '-' \
+ snap_num + '.png'
plt.savefig(figure_name)
def multi_T(dir_name, base_name, nproc, size, snap_num):
title = dir_name
rc('font', size=18)
fig = plt.figure()
fig.set_figwidth(10)
fig.set_figheight(9)
fig.subplots_adjust(left = 0.10, right = 0.85, top = 0.95, bottom = 0.10, \
wspace = 0.0, hspace = 0.0)
# plt.figtext(0.5, 0.94, title, ha = 'center', size = 'large')
majorLocator = ticker.MultipleLocator(20)
minorLocator = ticker.MultipleLocator(4)
nullFormatter = ticker.NullFormatter()
cmap = mpl.cm.jet
norm = mpl.colors.LogNorm()
vmin = 1.0E3
vmax = 1.0E7
#######################################################################
########################### X - Z ####################################
#######################################################################
rc('font', size=18)
plt.clf()
fig = plt.figure()
fig.set_figwidth(10)
fig.set_figheight(9)
fig.subplots_adjust(left = 0.10, right = 0.85, top = 0.95, bottom = 0.10, \
wspace = 0.0, hspace = 0.0)
# plt.figtext(0.5, 0.94, title, ha = 'center', size = 'large')
############################### snap 0 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[0], nproc)
ax = fig.add_subplot(221)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
# plt.xlabel('X [kpc]')
plt.ylabel('Z [kpc]')
plt.scatter(x[idw], z[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
ax.xaxis.set_major_formatter(nullFormatter)
# ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
################################# snap 1 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[1], nproc)
ax = fig.add_subplot(222)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
# plt.xlabel('X [kpc]')
# plt.ylabel('Z [kpc]')
plt.scatter(x[idw], z[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
ax.xaxis.set_major_formatter(nullFormatter)
ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
################################# snap 2 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[2], nproc)
ax = fig.add_subplot(223)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
plt.xlabel('X [kpc]')
plt.ylabel('Z [kpc]')
plt.scatter(x[idw], z[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
# ax.xaxis.set_major_formatter(nullFormatter)
# ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
################################ snap 3 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[3], nproc)
ax = fig.add_subplot(224)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
plt.xlabel('X [kpc]')
# plt.ylabel('Z [kpc]')
plt.scatter(x[idw], z[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
# ax.xaxis.set_major_formatter(nullFormatter)
ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
######################### color bar ##################################
# [xcoord, ycoord, width, height]
cbaxes = fig.add_axes([0.87, 0.10, 0.03, 0.85])
cb = plt.colorbar(orientation='vertical', \
ticks = ticker.LogLocator(base = 10.0), cax = cbaxes)
cb.set_label('Temperature [K]')
######################## label ######################################
label = 't = 0.5 Gyr'
plt.figtext(0.28, 0.9, label, ha='center', size=20)
label = 't = 1 Gyr'
plt.figtext(0.68, 0.9, label, ha='center', size=20)
label = 't = 2 Gyr'
plt.figtext(0.28, 0.47, label, ha='center', size=20)
label = 't = 3 Gyr'
plt.figtext(0.68, 0.47, label, ha='center', size=20)
########################## save ######################################
figure_name = 'TPanel-xz-' + dir_name + '.png'
plt.savefig(figure_name)
#######################################################################
########################### X - Y ####################################
#######################################################################
rc('font', size=18)
plt.clf()
fig = plt.figure()
fig.set_figwidth(10)
fig.set_figheight(9)
fig.subplots_adjust(left = 0.10, right = 0.85, top = 0.95, bottom = 0.10, \
wspace = 0.0, hspace = 0.0)
# plt.figtext(0.5, 0.94, title, ha = 'center', size = 'large')
############################### snap 0 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[0], nproc)
ax = fig.add_subplot(221)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
# plt.xlabel('X [kpc]')
plt.ylabel('Y [kpc]')
plt.scatter(x[idw], y[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
ax.xaxis.set_major_formatter(nullFormatter)
# ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
################################# snap 1 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[1], nproc)
ax = fig.add_subplot(222)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
# plt.xlabel('X [kpc]')
# plt.ylabel('Y [kpc]')
plt.scatter(x[idw], y[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
ax.xaxis.set_major_formatter(nullFormatter)
ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
################################# snap 2 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[2], nproc)
ax = fig.add_subplot(223)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
plt.xlabel('X [kpc]')
plt.ylabel('Y [kpc]')
plt.scatter(x[idw], y[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
# ax.xaxis.set_major_formatter(nullFormatter)
# ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
################################ snap 3 ####################################
ntot, time, rarr = mp.read_snap_paral(mp.prefix, dir_name, base_name,
snap_num[3], nproc)
ax = fig.add_subplot(224)
arr = np.sort(rarr, order='T')
arr = arr[::-1]
x = arr['pos'][:, 0]
y = arr['pos'][:, 1]
z = arr['pos'][:, 2]
type = arr['type'][:]
r = np.sqrt(x * x + y * y + z * z)
T = arr['T'][:]
boolarr = np.logical_and(type == mp.TYPE_WARM, r < 50.0)
# boolarr = np.logical_and(boolarr, np.abs(z) < 1.0)
idw = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_COLD, r < 50.0)
idc = np.where(boolarr)
boolarr = np.logical_and(type == mp.TYPE_STAR_PARAL, r < 50.0)
ids = np.where(boolarr)
width, height = rcParams['figure.figsize']
sz = min(width, height)
plt.xlabel('X [kpc]')
# plt.ylabel('Y [kpc]')
plt.scatter(x[idw], y[idw], s = 1, c = T[idw], marker = 'o', \
cmap = cmap, norm = norm, vmin = vmin, vmax = vmax, edgecolors = 'none')
ax.xaxis.set_major_locator(majorLocator)
ax.xaxis.set_minor_locator(minorLocator)
ax.yaxis.set_major_locator(majorLocator)
ax.yaxis.set_minor_locator(minorLocator)
# ax.xaxis.set_major_formatter(nullFormatter)
ax.yaxis.set_major_formatter(nullFormatter)
plt.xlim(-size, size * 0.99)
plt.ylim(-size, size)
######################### color bar ##################################
# [xcoord, ycoord, width, height]
cbaxes = fig.add_axes([0.87, 0.10, 0.03, 0.85])
cb = plt.colorbar(orientation='vertical', \
ticks = ticker.LogLocator(base = 10.0), cax = cbaxes)
cb.set_label('Temperature [K]')
######################## label ######################################
label = 't = 0.5 Gyr'
plt.figtext(0.28, 0.9, label, ha='center', size=20)
label = 't = 1 Gyr'
plt.figtext(0.68, 0.9, label, ha='center', size=20)
label = 't = 2 Gyr'
plt.figtext(0.28, 0.47, label, ha='center', size=20)
label = 't = 3 Gyr'
plt.figtext(0.68, 0.47, label, ha='center', size=20)
########################## save ######################################
figure_name = 'TPanel-xy-' + dir_name + '.png'
plt.savefig(figure_name)
exact = np.zeros((model.nt,model.nx))
for k in range(0,model.nt):
exact[k,:] = ufunc(model.dt*k,model.x,model.κ)
error_kappas[j,i] = LA.norm(Strang_Beam_kap- exact,np.inf)
xx, yy = np.meshgrid(1/nxs,kappas)
Pe = 4*xx/yy
from matplotlib import ticker, cm
fig, ax = plt.subplots(1, 2, sharex=True, sharey=True, figsize=(17, 8))
pes = ax[0].contourf(1/nxs,kappas,Pe,levels = np.logspace(-3,1,9), locator=ticker.LogLocator(), cmap=cm.PuBu_r)
cbar0 = fig.colorbar(pes, ax = ax[0])
cbar0.ax.set_title('$Pe$',fontsize='large', pad = 20)
ax[0].set_ylabel(r'$\kappa$',fontsize='xx-large')
ax[0].set_xlabel(r'$\Delta x$',fontsize='xx-large')
ax[0].set_title('Peclet Number',fontsize='large')
cs = ax[1].contourf(1/nxs,kappas,error_kappas, levels = np.logspace(-1,2,19), locator=ticker.LogLocator(), cmap=cm.viridis)
cbar1 = fig.colorbar(cs, ax = ax[1])
cbar1.ax.set_title('$||U - u||_\infty$',fontsize='large', pad = 20)
ax[1].set_xlabel(r'$\Delta x$',fontsize='xx-large')
ax[1].set_title('Error',fontsize='large')
plt.tight_layout()
def performance(tol):
"""
============================================================================
Assesses the performance of Bracket Descent and Scipy L-BFGS-B Methods
============================================================================
Parameters
------------
tol : float
Determines the tolerance for minimization
Returns
------------
This function will produce 4 figures.
The first 3 will represent a comparison of the precison of each method while
the 4th will represent a comparison of the timing.
The first three show the location of the computed minima for initial guesses
of [-100,-3], [-50,-3], [-10,-3] and [-1,-3]. These are overlayed onto the
original cost function; the Scipy L-BFGS-B results are represented by red
diamonds while the Bracket Descent results are represented by blue diamonds.
The three figures represent the cases when the noise amplitude is set to 0,
1, and 10.
The final figure consists of four subplots, the upper row represents the
computational time taken for convergence, given an initial x starting point,
while the lower represents the number of iterations requried. In each case
the Scipy L-BFGS-B method is shown on the left and the Bracket descent is
shown on the right. A legend on each plot differentiates the cases when the
Noise Ampplitude is set to 0, 1, and 10.
Trends Observed
----------------
For all cases, the Scipy minimization function appears to be more consistent
(to rely less on the initial guess) than the fortran Bracket Descent method.
This is seen in figures hw241-hw243, where the B-D results are seen to cover
a broader spead of final coordinated. These figures also illustrate that as
the level of noise of the cost function is increased, the Scipy L-BFGS-B
method becomes increasingly favourable over the Bracket descent approach,
producing more precise results each time.
This is a result of the lack of consideration for noise within the Bracket
Descent method; that is to say that any random fluctations which result in
two neighbouring points (along the convergence path) lying within the
tolerance limit will be assumed to be the true minimum of the function as
defined by the B-D method. However, it is likely that the Scipy L-BFGS-B
method is adapted to smooth out noisy functions and hence find the true
minimum more reliably.
A consideration of figure hw244, however, demonstrates an advantage of the
B-D method over the Scipy L-BGFS-B minimization in the form of timing. It can
be seen that despite requiring more iterations before converging to within a
set tolerance, the total computational time is less to within a factor of 10.
"""
plt.close('all')
count = 0
hw2.tol = tol
nintb = []
nintl = []
tlbfgsb = []
txfbd = []
lbfgsx = []
lbfgsy = []
xfbdx = []
xfbdy = []
cost.c_noise = True
for cost.c_noise_amp in [0., 1., 10.]:
count = count + 1
for [x, y] in [[-100., -3.], [-50., -3.], [-10., -3.], [-1., -3.]]:
t12 = 0
t34 = 0
for i in range(0, 1000):
t1 = time()
scipy.optimize.minimize(cost.costj, [x, y],
method='L-BFGS-B',
tol=tol)
t2 = time()
t12 = t12 + (t2 - t1)
t3 = time()
hw2.bracket_descent([x, y])
t4 = time()
t34 = t34 + (t4 - t3)
tlbfgsb.append(t12 / 1000)
txfbd.append(t34 / 1000)
info = scipy.optimize.minimize(cost.costj, [x, y],
method='L-BFGS-B',
tol=tol)
xfbd, jfbd, i2 = hw2.bracket_descent([x, y])
# print('method: ', 'Fortran Bracket Descent')
# print('Value: ', jfbd)
# print('number of iterations:', i2)
# print('x: ', xfbd)
# print('c_noise: ', cost.c_noise)
# print(' ')
#print(info)
x = info.x
lbfgsx.append(x[0])
lbfgsy.append(x[1])
xfbdx.append(xfbd[0])
xfbdy.append(xfbd[1])
nint = info.nit
nintl.append(nint)
nintb.append(i2)
Minx = 1 + (min([
min(xfbdx[(count - 1) * 4:count * 4]),
min(lbfgsx[(count - 1) * 4:count * 4])
]) - 1) * 1.1
Maxx = 1 + (max([
max(xfbdx[(count - 1) * 4:count * 4]),
max(lbfgsx[(count - 1) * 4:count * 4])
]) - 1) * 1.1
Miny = 1 + (min([
min(xfbdy[(count - 1) * 4:count * 4]),
min(lbfgsy[(count - 1) * 4:count * 4])
]) - 1) * 1.1
Maxy = 1 + (max([
max(xfbdy[(count - 1) * 4:count * 4]),
max(lbfgsy[(count - 1) * 4:count * 4])
]) - 1) * 1.1
[X, Y] = np.linspace(Minx, Maxx, 200), np.linspace(Miny, Maxy, 200)
#calculate noiseless cost function at each point on 2D grid
j = [[cost.costj([xi, yi]) for xi in X] for yi in Y]
#create contour plots of cost functions with and without noise
fig, p4 = plt.subplots()
cp = p4.contourf(X, Y, j, locator=ticker.LogLocator(), cmap=cm.GnBu)
cbar = fig.colorbar(cp)
BD, = p4.plot(xfbdx[(count - 1) * 4:count * 4],
xfbdy[(count - 1) * 4:count * 4],
'b',
linestyle='None',
marker='d',
markersize=6)
Scipy, = p4.plot(lbfgsx[(count - 1) * 4:count * 4],
lbfgsy[(count - 1) * 4:count * 4],
'r',
linestyle='None',
marker='d',
markersize=6)
BD.set_label('Fortran Bracket Descent')
Scipy.set_label('Scipy optimize L-BFGS-B')
plt.legend(loc='upper left', fontsize='small')
plt.suptitle(
'Rosemary Teague, performance \n Comparison of converged values, Noise='
+ str(int(cost.c_noise_amp)))
#plt.tight_layout(pad=5)
plt.savefig('hw24' + str(count), dpi=700)
print(tlbfgsb)
plt.close('all')
f4, (p414, p424) = plt.subplots(2, 2, sharey=True)
one, = p414[0].plot(
tlbfgsb[:4],
[np.abs(-100.), np.abs(-50.),
np.abs(-10.), np.abs(-1.)],
'r',
marker='x',
markersize=12)
two, = p414[0].plot(
tlbfgsb[4:8],
[np.abs(-100.), np.abs(-50.),
np.abs(-10.), np.abs(-1.)],
'm',
marker='x',
markersize=12)
three, = p414[0].plot(
tlbfgsb[8:],
[np.abs(-100.), np.abs(-50.),
np.abs(-10.), np.abs(-1.)],
'#c79fef',
marker='x',
markersize=12)
one.set_label('No Noise')
two.set_label('Noise = 1.0')
three.set_label('Noise = 10.0')
p414[0].set_title('Scipy Optimise L-BFGS-B')
p414[0].set_xlabel('Time Taken')
p414[0].legend(loc='upper right', fontsize='x-small')
p414[0].xaxis.set_ticks(np.linspace(min(tlbfgsb), max(tlbfgsb), 3))
p414[0].ticklabel_format(useOffset=False)
uno, = p414[1].plot(txfbd[:4], [
np.abs(-100. - xfbdx[0]),
np.abs(-50. - xfbdx[1]),
np.abs(-10. - xfbdx[2]),
np.abs(-1. - xfbdx[3])
],
'b',
marker='x',
markersize=12)
dos, = p414[1].plot(txfbd[4:8], [
np.abs(-100. - xfbdx[4]),
np.abs(-50. - xfbdx[5]),
np.abs(-10. - xfbdx[6]),
np.abs(-1. - xfbdx[7])
],
'g',
marker='x',
markersize=12)
tres, = p414[1].plot(txfbd[8:], [
np.abs(-100. - xfbdx[8]),
np.abs(-50. - xfbdx[9]),
np.abs(-10. - xfbdx[10]),
np.abs(-1. - xfbdx[11])
],
'c',
marker='x',
markersize=12)
uno.set_label('No Noise')
dos.set_label('Noise = 1.0')
tres.set_label('Noise = 10.0')
p414[1].set_title('Fortran Bracket Descent')
p414[1].set_xlabel('Time Taken')
p414[1].legend(loc='upper left', fontsize='x-small')
p414[1].xaxis.set_ticks(np.linspace(min(txfbd), max(txfbd), 3))
one1, = p424[0].plot(nintl[:4], [
np.abs(-100. - lbfgsx[0]),
np.abs(-50. - lbfgsx[1]),
np.abs(-10. - lbfgsx[2]),
np.abs(-1. - lbfgsx[3])
],
'r',
marker='x',
markersize=12)
two2, = p424[0].plot(nintl[4:8], [
np.abs(-100. - lbfgsx[4]),
np.abs(-50. - lbfgsx[5]),
np.abs(-10. - lbfgsx[6]),
np.abs(-1. - lbfgsx[7])
],
'm',
marker='x',
markersize=12)
three3, = p424[0].plot(nintl[8:], [
np.abs(-100. - lbfgsx[8]),
np.abs(-50. - lbfgsx[9]),
np.abs(-10. - lbfgsx[10]),
np.abs(-1. - lbfgsx[11])
],
'#c79fef',
marker='x',
markersize=12)
one1.set_label('No Noise')
two2.set_label('Noise = 1.0')
three3.set_label('Noise = 10.0')
p424[0].set_xlabel('Number of Iterations')
p424[0].legend(loc='upper left', fontsize='x-small')
p424[0].ticklabel_format(useOffset=False)
uno1, = p424[1].plot(nintb[:4], [
np.abs(-100. - xfbdx[0]),
np.abs(-50. - xfbdx[1]),
np.abs(-10. - xfbdx[2]),
np.abs(-1. - xfbdx[3])
],
'b',
marker='x',
markersize=12)
dos2, = p424[1].plot(nintb[4:8], [
np.abs(-100. - xfbdx[4]),
np.abs(-50. - xfbdx[5]),
np.abs(-10. - xfbdx[6]),
np.abs(-1. - xfbdx[7])
],
'g',
marker='x',
markersize=12)
tres3, = p424[1].plot(nintb[8:], [
np.abs(-100. - xfbdx[8]),
np.abs(-50. - xfbdx[9]),
np.abs(-10. - xfbdx[10]),
np.abs(-1. - xfbdx[11])
],
'c',
marker='x',
markersize=12)
uno1.set_label('No Noise')
dos2.set_label('Noise = 1.0')
tres3.set_label('Noise = 10.0')
p424[1].set_xlabel('Number of Iterations')
p424[1].legend(loc='upper left', fontsize='x-small')
f4.text(0.04,
0.5,
'Initial x-distance from Converged minimum',
va='center',
rotation='vertical')
plt.suptitle(
'Rosemary Teague, performance \n Time taken for values to converge',
fontsize='large')
plt.tight_layout(pad=3.5, h_pad=1, w_pad=1)
plt.savefig('hw244', dpi=700)
f = figure(figsize =(3.2,2.5), dpi = 220*2.0/3.0 * 2.0)
N, K = meshgrid(nlist, klist, indexing = 'ij')
ylabel('k', fontsize = 10)
xlabel('n',fontsize = 10)
title('$\lambda$ = %i nm' % lamda)
minorticks_on()
gca().tick_params(axis='both', which='major', labelsize=10)
if use_logscale:
cf_levels = compute_outer_log_levels(error_map, base10delta = 0.1)
CF = contourf(N,K, error_map, norm = LogNorm(), levels = cf_levels , locator=ticker.LogLocator() )
cbar = colorbar(CF, ticks = compute_inner_log_levels(error_map, base10delta = 0.2))
cbar.ax.set_ylabel('RMS Error (%)')
#cbar.ax.yaxis.set_ticks(log10(cf_levels) , minor=True)
else:
cf_level_delta = 0.5
error_min = error_map.min()
error_max = error_map.max()
cfmax = (error_max - error_min) *0.15 +error_min
cf_levels = arange( floor(error_min), ceil(cfmax) ,cf_level_delta )
CF = contourf(N,K, error_map, levels = cf_levels )
cbar = colorbar(CF)
cbar.ax.set_ylabel('RMS Error (%)')
min_indices = find_min_indices_2d_array(error_map)
#****************************************
#Plot Number 1 (start)
#****************************************
fig1, ax1 = plt.subplots(2, 1, sharex=True)
plt.xlabel('$P/P_{sat}$', fontname="Arial", fontsize=axis_Label_font_size)
plt.ylabel(' $\u03BE$ (No. waters / nm$^2$)',
fontname="Arial",
fontsize=axis_Label_font_size)
ax1[0].set_xticks(np.arange(0, 100, 10))
ax1[1].set_xticks(np.arange(0, 100, 10))
ax1[0].set_yticks(np.arange(-200, 200, E_per_nm_sq_10A_step))
ax1[1].set_yticks(np.arange(-200, 200, E_per_nm_sq_16A_step))
ax1[0].xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=5))
ax1[1].xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=5))
WSU_color = '#ff7f0e'
NDU_color = '#1f77b4'
ax1[0].errorbar(P_div_Po_10A_ads_Cass_list,
Avg_of_E_per_area_10A_ads_Cass_list,
Std_Dev_E_per_area_10A_ads_Cass_list,
color=NDU_color,
marker='D',
linestyle='-',
markersize=PointSizes + 2,
linewidth=ConnectionLineSizes,
fillstyle='full',
label="Cassandra - adsorb",
def psd_plot(pore_widths,
pore_dist,
method=None,
labeldiff='distribution',
labelcum='cumulative',
line_style=None,
log=True,
right=None,
left=None,
ax=None):
"""
Draw a pore size distribution plot.
Parameters
----------
pore_widths : array
Array of the pore radii which will become the x axis.
pore_dist : array
Contribution of each pore radius which will make up the y axis.
method : str
The method used. Will be a string part of the title.
labeldiff : str
The label for the plotted data, which will appear in the legend.
labelcum : str, optional
The label for the cumulative data, which will appear in the legend.
Set to None to remove cumulative distribution
line_style : dict, optional
The style dictionary to send to the plot() function.
log : int
Whether to display a logarithmic graph.
right : int
Higher bound of the selected pore widths.
right : int
Lower bound of the selected pore widths.
ax : matplotlib axes object, default None
The axes object where to plot the graph if a new figure is
not desired.
Returns
-------
matplotlib.axes
Matplotlib axes of the graph generated. The user can then apply their
own styling if desired.
"""
# Generate the figure if needed
if ax is None:
fig = plt.figure(figsize=(15, 5))
ax = fig.add_subplot(111)
lst = {'marker': '', 'color': 'k'}
if line_style is not None:
lst.update(line_style)
l1 = ax.plot(pore_widths, pore_dist, label=labeldiff, **lst)
if labelcum:
ax2 = ax.twinx()
l2 = ax2.plot(pore_widths[1:],
numpy.cumsum(pore_dist[1:] * numpy.diff(pore_widths)),
marker='',
color='r',
linestyle="--",
label=labelcum)
# Func formatter
def formatter(x, pos):
return "{0:g}".format(x)
if log:
ax.set_xscale('log')
ax.xaxis.set_minor_formatter(ticker.FuncFormatter(formatter))
ax.xaxis.set_major_formatter(ticker.FuncFormatter(formatter))
ax.xaxis.set_major_locator(
ticker.LogLocator(base=10.0, numticks=15, numdecs=20))
ax.tick_params(axis='x',
which='minor',
width=0.75,
length=2.5,
**TICK_STYLE)
ax.tick_params(axis='x',
which='major',
width=2,
length=10,
**TICK_STYLE)
ax.set_xlim(left=left, right=right)
else:
if not left:
left = 0
ax.set_xlim(left=left, right=right)
ax.tick_params(axis='both', which='major', **TICK_STYLE)
ax.set_title("PSD plot " + str(method), **TITLE_STYLE)
ax.set_xlabel('Pore width (nm)', **LABEL_STYLE)
ax.set_ylabel('Distribution (dV/dw)', **LABEL_STYLE)
if labelcum:
ax2.set_ylabel('Cumulative Vol ($cm^3 g^{-1}$)', **LABEL_STYLE)
lns = l1
if labelcum:
lns = l1 + l2
labs = [l.get_label() for l in lns]
ax.legend(lns, labs, loc='lower right')
ax.set_ylim(bottom=0)
if labelcum:
ax2.set_ylim(bottom=0)
ax.grid(True)
return ax
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import ticker, cm
R = 1 + 0.3j
L = -2 - 0.1j
def y(omega, p, R=R, L=L):
return omega * (R * (1 + p) + L * (1 - p))
ps = np.linspace(0, 1.)
omegas = np.logspace(-1, 1)
omegabar = 1
pbar = 1.
def chisquare(omega, p):
return abs(y(omega, p) - y(omegabar, pbar))**2
O, P = np.meshgrid(omegas, ps)
cs = plt.contourf(O, P, chisquare(O, P), locator=ticker.LogLocator(numticks=9))
plt.colorbar(cs)
plt.xscale('log')
# %%
coeffs.write_file("/tmp/REXI_"+rexi_method+"_"+unique_id_string+"_txt.txt", False)
coeffs.write_file("/tmp/REXI_"+rexi_method+"_"+unique_id_string+"_bin.txt", True)
#Save last names used
met_names[imet]=unique_id_string
fig, axes = plt.subplots(m_met,2, figsize=(12,12))
plt.suptitle(r'$e^{ix}$'+" reconstruction error")
levels = np.geomspace(1e-16, 1, num=17)
cmap=cm.viridis
X, Y = np.meshgrid(x_points, K_tests )
axes1d=axes.reshape(-1)
for i, rexi_method in enumerate(rexi_methods):
cs = axes1d[i].contourf(X, Y, errors[i], levels, locator=ticker.LogLocator(), cmap=cmap)
axes1d[i].set(xlabel="x", ylabel="Poles", title=str(met_names[i]))
fig.tight_layout(pad=4.0)
fig.subplots_adjust(right=0.8)
fig.subplots_adjust(top=0.9)
cbar_ax = fig.add_axes([0.85, 0.15, 0.05, 0.7])
cbar=fig.colorbar(cs, cax=cbar_ax)
cbar.set_label('Error in exp(ix)')
plt.figtext(0.3,0.01, "lambda_include_imag="+str(lambda_include_imag)+" lambda_max_real="+str(lambda_max_real) )
plt.savefig("expi_recon.png")
plt.show()
def test_bad_locator_subs(sub):
ll = mticker.LogLocator()
with pytest.raises(ValueError):
ll.subs(sub)
for i in range(times_reduced_gp3.shape[0]):
parallel_efficiency_gp3[i] = times_reduced_gp3[0]/(times_reduced_gp3[i]*nprocs[i])
# -- We prepare the plot area.
bar_width = 1e-1 # Width of the bars in the bar plot.
fig = plt.figure(figsize=(4,2.5))
ax = fig.add_subplot(111)
vphys.adjust_spines(ax, ['left', 'bottom'], 2)
#ax.spines['bottom'].set_position(('outward',0.0))
ax.set_xlim(0.75,1.5*np.amax(nprocs))
ax.set_ylim(0,1.05)
ax.set_xscale('log')
ax.xaxis.set_major_locator(ticker.LogLocator(base=2))
ax.xaxis.set_major_formatter(ticker.StrMethodFormatter("{x:1.0f}"))
ax.yaxis.set_major_formatter(ticker.StrMethodFormatter("{x:1.1f}"))
ax.set_xticks(nprocs)
ax.set_xlabel(r"Number of processors")
ax.set_ylabel(r"Parallel Efficiency on GP3", rotation='horizontal', ha='left')
ax.yaxis.set_label_coords(-0.1,1.05)
ax.grid(True,color='gray', linestyle="-", lw=0.3, zorder=0)
# -- Remove minor tick marks.
for toc in ax.xaxis.get_minor_ticks():
toc.tick1On = toc.tick2On = False
width = 1e-1
def GenerateData():
plt.rcParams['font.serif'] = "Times New Roman"
plt.rcParams['font.family'] = "serif"
plt.rcParams['font.weight'] = "light"
plt.rcParams['font.size'] = 14
plt.rcParams['mathtext.fontset'] = 'cm'
plt.rcParams['mathtext.rm'] = 'serif'
M1 = 10
M2 = 3
A = 0.75
comx, comy = COM(M1, M2, A)
temp = Roche(comx, comy, M1, M2, A)
print("COM (", comx, ",", comy, ") = ", temp)
cblabel = "Gravitational Potential " r"$\Phi$"
Factor = 1
steps = 1000
bound = 1
#lines = [0, 2, 5, 6,7, 10, 50, 75, 100, 150, 200, 250, 300, 400, 500]
lines = 1000
xlist = np.linspace(-bound, bound, steps)
ylist = np.linspace(-bound, bound, steps)
x, y = np.meshgrid(xlist, ylist)
z = [[0 for k in range(steps)] for l in range(steps)]
i = 0
while i < steps:
j = 0
while j < steps:
z[i][j] = (Factor * Roche(xlist[i], ylist[j], M1, M2, A))
if (z[i][j] - 0)**2 < 3:
print("0 = (", xlist[i], ", ", ylist[j], ") ")
j = j + 1
i = i + 1
fig = plt.figure()
ax = fig.add_subplot(111)
ax.set_aspect("equal")
cp = plt.contourf(y,
x,
z,
lines,
cmap=plt.cm.bone,
locator=ticker.LogLocator(base=10.0, subs="all"),
vmax=100)
cb = fig.colorbar(cp, ticks=[1, 10, 100, 1000], label=cblabel)
ax.text(A / 2 - 0.06,
-0.26,
r"$\mathbf{M_2}$",
color='White',
fontsize=14,
fontweight='bold')
ax.text(-A / 2 - 0.05,
-0.26,
r"$\mathbf{M_1}$",
color='White',
fontsize=14,
fontweight='bold')
#cp2 = plt.contour(cp, levels=[5.667865], colors='r')
#cp3 = plt.contour(cp, levels=[5.66555], colors='r')
cp4 = plt.contour(cp, levels=[4.9281], colors='r')
plt.plot([A / 2], [0],
"o",
color="Yellow",
markersize=6 * (M2 / (M1 + M2)))
plt.plot([-A / 2], [0],
"o",
color="Yellow",
markersize=6 * (M1 / (M1 + M2)))
plt.axis('off')
# plt.legend(loc="best")
plt.savefig("RocheLobe.png", format='png', dpi=1200)
plt.show()
return x, y, z
def compute_dust_mass_volume_density():
# extra useful quantities (code units)
Rinf = par.gas.redge[0:len(par.gas.redge) - 1]
Rsup = par.gas.redge[1:len(par.gas.redge)]
surface = np.zeros(
par.gas.data.shape) # 2D array containing surface of each grid cell
surf = np.pi * (Rsup * Rsup - Rinf * Rinf
) / par.gas.nsec # surface of each grid cell (code units)
for th in range(par.gas.nsec):
surface[:, th] = surf
# Mass of gas in units of the star's mass
Mgas = np.sum(par.gas.data * surface)
# if hydro simulation is 2D, we first need to compute the dust's
# surface density before getting its mass volume density
if par.hydro2D == 'Yes':
dustcube = compute_dust_mass_surface_density()
print('--------- computing dust mass surface density ----------')
print('--------- computing dust mass volume density ----------')
DUSTOUT = open('dust_density.inp', 'w')
DUSTOUT.write('1 \n') # iformat
DUSTOUT.write(str(par.gas.nrad * par.gas.nsec * par.gas.ncol) +
' \n') # n cells
DUSTOUT.write(str(int(par.nbin)) + ' \n') # nbin size bins
# array (ncol, nbin, nrad, nsec)
rhodustcube = np.zeros(
(par.gas.ncol, par.nbin, par.gas.nrad, par.gas.nsec))
# dust aspect ratio as function of ibin and r (or actually, R, cylindrical radius)
hd = np.zeros((par.nbin, par.gas.nrad))
# work out averaged Stokes number per size bin with fargo3d
# Epstein regime assumed so far in the code -> St ~ pi/2 (s x rho_int) / Sigma_gas
# where Sigma_gas should denote the azimuthally-averaged gas surface density here
# so that St is understood as a 2D array (nbin, nrad)
if par.fargo3d == 'Yes':
Stokes_fargo3d = np.zeros((par.nbin, par.gas.nrad))
axirhogas = np.sum(par.gas.data,
axis=1) / par.gas.nsec # in code units
axirhogas *= (par.gas.cumass * 1e3) / (
(par.gas.culength * 1e2)**2.) # in g/cm^2
for ibin in range(par.nbin):
if par.dustfluids != 'No':
mysize = par.dust_size[par.dustfluids[0] - 1 + ibin]
else:
mysize = par.bins[ibin]
print('mysize = ', mysize)
Stokes_fargo3d[ibin, :] = 0.5 * np.pi * (
mysize * 1e2
) * par.dust_internal_density / axirhogas # since dust size is in meters...
#print('max Stokes number = ', Stokes_fargo3d[ibin,:].max())
# For the vertical expansion of the dust mass volume density, we
# need to define a 2D array for the dust's aspect ratio:
for ibin in range(par.nbin):
if par.fargo3d == 'No':
St = avgstokes[ibin] # avg stokes number for that bin
if par.fargo3d == 'Yes' and par.dustfluids != 'No':
St = Stokes_fargo3d[ibin]
# gas aspect ratio (par.gas.rmed[i] = R in code units)
hgas = par.aspectratio * (par.gas.rmed)**(par.flaringindex)
# vertical extension depends on grain Stokes number:
# T = theoretical: hd/hgas = sqrt(alpha/(St+alpha))
# T2 = theoretical: hd/hgas = sqrt(Dz/(St+Dz)) with Dz = 10xalpha here is the coefficient for
# vertical diffusion at midplane, which can differ from alpha
# F = extrapolation from the simulations by Fromang & Nelson 09
# G = Gaussian = same as gas (case of well-coupled dust for polarized intensity images)
if par.z_expansion == 'F':
hd[ibin, :] = 0.7 * hgas * ((St + 1. / St) / 1000.)**(0.2)
if par.z_expansion == 'T':
hd[ibin, :] = hgas * np.sqrt(par.alphaviscosity /
(par.alphaviscosity + St))
if par.z_expansion == 'T2':
hd[ibin, :] = hgas * np.sqrt(10.0 * par.alphaviscosity /
(10.0 * par.alphaviscosity + St))
if par.z_expansion == 'G':
hd[ibin, :] = hgas
# dust aspect ratio as function of ibin, r and phi (2D array for each size bin)
hd2D = np.zeros((par.nbin, par.gas.nrad, par.gas.nsec))
for th in range(par.gas.nsec):
hd2D[:, :, th] = hd # nbin, nrad, nsec
# grid radius function of ibin, r and phi (2D array for each size bin)
r2D = np.zeros((par.nbin, par.gas.nrad, par.gas.nsec))
for ibin in range(par.nbin):
for th in range(par.gas.nsec):
r2D[ibin, :, th] = par.gas.rmed
# work out exponential and normalization factors exp(-z^2 / 2H_d^2)
# with z = r cos(theta) and H_d = h_d x R = h_d x r sin(theta)
# r = spherical radius, R = cylindrical radius
for j in range(par.gas.ncol):
rhodustcube[j, :, :, :] = dustcube * np.exp(
-0.5 *
(np.cos(par.gas.tmed[j]) / hd2D)**2.0) # ncol, nbin, nrad, nsec
rhodustcube[j, :, :, :] /= (np.sqrt(2. * np.pi) * r2D * hd2D *
par.gas.culength * 1e2
) # quantity is now in g / cm^3
# Renormalize dust's mass volume density such that the sum over the 3D grid's volume of
# the dust's mass volume density x the volume of each grid cell does give us the right
# total dust mass, which equals ratio x Mgas. Do that every time except fargo3D simulations
# carried out in 2D used dust fluids
if par.hydro2D == 'Yes' and par.dustfluids == 'No':
rhofield = np.sum(rhodustcube, axis=1) # sum over dust bins
Redge, Cedge, Aedge = np.meshgrid(
par.gas.redge, par.gas.tedge,
par.gas.pedge) # ncol+1, nrad+1, Nsec+1
r2 = Redge * Redge
jacob = r2[:-1, :-1, :-1] * np.sin(Cedge[:-1, :-1, :-1])
dphi = Aedge[:-1, :-1, 1:] - Aedge[:-1, :-1, :-1] # same as 2pi/nsec
dr = Redge[:-1, 1:, :-1] - Redge[:-1, :-1, :-1] # same as Rsup-Rinf
dtheta = Cedge[1:, :-1, :-1] - Cedge[:-1, :-1, :-1]
# volume of a cell in cm^3
vol = jacob * dr * dphi * dtheta * (
(par.gas.culength * 1e2)**3) # ncol, nrad, Nsec
total_mass = np.sum(rhofield * vol)
normalization_factor = par.ratio * Mgas * (par.gas.cumass *
1e3) / total_mass
rhodustcube = rhodustcube * normalization_factor
if par.verbose == 'Yes':
print('total dust mass after vertical expansion [g] = ',
np.sum(np.sum(rhodustcube, axis=1) * vol),
' as normalization factor = ', normalization_factor)
# finally write mass volume densities for all size bins
for ibin in range(par.nbin):
print('dust species in bin', ibin, 'out of ', par.nbin - 1)
for k in range(par.gas.nsec):
for j in range(par.gas.ncol):
for i in range(par.gas.nrad):
DUSTOUT.write(str(rhodustcube[j, ibin, i, k]) + ' \n')
# print max of dust's mass volume density at each colatitude
if par.verbose == 'Yes':
for j in range(par.gas.ncol):
print('max(rho_dustcube) [g cm-3] for colatitude index j = ', j,
' = ', rhodustcube[j, :, :, :].max())
DUSTOUT.close()
# plot azimuthally-averaged dust density vs. radius and colatitude for smallest and largest bin sizes
if par.plot_dust_quantities == 'Yes':
from mpl_toolkits.axes_grid1 import make_axes_locatable
import matplotlib.ticker as ticker
from matplotlib.ticker import (MultipleLocator, FormatStrFormatter,
AutoMinorLocator, LogLocator,
LogFormatter)
matplotlib.rcParams.update({'font.size': 20})
matplotlib.rc('font', family='Arial')
fontcolor = 'white'
# azimuthally-averaged dust density:
axidens_smallest = np.sum(rhodustcube[:, 0, :, :],
axis=2) / par.gas.nsec # (nol,nrad)
axidens_largest = np.sum(rhodustcube[:, par.nbin - 1, :, :],
axis=2) / par.gas.nsec # (nol,nrad)
radius_matrix, theta_matrix = np.meshgrid(par.gas.redge, par.gas.tedge)
R = radius_matrix * np.sin(
theta_matrix) * par.gas.culength / 1.5e11 # in au
Z = radius_matrix * np.cos(
theta_matrix) * par.gas.culength / 1.5e11 # in au
# midplane dust mass volume density:
midplane_dens_smallest = rhodustcube[par.gas.ncol // 2 - 1,
0, :, :] # (nrad,nsec)
midplane_dens_largest = rhodustcube[par.gas.ncol // 2 - 1,
par.nbin - 1, :, :] # (nrad,nsec)
radius_matrix, theta_matrix = np.meshgrid(par.gas.redge, par.gas.pedge)
X = radius_matrix * np.cos(
theta_matrix) * par.gas.culength / 1.5e11 # in au
Y = radius_matrix * np.sin(
theta_matrix) * par.gas.culength / 1.5e11 # in au
print(
'--------- a) plotting azimuthally-averaged dust density (R,z) ----------'
)
# --- smallest bin size ---
fig = plt.figure(figsize=(8., 8.))
plt.subplots_adjust(left=0.17, right=0.92, top=0.88, bottom=0.1)
ax = plt.gca()
ax.tick_params(top='on',
right='on',
length=5,
width=1.0,
direction='out')
ax.tick_params(axis='x', which='minor', top=True)
ax.tick_params(axis='y', which='minor', right=True)
ax.set_xlabel('Radius [au]')
ax.set_ylabel('Altitude [au]')
ax.set_ylim(Z.min(), Z.max())
ax.set_xlim(R.min(), R.max())
if axidens_smallest.max() / axidens_smallest.min() > 1e3:
mynorm = matplotlib.colors.LogNorm(vmin=1e-3 *
axidens_smallest.max(),
vmax=axidens_smallest.max())
else:
mynorm = matplotlib.colors.LogNorm(vmin=axidens_smallest.min(),
vmax=axidens_smallest.max())
CF = ax.pcolormesh(R,
Z,
axidens_smallest,
cmap='nipy_spectral',
norm=mynorm)
divider = make_axes_locatable(ax)
cax = divider.append_axes("top", size="2.5%", pad=0.12)
cb = plt.colorbar(CF, cax=cax, orientation='horizontal')
cax.xaxis.tick_top()
cax.xaxis.set_tick_params(labelsize=20, direction='out')
cax.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=10))
cax.xaxis.set_label_position('top')
cax.set_xlabel('dust density ' + r'[g cm$^{-3}$]')
cax.xaxis.labelpad = 8
fileout = 'dust_density_smallest_Rz.pdf'
plt.savefig('./' + fileout, dpi=160)
plt.close(fig)
# --- repeat for largest bin size ---
fig = plt.figure(figsize=(8., 8.))
plt.subplots_adjust(left=0.17, right=0.92, top=0.88, bottom=0.1)
ax = plt.gca()
ax.tick_params(top='on',
right='on',
length=5,
width=1.0,
direction='out')
ax.tick_params(axis='x', which='minor', top=True)
ax.tick_params(axis='y', which='minor', right=True)
ax.set_xlabel('Radius [au]')
ax.set_ylabel('Altitude [au]')
ax.set_ylim(Z.min(), Z.max())
ax.set_xlim(R.min(), R.max())
if axidens_largest.max() / axidens_largest.min() > 1e3:
mynorm = matplotlib.colors.LogNorm(vmin=1e-3 *
axidens_largest.max(),
vmax=axidens_largest.max())
else:
mynorm = matplotlib.colors.LogNorm(vmin=axidens_largest.min(),
vmax=axidens_largest.max())
CF = ax.pcolormesh(R,
Z,
axidens_largest,
cmap='nipy_spectral',
norm=mynorm)
divider = make_axes_locatable(ax)
cax = divider.append_axes("top", size="2.5%", pad=0.12)
cb = plt.colorbar(CF, cax=cax, orientation='horizontal')
cax.xaxis.tick_top()
cax.xaxis.set_tick_params(labelsize=20, direction='out')
cax.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=10))
cax.xaxis.set_label_position('top')
cax.set_xlabel('dust density ' + r'[g cm$^{-3}$]')
cax.xaxis.labelpad = 8
fileout = 'dust_density_largest_Rz.pdf'
plt.savefig('./' + fileout, dpi=160)
plt.close(fig)
print('--------- b) plotting dust density (x,y) ----------')
# --- smallest bin size ---
fig = plt.figure(figsize=(8., 8.))
plt.subplots_adjust(left=0.17, right=0.92, top=0.88, bottom=0.1)
ax = plt.gca()
ax.tick_params(top='on',
right='on',
length=5,
width=1.0,
direction='out')
ax.tick_params(axis='x', which='minor', top=True)
ax.tick_params(axis='y', which='minor', right=True)
ax.set_xlabel('x [au]')
ax.set_ylabel('y [au]')
ax.set_ylim(Y.min(), Y.max())
ax.set_xlim(X.min(), X.max())
if midplane_dens_smallest.max() / midplane_dens_smallest.min() > 1e3:
mynorm = matplotlib.colors.LogNorm(
vmin=1e-3 * midplane_dens_smallest.max(),
vmax=midplane_dens_smallest.max())
else:
mynorm = matplotlib.colors.LogNorm(
vmin=midplane_dens_smallest.min(),
vmax=midplane_dens_smallest.max())
midplane_dens_smallest = np.transpose(midplane_dens_smallest)
CF = ax.pcolormesh(X,
Y,
midplane_dens_smallest,
cmap='nipy_spectral',
norm=mynorm,
rasterized=True)
#CF = ax.pcolormesh(X,Y,midplane_dens_smallest,cmap='nipy_spectral',norm=mynorm)
divider = make_axes_locatable(ax)
cax = divider.append_axes("top", size="2.5%", pad=0.12)
cb = plt.colorbar(CF, cax=cax, orientation='horizontal')
cax.xaxis.tick_top()
cax.xaxis.set_tick_params(labelsize=20, direction='out')
cax.xaxis.set_label_position('top')
cax.set_xlabel('midplane dust density ' + r'[g cm$^{-3}$]')
cax.xaxis.labelpad = 8
fileout = 'dust_density_smallest_midplane.pdf'
plt.savefig('./' + fileout, dpi=160)
plt.close(fig)
# --- repeat for largest bin size ---
fig = plt.figure(figsize=(8., 8.))
plt.subplots_adjust(left=0.17, right=0.92, top=0.88, bottom=0.1)
ax = plt.gca()
ax.tick_params(top='on',
right='on',
length=5,
width=1.0,
direction='out')
ax.tick_params(axis='x', which='minor', top=True)
ax.tick_params(axis='y', which='minor', right=True)
ax.set_xlabel('x [au]')
ax.set_ylabel('y [au]')
ax.set_ylim(Y.min(), Y.max())
ax.set_xlim(X.min(), X.max())
if midplane_dens_largest.max() / midplane_dens_largest.min() > 1e3:
mynorm = matplotlib.colors.LogNorm(
vmin=1e-3 * midplane_dens_largest.max(),
vmax=midplane_dens_largest.max())
else:
mynorm = matplotlib.colors.LogNorm(
vmin=midplane_dens_largest.min(),
vmax=midplane_dens_largest.max())
midplane_dens_largest = np.transpose(midplane_dens_largest)
CF = ax.pcolormesh(X,
Y,
midplane_dens_largest,
cmap='nipy_spectral',
norm=mynorm,
rasterized=True)
#CF = ax.pcolormesh(X,Y,midplane_dens_largest,cmap='nipy_spectral',norm=mynorm)
divider = make_axes_locatable(ax)
cax = divider.append_axes("top", size="2.5%", pad=0.12)
cb = plt.colorbar(CF, cax=cax, orientation='horizontal')
cax.xaxis.tick_top()
cax.xaxis.set_tick_params(labelsize=20, direction='out')
cax.xaxis.set_label_position('top')
cax.set_xlabel('midplane dust density ' + r'[g cm$^{-3}$]')
cax.xaxis.labelpad = 8
fileout = 'dust_density_largest_midplane.pdf'
plt.savefig('./' + fileout, dpi=160)
plt.close(fig)
# free RAM memory
del rhodustcube, dustcube, hd2D, r2D
def initial_enthalpy_plot(loading,
enthalpy,
fitted_enthalpy,
log=False,
title=None,
extras=None,
ax=None):
"""
Draws the initial enthalpy calculation plot.
Parameters
----------
loading : array
Loadings for which the initial enthalpy was calculated.
enthalpy : array
The enthalpy corresponding to each loading.
fitted_enthalpy : array
The predicted enthalpy corresponding to each loading.
log : int
Whether to display a logarithmic graph
title : str
Name of the material to put in the title.
ax : matplotlib axes object, default None
The axes object where to plot the graph if a new figure is
not desired.
Returns
-------
matplotlib.axes
Matplotlib axes of the graph generated. The user can then apply their
own styling if desired.
"""
# Generate the figure if needed
if ax is None:
_, ax = plt.subplots()
ax.plot(loading,
enthalpy,
marker='o',
color='black',
label='original',
linestyle='')
ax.plot(loading, fitted_enthalpy, color='r', label='fitted', linestyle='-')
if extras is not None:
for param in extras:
ax.plot(param[0], param[1], label=param[2], linestyle='--')
if log:
ax.set_xscale('log')
ax.xaxis.set_major_locator(
ticker.LogLocator(base=10.0, numticks=15, numdecs=20))
ax.set_title(title + " initial enthalpy fit")
ax.set_xlabel('Loading')
ax.set_ylabel('Enthalpy')
ax.legend(loc='best')
ax.set_ylim(bottom=0, top=(max(enthalpy) * 1.2))
ax.set_xlim(left=0)
ax.grid(True)
return ax
def bracket_descent_test(xg, display=False, compare=False, i=1):
"""
======================================================================================
Use the Bracket Descent method to minimize a cost function, j, defined in cost module.
======================================================================================
Parameters
----------
xg : list
Initial guess
display : Boolean, Optional
If set to True, figures will be created to illustrate the optimization
path taken and the distance from convergence at each step.
compare : Boolean, optional
If set to True, a figure will be created to directly compare Newton and
Bracket Descent methods.
i=1 : Integer, Optional
Sets the name of the figures as hw231(/2)_i.png.
Returns
---------
xf : ndarray
Computed location of minimum
jf : float
Computed minimum
output : Tuple
containing the time taken for the minimia to be found for each of newton and
bracket descent methods. An average over 10 tests is taken, only set if
compare parameter set to True, otherwise empty.
Calling this function will produce two figures. The first will containing two
subplots illustrating the location of each step in the minimization path, overlayed
over the initial cost function, and the distance of j from the final, computed
minimum at each iteration.
The second plot (which is only produced when 'compare' is set to True) demonstrates
the distance of each step from the final, converged minimum at each iteration.
This shows that the newton method requires significantly fewer steps and is hence
faster.
Trends Observed
----------------
Figures hw321_i show the path taken during a bracket descent conversion is much
longer than in a newton conversion (shown in figures hw22i). This is because
the B-D method limits the size of a step to 2*L where L is definied by the size
of an equilateral triangle whose centroid moved with each step. The method is
furthermore designed such that this triangle will only decrease in size per
iteration, and hence the maximum length a step can take can only be
decreased (not increased) throughout the convergence. The figures further show
that steps appear to be taken initially perpendicular to the curvature, finding
the minimum along that strip, and then converging in down the parallel Path
until they reach a level of tolerance.
In contrast, the Newton approach is not limited in the size of the steps it is
able to take and can hence converge in a much smaller number of iterations.
This is a result of the use of gradients in this method. Figures hw22i illustrate
how each step travels through many bands on the contour plot (representing
differences of 1 order of magnitude each) as the method searches for the
direction of minimisation.
"""
cost.c_noise = False
hw2.tol = 10**(-6)
hw2.itermax = 1000
t34 = 0
output = ()
if compare:
N = 10
else:
N = 1
for j in range(1, N):
t3 = time()
hw2.bracket_descent(xg)
t4 = time()
t34 = t34 + (t4 - t3)
X, Y = hw2.xpath
xf = [X[-1], Y[-1]]
jf = hw2.jpath[-1]
d1 = np.sqrt((X - xf[0])**2 + (Y - xf[1])**2)
if display:
Minx = min(X) - 1
Maxx = max(X) + 1
Miny = min(Y) - 1
Maxy = max(Y) + 1
[Xj, Yj] = np.linspace(Minx, Maxx, 200), np.linspace(Miny, Maxy, 200)
#calculate noiseless cost function at each point on 2D grid
j = [[cost.costj([xi, yi]) for xi in Xj] for yi in Yj]
f, (p1, p2) = plt.subplots(1, 2)
p1.contourf(Xj, Yj, j, locator=ticker.LogLocator(), cmap=cm.GnBu)
p1.plot(X, Y, 'g', marker='d')
p1.set_xlabel('X1-location')
p1.set_ylabel('X2-location')
p1.set_title('Convergence Path')
p2.semilogy(np.linspace(1, len(X), len(X)), hw2.jpath)
p2.set_xlabel('Iteration number')
p2.set_ylabel('distance from converged minimum')
p2.set_title('Rate')
plt.suptitle('Rosemary Teague, bracket_descent_test, initial guess =' +
str(xg) + ' \n Rate of convergence of a cost function')
plt.tight_layout(pad=4)
plt.savefig('hw231_' + str(i), dpi=700)
if compare:
plt.close('all')
One, = plt.loglog(np.linspace(1, len(X), len(X)), hw2.jpath)
xf2, jf2, outputn = newton_test(xg, timing=True)
X2, Y2 = outputn[1], outputn[2]
d2 = np.sqrt((X2 - xf2[0])**2 + (Y2 - xf2[1])**2)
print(np.linspace(1, len(X2), len(X2)), outputn[3])
Two, = plt.loglog(np.linspace(1, len(X2), len(X2)), outputn[3])
One.set_label('Bracket Descent')
Two.set_label('Newton')
plt.xlabel('Iteration number')
plt.ylabel('Distance from converged minimum')
plt.legend()
plt.title('Rosemary Teague, bracket_descent_test, initial guess =' +
str(xg) +
' \n Comparison of Newton and Bracket Descent Methods')
plt.savefig('hw232_' + str(i), dpi=700)
output = (outputn[0], t34 / N)
return xf, jf, output
def plot_data(df,
title='',
min=None,
max=None,
ylog=False,
stack=False,
subplot=111,
fig=None,
yticks=None,
gf=False,
timebase=28,
c_mark=False):
# Create new fig
if fig is None:
fig = plt.figure()
# Add to existing fig
ax = fig.add_subplot(subplot)
# Create x and y data for plotting
x = df.index.to_numpy()
x_t = df.index.to_numpy(dtype='float64')
y = df.to_numpy()
# If the user wants a graph of Growth Rate, do some spooky mathemagic....
# This part needs better commenting!
if gf:
x = x[2:]
y = y[2:]
x_t = x_t[2:]
y = np.diff(y, axis=0)
y = np.insert(y, 0, y[0], axis=0)
y = np.insert(y, -1, y[-1], axis=0)
y = convolve2d(y, [[0.25], [0.5], [0.25]], mode='valid')
y = np.diff(np.log(y + 1e-3), axis=0)
y = np.insert(y, 0, y[0], axis=0)
y = np.insert(y, -1, y[-1], axis=0)
y = convolve2d(y, [[0.25], [0.5], [0.25]], mode='valid')
x = x[2:]
x_t = x_t[2:]
#print(x.shape, y.shape)
if stack:
ax.stackplot(x, y.transpose(), labels=df.columns)
else:
# Mark data point every 7 days
ax.plot(x, y, markevery=7)
# Set axes, grid and ticks
if min is not None and max is not None:
ax.set_ylim([min, max])
if ylog:
ax.set_yscale('log')
locmin = plticker.LogLocator(base=10.0, subs=(1, 10))
ax.yaxis.set_major_formatter(plticker.FormatStrFormatter("%.0f"))
ax.yaxis.set_major_locator(locmin)
ax.grid(axis='y', which='major', linewidth=0.8)
ax.grid(axis='y', which='minor', linewidth=0.3)
ax.grid(axis='x', which='both')
ax.xaxis.set_major_locator(plticker.MultipleLocator(base=timebase))
ax.xaxis.set_minor_locator(plticker.MultipleLocator(base=timebase / 2))
if stack:
ax.legend(loc='upper left')
else:
for i, line in enumerate(ax.get_lines()):
line.set_marker(MARKERS[i])
#line.get_color()
if gf:
# calc the trendline
z = np.polyfit(x_t[4:], y[4:, i], 1)
p = np.poly1d(z)
ax.plot(x, p(x_t), "--", color=line.get_color())
ax.legend(ax.get_lines(), df.columns, loc='upper left')
# Set graph title
ax.set_title(title)
ax.yaxis.tick_right()
# place a text box in upper left in axes coords
if c_mark:
ax.text(0.99,
0.05,
"(c) E. Maitland 2020",
transform=ax.transAxes,
fontsize=8.5,
va='top',
ha='right') # , bbox=props)
return ax
def newton_test(xg, display=False, i=1, timing=False):
"""
============================================================================
Use Newton's method to minimize a cost function, j, defined in cost module.
============================================================================
Parameters
----------
xg : list
Initial guess
display : Boolean, Optional
If set to True, figures will be created to illustrate the optimization
path taken and the distance from convergence at each step.
i=1 : Integer, Optional
Sets the name of the figures as hw22i.png
timing : Boolean, Optional
If set to true, an average time will be calculated for the completion
of finding a minimum and will be appended to the tuple output.
Returns
---------
xf : ndarray
Computed location of minimum
jf : float
Computed minimum
output : Tuple
containing the time taken for the minimia to be found. An
average over 10 tests, only set if timining parameter set to True, otherwise
empty.
Calling this function will produce a figure containing two subplots. The first
will illustrate the location of each step in the minimization path, overlayed
over the initial cost function. The second will illustrate the distance from
the final, computed minimum at each iteration.
"""
cost.c_noise = False
hw2.tol = 10**(-6)
hw2.itermax = 1000
t21 = 0
output = ()
if timing:
N = 10
else:
N = 1
for j in range(1, N):
t1 = time()
hw2.newton(xg)
t2 = time()
t21 = t21 + (t2 - t1)
X, Y = hw2.xpath
xf = [X[-1], Y[-1]]
jpathn = [j for j in hw2.jpath]
jf = hw2.jpath[-1]
output = (t21 / N, X, Y, jpathn)
if display:
Minx = min(X) - 1
Maxx = max(X) + 1
Miny = min(Y) - 1
Maxy = max(Y) + 1
[Xj, Yj] = np.linspace(Minx, Maxx, 200), np.linspace(Miny, Maxy, 200)
#calculate noiseless cost function at each point on 2D grid
j = [[cost.costj([xi, yi]) for xi in Xj] for yi in Yj]
f, (p1, p2) = plt.subplots(1, 2)
p1.contourf(Xj, Yj, j, locator=ticker.LogLocator(), cmap=cm.GnBu)
p1.plot(X, Y, 'g', marker='d')
p1.set_xlim(min(X) - 1, max(X) + 1)
p1.set_xlabel('X1-location')
p1.set_ylabel('X2-location')
p1.set_title('Convergence Path')
p2.plot(np.linspace(0, len(X) - 1, len(X)), hw2.jpath - jf)
p2.set_xlabel('Iteration number')
p2.set_ylabel('distance from converged minimum')
p2.set_title('Rate')
plt.suptitle('Rosemary Teague, Newton_test, initial guess =' +
str(xg) + ' \n Convergence of a cost function')
plt.tight_layout(pad=4)
plt.savefig('hw22' + str(i), dpi=700)
return xf, jf, output
def log_minor_ticks(ax): locmin = ticker.LogLocator(base=10.0, subs=(0.1,0.2,0.4,0.6,0.8,1,2,4,6,8,10)) ax.yaxis.set_minor_locator(locmin) ax.yaxis.set_minor_formatter(ticker.NullFormatter())
setup(ax)
ax.plot(range(0, 5), [0]*5, color='White')
ax.xaxis.set_major_locator(ticker.IndexLocator(base=.5, offset=.25))
ax.text(0.0, 0.5, "IndexLocator(base=0.5, offset=0.25)",
fontsize=14, transform=ax.transAxes)
# Auto Locator
ax = plt.subplot(n, 1, 6)
setup(ax)
ax.xaxis.set_major_locator(ticker.AutoLocator())
ax.xaxis.set_minor_locator(ticker.AutoMinorLocator())
ax.text(0.0, 0.5, "AutoLocator()", fontsize=14, transform=ax.transAxes)
# MaxN Locator
ax = plt.subplot(n, 1, 7)
setup(ax)
ax.xaxis.set_major_locator(ticker.MaxNLocator(4))
ax.xaxis.set_minor_locator(ticker.MaxNLocator(40))
ax.text(0.0, 0.5, "MaxNLocator(n=4)", fontsize=14, transform=ax.transAxes)
# Log Locator
ax = plt.subplot(n, 1, 8)
setup(ax)
ax.set_xlim(10**3, 10**10)
ax.set_xscale('log')
ax.xaxis.set_major_locator(ticker.LogLocator(base=10.0, numticks=15))
ax.text(0.0, 0.5, "LogLocator(base=10, numticks=15)",
fontsize=15, transform=ax.transAxes)
plt.show()
#b=parsec
for j in range(N_M):
n_p = rho/M_p_bins[j]
for i in range(N_a):
b = calc_b_max(M_p_bins[j], v_rms, a_bins[i], m1, m2)
for k in range(N_enc):
(notBound, a_new, e_new) = impulseEncounter(m1, m2, v_rms, b, a_bins[i], e, M_p_bins[j])
a_frac_avg[i,j] += (a_new-a_bins[i])/a_bins[i]
#Normalise a_frac_avg
a_frac_avg /= N_enc
#Plot
plt.title(r'Absolute average fractional change in semi-major axis due to single encounter at $b=b_{\mathrm{max}}$')
ax = plt.gca()
cs = ax.contourf(a_bins/au, M_p_bins/(2.0*10.0**30.0), np.transpose(np.absolute(a_frac_avg)), locator=ticker.LogLocator())
plt.colorbar(cs)
plt.ylabel(r'Perturber mass, $M_\odot$')
plt.xlabel('Initial semi-major axis, au')
plt.xscale('log')
plt.yscale('log')
plt.show()
def plot_smps(d_mtx,
fit_lognormal_modes=False,
fig=None,
ax=None,
zlim=[10**1, 10**6],
xlim=None,
logscale_z=True,
title='Size distribution',
xlabel='Timestamp',
ylabel='Mobility diameter (nm)',
zlabel='dN/d(log$_{10}$d$_0$) ($cm^{-3}$)',
saveorshowplot='show',
output_path=None,
output_filename='SMPS.png'):
#https://matplotlib.org/examples/images_contours_and_fields/pcolormesh_levels.html
d_mtx = _fill_smps_times(d_mtx)
# Setup data input
x0 = np.array([dates.date2num(d) for d in d_mtx.index]) # Time axis
try: #try tsi first
y0 = np.array([float(s) for s in d_mtx.columns.values]) # size axis
except: #otherwise its grimm!
y0 = np.array([float(s.split(' ')[0]) for s in d_mtx.columns.values])
x, y = np.meshgrid(x0, y0)
z = d_mtx.as_matrix().transpose()
with np.errstate(
invalid='ignore'): # ignore runtimewarning about nan values
z[z < zlim[0]] = zlim[
0] # mask bad values so that there are no holes in the data
# Plot contour
cmap = plt.get_cmap('jet')
if logscale_z:
tick_loc = tck.LogLocator()
cont_levels = np.logspace(np.log10(zlim[0]), np.log10(zlim[1]), 100)
z_loc_arr = np.logspace(np.log10(zlim[0]), np.log10(zlim[1]), 6)
else:
tick_loc = tck.LinearLocator()
cont_levels = np.linspace(zlim[0], zlim[1], 100)
z_loc_arr = np.array(
[round_to_1(x) for x in np.linspace(zlim[0], zlim[1], 6)])
if (fig is None) and (ax is None):
fig, ax = plt.subplots(nrows=1, figsize=(15, 5))
elif (fig is None) or (ax is None):
assert True, \
'You must pass both a figure and axis object in to the \
size distribution plotting routine'
cf = ax.contourf(x,
np.log(y),
z,
levels=cont_levels,
locator=tick_loc,
cmap=cmap,
interpolation=None)
ax.set_title(title)
# Format y axis labels
y_loc_arr = np.logspace(np.log10(y0[0]), np.log10(y0[-1]), 7)
rounding = [round_to_1(y) for y in y_loc_arr[1:-1]]
y_loc_arr = np.append(np.insert(rounding, 0, np.ceil(y_loc_arr[0])),
np.floor(y_loc_arr[-1]))
y_loc = tck.FixedLocator(np.log(y_loc_arr))
y_loc_min = tck.FixedLocator(
np.log(
np.concatenate((np.arange(10, 100, 10), np.arange(100, 700,
100)))))
ax.yaxis.set_major_locator(y_loc)
ax.set_yticklabels(y_loc_arr)
ax.yaxis.set_minor_locator(y_loc_min)
ax.set_ylabel(ylabel)
# Format x axis labels
if xlim is not None:
ax.set_xlim(xlim)
x_loc_arr = get_time_ticks(x0, 6)
x_loc = tck.FixedLocator(x_loc_arr)
x_labels = np.array(
[dates.num2date(dt).strftime('%d-%b\n%H:%M') for dt in x_loc_arr])
ax.xaxis.set_major_locator(x_loc)
ax.set_xticklabels(x_labels)
ax.set_xlabel(xlabel)
# Format colorbar
cbar = fig.colorbar(cf, ax=ax, ticks=z_loc_arr, pad=0.01)
cbar.ax.set_xticklabels(z_loc_arr)
cbar.set_label(zlabel)
if fit_lognormal_modes:
# Calculate the maximum in each mode using a lognormal fitting procedure
mode_max = mode_max_from_dist_fit(d_mtx)
# Overlay the mode sizes
ax.plot(np.log(mode_max), '-k')
atmosplots.saveorshowplot(plt, saveorshowplot, output_path,
output_filename)
return
# 生成网格数据 X, Y = np.meshgrid(x_value, y_value) #显示3D图像 fig = plt.figure() ax = plt.gca(projection='3d') #ax.plot(X,Y,f(X,Y)) ax.plot_surface(X,Y,f(X,Y),cmap='jet') #plt.show() #显示等高线 plt.figure() # 填充等高线的颜色, 8是等高线分为几部分 plt.contourf(X, Y, f(X, Y), 5, alpha=0) # 绘制等高线 C = plt.contour(X, Y, f(X, Y), 8, locator=ticker.LogLocator(), colors='black', linewidth=0.01) # 绘制等高线数据 plt.clabel(C, inline=True, fontsize=10) alpha = 0.09 beta = 0.8 x = np.array([[-0.2],[0.4]]) #随机选取一个起始点 eta = 0.000001 #学习率 #print(df(x)) xv = [x[0,0]] yv = [x[1,0]] plt.plot(x[0, 0], x[1, 0], marker='o') #plt.show()
win_levels = np.array([0, args.boundary])
all_levels = np.array([0, 1, 10, 100, 1000])
norm = cm.colors.Normalize(vmax=abs(z).max(), vmin=-abs(z).max())
cmap = cm.PRGn
fig, ax = plt.subplots()
ax.set_yscale("log")
ax.set_xlabel('Lag in arrival time (hrs)')
ax.set_ylabel('Challenger inoculum (CFU)')
cset1 = plt.contourf(x,
y,
z,
win_levels,
locator=ticker.LogLocator(),
colors=('#289600', '#ffffff'))
#cmap=cm.get_cmap(cmap, len(levels) - 1), norm=norm)
# t_com
plt.axvline(x=3.76, color='k', linestyle='--', label='t_com')
plt.text(2.7, 3000, 't_com', rotation=90)
#cset2 = plt.contour(x, y, z, all_levels, locator=ticker.LogLocator(), colors='k', linewidths = 1)
#plt.clabel(cset2, fmt='%1.0f', inline=1, fontsize=10)
# Resident wins boundary
cset3 = plt.contour(x, y, z, win_levels, colors='k', linewidths=2)
plt.title('Isogenic challenger')
#plt.colorbar(cset1) # legend
plt.savefig(args.output + ".pdf")
def pps_eigenmodes_eval(self):
## LOAD ComputeSaveNormalizedEigenError: fig1
fig1_data = np.load(self.eval_dir + '/ComputeSaveNormalizedEigenError_fig1.npz')
normalized_relative_error = fig1_data['nre']
true_tsnap = fig1_data['tt']
linearEvolvingEigen = fig1_data['le']
relative_error = self.params['relative_error']
## draw ComputeSaveNormalizedEigenError: fig1
# plot normalized relative error for each eigenmodes
plt.figure(figsize=FIG_SIZE)
cmap = self.get_cmap(normalized_relative_error.shape[1])
for i in range(normalized_relative_error.shape[1]):
plt.plot(true_tsnap, normalized_relative_error[:, i], '-', c=cmap(i), label= str(i + 1) + 'th-eigenvalue: ' + "{0:.3f}".format(linearEvolvingEigen[i,i]))
plt.xlabel('time',fontsize = 32)
if relative_error:
plt.ylabel('normalized error',fontsize = 32)
else:
plt.ylabel('error',fontsize = 32)
plt.yscale('log')
lgd = plt.legend(bbox_to_anchor=(1, 0.5))
plt.savefig(self.pps_dir + '/normalized_relative_eigen_error.png',
bbox_extra_artists=(lgd,),
bbox_inches='tight')
plt.close()
## LOAD ComputeSaveNormalizedEigenError: fig2
fig2_data = np.load(self.eval_dir + '/ComputeSaveNormalizedEigenError_fig2.npz')
mean_normalized_relative_error = fig2_data['mre']
small_to_large_error_eigen_index = fig2_data['stli']
small_to_large_error_eigen_index_kou = fig2_data['stli_kou']
abs_sum_kou = fig2_data['abs_sum_kou']
error_reconstruct_state_list = fig2_data['ersl']
# bool_index_further = fig2_data['iestli'] ## it is removed..
self.small_to_large_error_eigen_index = small_to_large_error_eigen_index
# self.bool_index_further = bool_index_further
## draw ComputeSaveNormalizedEigenError: fig2
# error ordered
fig= plt.figure(figsize=FIG_SIZE)
ax1 = fig.add_subplot(111)
ax2 = ax1.twinx()
ax1.plot(range(1, normalized_relative_error.shape[1] + 1),
mean_normalized_relative_error[small_to_large_error_eigen_index],
'b-^', label='max relative eigenfunction error')
ax1.set_xlabel(r'number of selected eigenmodes $\hat{L}$',fontsize = 32)
# ax1.legend(bbox_to_anchor=(1, 0.5))
ax1.set_yscale('log')
if relative_error:
ax1.set_ylabel('max linear evolving normalized error', color='b',fontsize = 32)
else:
ax1.set_ylabel('max error', color='b',fontsize = 32)
# plot error from reconstruction state from eigenfunction values
ax2.plot(range(1, normalized_relative_error.shape[1] + 1),
error_reconstruct_state_list,'r-o',
label='reconstruction normalized error')
if relative_error:
ax2.set_ylabel('reconstruction normalized error', color='r',fontsize = 32)
else:
ax2.set_ylabel('reconstruction error', color='r',fontsize = 32)
# ax2.set_ylim([-1,20])
ax2.set_yscale('log')
# set up ticks
yticks = ticker.LogLocator()
ax1.yaxis.set_major_locator(yticks)
ax2.yaxis.set_major_locator(yticks)
# ax2.legend(bbox_to_anchor=(1, 0.5))
plt.savefig(self.pps_dir + '/reConstr_decay_normalized_relative_eigen_error.png',
bbox_inches='tight')
plt.close()
## LOAD ComputeSaveNormalizedEigenError: fig3
fig3_data = np.load(self.eval_dir + '/ComputeSaveNormalizedEigenError_fig3.npz')
top_k_modes_list = fig3_data['tkm_index_list']
self.top_k_modes_list = top_k_modes_list
# print out kou's result
print('as a comparison: index chosen by Kou criterion: ')
print(small_to_large_error_eigen_index_kou + 1)
print('corresponding abs sum:')
print(abs_sum_kou)
self.index_selected_in_full = self.small_to_large_error_eigen_index[:self.top_k_modes_list[-1] + 1]
# self.index_selected_in_full = self.small_to_large_error_eigen_index
## draw ComputeSaveNormalizedEigenError: fig3
# fig. 3: plot normalized relative error for top K smallest error eigenmodes
plt.figure(figsize=FIG_SIZE)
cmap = self.get_cmap(len(top_k_modes_list))
for i in top_k_modes_list:
i_s = small_to_large_error_eigen_index[i]
plt.plot(true_tsnap, normalized_relative_error[:, i_s], '-', c=cmap(i),
label= str(i_s + 1) + 'th-eigenvalue: ' + "{0:.3f}".format(linearEvolvingEigen[i_s,i_s]))
# print eigenvectors
# print 'no. eigen vectors ', i_s+1
# print self.model.KoopmanEigenV[:, i_s]
plt.xlabel('time',fontsize = 32)
if relative_error:
plt.ylabel('normalized error',fontsize = 32)
else:
plt.ylabel('error',fontsize = 32)
plt.yscale('log')
lgd = plt.legend(bbox_to_anchor=(1, 0.5))
plt.savefig(self.pps_dir + '/top_' + str(len(top_k_modes_list)) + '_normalized_relative_eigen_error.png',
bbox_extra_artists=(lgd,),
bbox_inches='tight')
plt.close()
# load MTENET
fig4_data = np.load(self.eval_dir + '/MultiTaskElasticNet_result.npz')
alphas_enet = fig4_data['alphas_enet']
coefs_enet = fig4_data['coefs_enet']
residual_array = fig4_data['residual_array']
# coefficients vs alpha & number non-zero
num_target_components = coefs_enet.shape[0]
alphas_enet_log_negative = -np.log10(alphas_enet)
# print("coef_enet real= ", np.real(coefs_enet))
# print("coef_enet imag= ", np.imag(coefs_enet))
for i_component in range(num_target_components):
plt.figure(figsize=FIG_SIZE)
cmap = self.get_cmap(len(top_k_modes_list))
for i in top_k_modes_list:
i_s = small_to_large_error_eigen_index[i]
plt.plot(alphas_enet_log_negative, abs(coefs_enet[i_component,i,:]), '-*', c=cmap(i),
label = 'No. ' + str(i + 1) + ', index = ' + str(i_s+1))
max_val = np.max(abs(coefs_enet[i_component, :, -1]))
min_val = np.min(abs(coefs_enet[i_component, :, -1]))
diss = (max_val - min_val)/2
mean = (max_val + min_val)/2
plt.xlabel(r'-$\log_{10}(\alpha)$',fontsize = 32)
plt.ylabel('abs of coefficients',fontsize = 32)
plt.ylim([mean - diss*1.05, mean + diss*3])
lgd = plt.legend(bbox_to_anchor=(1, 0.5))
plt.savefig(self.pps_dir + '/multi-elastic-net-coef-' + str(i_component+1) + '.png',
bbox_extra_artists=(lgd,),
bbox_inches='tight')
plt.close()
# total number of non-zero terms1
plt.figure(figsize=FIG_SIZE)
num_non_zeros = [len((coefs_enet[i_component, abs(coefs_enet[i_component, :, ii]) >0*np.max(abs(coefs_enet[i_component,:,ii])), ii]))
for ii in range(coefs_enet.shape[2])]
plt.plot(alphas_enet_log_negative, num_non_zeros , 'k^-')
plt.xlabel(r'-$\log_{10}(\alpha)$',fontsize = 32)
plt.ylabel('number of selected features',fontsize = 32)
lgd = plt.legend(bbox_to_anchor=(1, 0.5))
plt.savefig(self.pps_dir + '/multi-elastic-net-coef-non-zeros-' + str(i_component+1) + '.png',
bbox_extra_artists=(lgd,),
bbox_inches='tight')
plt.close()
num_non_zero_all_alpha = []
for ii in range(coefs_enet.shape[2]):
non_zero_index_per_alpha = []
for i_component in range(num_target_components):
# non_zero_index_per_alpha_per_target = abs(coefs_enet[i_component, :, ii]) > 0
non_zero_index_per_alpha_per_target = abs(coefs_enet[i_component, :, ii]) > 0*np.max(abs(coefs_enet[i_component, :, ii]))
non_zero_index_per_alpha.append(non_zero_index_per_alpha_per_target)
non_zero_index_per_alpha_all_targets = np.logical_or.reduce(non_zero_index_per_alpha)
num_non_zero_all_alpha.append(np.sum(non_zero_index_per_alpha_all_targets))
num_non_zero_all_alpha = np.array(num_non_zero_all_alpha)
# total residual vs alpha AND number of non-zero modes vs alpha
fig=plt.figure(figsize=FIG_SIZE)
ax1 = fig.add_subplot(111)
ax2 = ax1.twinx()
ax1.plot(alphas_enet_log_negative, residual_array, 'k*-')
ax1.set_xlabel(r'-$\log_{10}(\alpha)$',fontsize = 32)
ax1.set_ylabel('normalized reconstruction MSE',color='k',fontsize = 32)
# ax1.set_yscale('log')
ax2.plot(alphas_enet_log_negative, num_non_zero_all_alpha,'r*-')
ax2.set_ylabel('number of selected features',color='r',fontsize = 32)
lgd = plt.legend(bbox_to_anchor=(1, 0.5))
plt.savefig(self.pps_dir + '/multi-elastic-net-mse.png',
bbox_extra_artists=(lgd,),
bbox_inches='tight')
plt.close()
# In[70]:
freqs = r[0]
vols = r[1:]
ls = [1.53e-3] * 33
ls = np.round(np.cumsum(ls), 5)
# In[104]:
plt.close('all')
plt.figure(figsize=(10, 10))
X, Y = np.meshgrid(freqs, ls)
plt.contourf(X,
Y,
np.abs(vols),
locator=ticker.LogLocator(base=2),
cmap=plt.cm.YlGnBu_r)
plt.xlabel("frequency")
plt.ylabel('distance')
plt.colorbar()
plt.show()
# In[98]:
plt.close('all')
#plt.figure(figsize=[10,10])
#plt.plot(vs_mat[1,:])
#plt.show()
# In[90]:
def plot_surface(
self,
surf,
minvalue=None,
maxvalue=None,
contourlevels=None,
xlabelrotation=None,
colormap=None,
logarithmic=False,
): # pylint: disable=too-many-statements
"""Input a surface and plot it."""
# need a deep copy to avoid changes in the original surf
logger.info("The key contourlevels %s is not in use", contourlevels)
usesurf = surf.copy()
if usesurf.yflip < 0:
usesurf.swapaxes()
if abs(surf.rotation) > 0.001:
usesurf.unrotate()
xi, yi, zi = usesurf.get_xyz_values()
zimask = ma.getmaskarray(zi).copy() # yes need a copy!
legendticks = None
if minvalue is not None and maxvalue is not None:
minv = float(minvalue)
maxv = float(maxvalue)
step = (maxv - minv) / 10.0
legendticks = []
for i in range(10 + 1):
llabel = float("{0:9.4f}".format(minv + step * i))
legendticks.append(llabel)
zi.unshare_mask()
zi[zi < minv] = minv
zi[zi > maxv] = maxv
# need to restore the mask:
zi.mask = zimask
# note use surf.min, not usesurf.min here ...
notetxt = ("Note: map values are truncated from [" +
str(surf.values.min()) + ", " + str(surf.values.max()) +
"] " + "to interval [" + str(minvalue) + ", " +
str(maxvalue) + "]")
self._fig.text(0.99,
0.02,
notetxt,
ha="right",
va="center",
fontsize=8)
logger.info("Legendticks: %s", legendticks)
if minvalue is None:
minvalue = usesurf.values.min()
if maxvalue is None:
maxvalue = usesurf.values.max()
# this will override current instance colormap locally, and is
# therefore reset afterwards
keepcolor = self.colormap
if colormap is not None:
self.colormap = colormap
levels = np.linspace(minvalue, maxvalue, self.contourlevels)
logger.debug("Number of contour levels: %s", levels)
plt.setp(self._ax.xaxis.get_majorticklabels(), rotation=xlabelrotation)
# zi = ma.masked_where(zimask, zi)
# zi = ma.masked_greater(zi, _cxtgeo.UNDEF_LIMIT)
if ma.std(zi) > 1e-07:
uselevels = levels
else:
uselevels = 1
try:
if logarithmic is False:
locator = None
ticks = legendticks
im = self._ax.contourf(xi,
yi,
zi,
uselevels,
locator=locator,
cmap=self.colormap)
else:
logger.info("use LogLocator")
locator = ticker.LogLocator()
ticks = None
uselevels = None
im = self._ax.contourf(xi,
yi,
zi,
locator=locator,
cmap=self.colormap)
self._fig.colorbar(im, ticks=ticks)
except ValueError as err:
logger.warning("Could not make plot: %s", err)
plt.gca().set_aspect("equal", adjustable="box")
self.colormap = keepcolor
def parameter_variation(m,
sh,
thing_to_vary,
thing_to_divide_by,
factors,
num_points,
draw_plot=True):
"""
Explore the effect of varying different parameters on the upconversion yield of the acceptor.
Parameters
----------
m : tripletpairs.kineticmodelling.steadystatemodels.Merrifield or tripletpairs.kineticmodelling.steadystatemodels.MerrifieldExplicit1TT
A pre-prepared instance of either :class:`tripletpairs.kineticmodelling.steadystatemodels.Merrifield` or :class:`tripletpairs.kineticmodelling.steadystatemodels.MerrifieldExplicit1TT`.
sh : tripletpairs.spin.SpinHamiltonian
A pre-prepared instance of :class:`tripletpairs.spin.SpinHamiltonian`.
thing_to_vary : str
The name of the rate constant to vary.
thing_to_divide_by : str or None
The name of the rate constant to normalise to, if desired.
factors : 2-tuple of float
Rate constant will be varied geometrically between its starting value divided by the first entry in factors and its starting value multiplied by the second entry in factors.
num_points : int
Number of rate constant values to sample.
draw_plot : bool, optional
Whether to draw a plot of the result. The default is True.
Raises
------
TypeError
If the model given is invalid.
ValueError
If either of the given parameters is invalid.
Returns
-------
rates : numpy.ndarray
The rate constant values of thing_to_vary, note that these have not been divided by anything.
ucy_actual : numpy.ndarray
The upconversion yield as a function of rates.
ucy_nospin : numpy.ndarray
The upconversion yield as a function of rates, assuming no spin statistical effects.
"""
if (m.model_name not in ['Merrifield', 'MerrifieldExplicit1TT'
]) or (m._time_resolved):
raise TypeError('invalid model')
m.initial_weighting = {'T1': 1}
ucy_nospin = np.zeros(num_points)
ucy_actual = np.zeros(num_points)
if thing_to_vary not in m.rates:
if thing_to_vary not in ['G', 'J']:
raise ValueError('invalid thing_to_vary')
if thing_to_divide_by is not None:
if thing_to_divide_by not in m.rates:
raise ValueError('invalid thing_to_divide_by')
elif thing_to_vary in ['G', 'J']:
raise ValueError(
'thing_to_divide_by must be None if thing_to_vary is G or J')
if thing_to_vary == 'J':
vars_object = sh
else:
vars_object = m
orig = vars(vars_object)[thing_to_vary]
rates = np.geomspace(
vars(vars_object)[thing_to_vary] / factors[0],
vars(vars_object)[thing_to_vary] * factors[1], num_points)
for i, rate in enumerate(rates):
vars(vars_object)[thing_to_vary] = rate
sh.calculate_everything()
m.cslsq = sh.cslsq
m.simulate()
ucy0 = 2 * m.kSNR * m.S1 / m.G
m.cslsq = np.ones(9) / 9
m.simulate()
ucy1 = 2 * m.kSNR * m.S1 / m.G
ucy_nospin[i] = 100 * ucy1
ucy_actual[i] = 100 * ucy0
if draw_plot:
rate_labels = {
'kSF': 'Forwards SF Rate',
'k_SF': 'Backwards SF Rate',
'kHOP': 'Fowards Hop Rate',
'k_HOP': 'Backwards Hop Rate',
'kHOP2': 'Spin Loss Rate',
'kDISS': 'Spin Loss Rate',
'kTTA': 'TTA Rate',
'kTTNR': r'$^1$(TT) Decay Rate',
'kTNR': 'Triplet Decay Rate',
'kSNR': 'Singlet Decay Rate',
'G': 'Generation Rate',
'J': 'Exchange Energy'
}
if thing_to_divide_by is None:
x = rates
vline = orig
xlabel_text = rate_labels[thing_to_vary]
if thing_to_vary == 'kTTA':
xlabel_unit = r' (nm$^3$ns$^{-1}$)'
elif thing_to_vary == 'G':
xlabel_unit = r' (nm$^{-3}$ns$^{-1}$)'
elif thing_to_vary == 'J':
xlabel_unit = r' ($\mu$eV)'
else:
xlabel_unit = r' (ns$^{-1}$)'
else:
x = rates / vars(vars_object)[thing_to_divide_by]
vline = orig / vars(vars_object)[thing_to_divide_by]
xlabel_text = rate_labels[thing_to_vary] + '/' + rate_labels[
thing_to_divide_by]
if thing_to_vary == 'kTTA':
xlabel_unit = r' (nm$^3$)'
elif thing_to_divide_by == 'kTTA':
xlabel_unit = r' (nm$^{-3}$)'
else:
xlabel_unit = ''
fig, ax1 = plt.subplots(figsize=(5, 4))
ax1.semilogx(x, ucy_nospin, 'b--', label='loss')
ax1.semilogx(x, ucy_actual, 'b-', label='loss')
ax1.set_ylim([0, 100])
ax1.set_xlabel(xlabel_text + xlabel_unit, fontsize=20)
ax1.set_ylabel('Upconversion Yield (%)', fontsize=20, color='b')
ax1.tick_params(axis='y', labelcolor='b')
ax2 = ax1.twinx()
gain = 100 * (ucy_nospin - ucy_actual) / ucy_actual
ax2.semilogx(x, gain, 'r:')
ax2.set_ylim([0, 1.1 * max(gain)])
ax2.set_ylabel('Potential Gain (%)', fontsize=20, color='r')
ax2.tick_params(axis='y', labelcolor='r')
ax1.xaxis.set_major_locator(mticker.LogLocator(numticks=12))
ax1.xaxis.set_minor_locator(
mticker.LogLocator(subs=np.linspace(0.1, 0.9, 9), numticks=12))
ax1.set_xlim([min(x), max(x)])
ax1.axvline(vline, color='k', linestyle=':', linewidth=1, alpha=0.5)
for ax in [ax1, ax2]:
ax.tick_params(axis='both',
which='major',
labelsize=20,
width=1.4,
length=6)
ax.tick_params(axis='both',
which='minor',
labelsize=20,
width=1.4,
length=3)
for axis in ['top', 'bottom', 'left', 'right']:
ax.spines[axis].set_linewidth(1.4)
vars(vars_object)[thing_to_vary] = orig
return rates, ucy_actual, ucy_nospin
def histogram_1d(data,
nbins=None,
l_adjust_bins=False,
l_xlog=False,
x_label='',
y_label='',
legend_label=[],
l_color=True,
l_percentage=True,
l_rel_mode=False,
l_pope=False):
"""Probability distributions for multiple variables in a xarray-dataset."""
fig, ax = plt.subplots(figsize=(7 * 0.8, 5 * 0.8))
linestyle = [
'solid', 'dashed', 'dotted', (0, (1, 1)), (0, (3, 5, 1, 5)),
(0, (3, 1, 1, 1, 1, 1))
]
color = [
col.sol['magenta'], col.sol['blue'], 'k', col.sol['green'],
col.sol['red'], col.sol['magenta'], col.sol['cyan']
]
lw = [1., 2., 2., 2., 2., 2.]
for i, var in enumerate(data):
# var = dataset[variable]
if type(nbins) == int:
bins = np.linspace(start=var.min(), stop=var.max(),
num=nbins + 1) # 50
else:
if l_adjust_bins:
bins = np.linspace(start=m.sqrt(max(0, var.min())),
stop=m.sqrt(var.max()),
num=20)**2
else:
bins = nbins[i]
# sns.distplot(var[var.notnull()], bins=bins, kde=False, norm_hist=True) # hist_kws={'log': True})
total = var.notnull().sum().values
# metric_clean = var.fillna(-1) # works if var is positive and thus bins as well
metric_clean = var.where(
var.notnull(),
drop=True) # works if var is positive and thus bins as well
h, edges = np.histogram(metric_clean, bins=bins) # , density=True)
if l_rel_mode:
total = h.max()
bin_centre = 0.5 * (edges[1:] + edges[:-1])
dx = edges[1:] - edges[:-1]
dlogx = dx / (bin_centre * m.log(10, m.e))
if l_xlog:
h_normed = h / dlogx / total # equals density=True
else:
if l_percentage:
h_normed = h / total * 100
else:
h_normed = h / dx / total # equals density=True
if l_color:
plt.plot(bin_centre,
h_normed,
linestyle='-',
marker='o',
color=color[i],
linewidth=lw[i])
else:
h_normed_ext = np.zeros(shape=len(h_normed) + 1)
h_normed_ext[0] = h_normed[0]
h_normed_ext[1:] = h_normed
# plot a step function instead of a continuous line
h_to_plot = h_normed_ext
# h_to_plot = np.cumsum(h_normed_ext)
plt.step(edges, h_to_plot, color=color[i],
linewidth=2.) #, linestyle=linestyle[i])
if l_xlog:
plt.xscale('log')
x_major = ticker.LogLocator(base=10.0, numticks=10)
ax.xaxis.set_major_locator(x_major)
x_minor = ticker.LogLocator(base=10.0,
subs=np.arange(1.0, 10.0) * 0.1,
numticks=20)
ax.xaxis.set_minor_locator(x_minor)
ax.xaxis.set_minor_formatter(ticker.NullFormatter())
plt.ylabel(y_label)
plt.xlabel(x_label) #, **font)
if len(legend_label) != 0:
lg = plt.legend(legend_label, fontsize=14)
# this sets only the legend background color to transparent (not the surrounding box)
lg.get_frame().set_facecolor('none')
if l_pope:
ax.set_xticks([10, 30, 50, 70, 90], minor=True)
ax.grid(b=True, which='both', axis='x')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
if not l_pope:
ax.xaxis.set_major_formatter(ticker.ScalarFormatter())
ax.xaxis.set_major_formatter(
ticker.FuncFormatter(lambda y, _: '{:g}'.format(y)))
ax.tick_params(axis='x', length=8)
ax.spines['bottom'].set_position('zero')
ax.tick_params(axis='x', direction='out')
# ax.xaxis.set_ticks_position('none') # 'left', 'right'
# ax.set_xlim(6)
ax.tick_params(axis='y', direction='out')
ax.yaxis.set_ticks_position('none') # 'left', 'right'
return fig
