def load_data_and_reshape(dirname, names, remap_infinity=False):
d = {}
for name in names:
# print("reading " + name)
if "estimate_MFPT" in dirname:
# Use a masked array. Mask missing values...
m = np.genfromtxt(dirname + "/" + name + ".dat",
usemask=True,
missing_values="NaN,nan")
# ... mask the diagonal...
np.fill_diagonal(m, np.ma.masked)
# ... and those where mfpte-len < 5...
filename = dirname + "/MFPTE_len.dat"
mfpte_len = np.genfromtxt(filename)
min_vals = 5 # an attempt at reliability
m[mfpte_len < min_vals] = np.ma.masked
# FIXME mask those where MFPT < 0.1?
else:
m = np.genfromtxt(dirname + "/" + name + ".dat")
if remap_infinity:
# substitute an arbitrary large value for any infinities
map_infinity_to_large(m)
d[name] = m.reshape(len(m)**2)
return d
def load_data_and_reshape(dirname, names, remap_infinity=False):
d = {}
for name in names:
# print("reading " + name)
if "estimate_MFPT" in dirname:
# Use a masked array. Mask missing values...
m = np.genfromtxt(dirname + "/" + name + ".dat",
usemask=True, missing_values="NaN,nan")
# ... mask the diagonal...
np.fill_diagonal(m, np.ma.masked)
# ... and those where mfpte-len < 5...
filename = dirname + "/MFPTE_len.dat"
mfpte_len = np.genfromtxt(filename)
min_vals = 5 # an attempt at reliability
m[mfpte_len < min_vals] = np.ma.masked
# FIXME mask those where MFPT < 0.1?
else:
m = np.genfromtxt(dirname + "/" + name + ".dat")
if remap_infinity:
# substitute an arbitrary large value for any infinities
map_infinity_to_large(m)
d[name] = m.reshape(len(m)**2)
return d
def make_grid(w, names, filename, colour_map=None, bar=True):
# we dont rescale the data. matshow() internally scales the data
# so that the smallest numbers go to black and largest to white.
# A uniform array will cause a "RuntimeWarning: invalid value
# encountered in divide" when calculating the colorbar. So ignore
# that.
old = np.seterr(invalid='ignore')
# Can put NaN on the diagonal to avoid plotting it -- makes a bit
# more space available for other data. But misleading.
# w += np.diag(np.ones(len(w)) * np.nan)
# if w contains any NaNs we'll get a misleading result: they won't
# be plotted, so will appear white, as will the largest finite
# values of w. So first, replace them with a large value -- 100
# times the largest finite value.
map_infinity_to_large(w)
side = 8.0
figsize = (side, side)
fig = plt.figure(figsize=figsize)
ax = fig.add_subplot(1, 1, 1)
# consider other colour maps: cm.gray_r for reversed, autumn, hot,
# gist_earth, copper, ocean, some others, or a custom one for
# nicer images (not for publication, maybe).
# im = ax.matshow(w, cmap=cm.gray, interpolation="none")
if colour_map is None: colour_map = cm.gray
im = ax.matshow(w, cmap=colour_map, interpolation="nearest")
if bar:
fig.colorbar(im, shrink=0.775)
if names:
# Turn labels on
ax.xaxis.set_major_locator(MyLocator(1, 0))
ax.yaxis.set_major_locator(MyLocator(1, 0))
ax.set_xticklabels(names, range(len(names)), rotation=90, size=1.0)
ax.set_yticklabels(names, range(len(names)), size=1.0)
else:
# Turn them off
ax.set_xticklabels([], [])
ax.set_yticklabels([], [])
ax.tick_params(length=0, pad=3.0)
fig.savefig(filename + ".pdf", dpi=300, bbox_inches='tight')
fig.savefig(filename + ".eps", dpi=300, bbox_inches='tight')
fig.savefig(filename + ".png", dpi=300, bbox_inches='tight')
plt.close(fig)
# restore old error settings
np.seterr(**old)
def make_grid(w, names, filename, colour_map=None, bar=True):
# we dont rescale the data. matshow() internally scales the data
# so that the smallest numbers go to black and largest to white.
# A uniform array will cause a "RuntimeWarning: invalid value
# encountered in divide" when calculating the colorbar. So ignore
# that.
old = np.seterr(invalid='ignore')
# Can put NaN on the diagonal to avoid plotting it -- makes a bit
# more space available for other data. But misleading.
# w += np.diag(np.ones(len(w)) * np.nan)
# if w contains any NaNs we'll get a misleading result: they won't
# be plotted, so will appear white, as will the largest finite
# values of w. So first, replace them with a large value -- 100
# times the largest finite value.
map_infinity_to_large(w)
side = 8.0
figsize = (side, side)
fig = plt.figure(figsize=figsize)
ax = fig.add_subplot(1, 1, 1)
# consider other colour maps: cm.gray_r for reversed, autumn, hot,
# gist_earth, copper, ocean, some others, or a custom one for
# nicer images (not for publication, maybe).
# im = ax.matshow(w, cmap=cm.gray, interpolation="none")
if colour_map is None: colour_map = cm.gray
im = ax.matshow(w, cmap=colour_map, interpolation="nearest")
if bar:
fig.colorbar(im, shrink=0.775)
if names:
# Turn labels on
ax.xaxis.set_major_locator(MyLocator(1, 0))
ax.yaxis.set_major_locator(MyLocator(1, 0))
ax.set_xticklabels(names, range(len(names)), rotation=90, size=1.0)
ax.set_yticklabels(names, range(len(names)), size=1.0)
else:
# Turn them off
ax.set_xticklabels([], [])
ax.set_yticklabels([], [])
ax.tick_params(length=0, pad=3.0)
fig.savefig(filename + ".pdf", dpi=300, bbox_inches='tight')
fig.savefig(filename + ".eps", dpi=300, bbox_inches='tight')
fig.savefig(filename + ".png", dpi=300, bbox_inches='tight')
plt.close(fig)
# restore old error settings
np.seterr(**old)