def plot_pseudospectra_single(perf_data, contour=False, levels=False, invert=True, linewidth=3.0):
W = perf_data.net.W
N = W.shape[0]
plt.figure()
fig = plt.gcf()
ax = fig.add_subplot(1, 1, 1)
if contour:
plot_pseudospectra_contour(W, bounds=[-3, 3, -3, 3], npts=50, ax=ax, colorbar=False, invert=False, log=False, levels=levels, linewidth=linewidth)
else:
plot_pseudospectra(W, bounds=[-3, 3, -3, 3], npts=50, ax=ax, colorbar=True, log=True, invert=invert)
plt.axhline(0.0, color='k', axes=ax)
plt.axvline(0.0, color='k', axes=ax)
cir = pylab.Circle((0.0, 0.0), radius=1.00, fc='gray', fill=False)
pylab.gca().add_patch(cir)
plt.xticks([], [])
plt.yticks([], [])
if not contour:
for m in range(N):
ev = perf_data.eigen_values[m]
ax.plot(ev.real, ev.imag, 'ko', markerfacecolor='w')
def plot_pseudospectra_by_perf(pdata, perf_attr='logit_perf', contour=False, levels=None, invert=True):
num_plots = 25
indx_off = [0, len(pdata)-num_plots]
weights = [[], []]
for k,offset in enumerate(indx_off):
pend = offset + num_plots
for m,p in enumerate(pdata[offset:pend]):
weights[k].append(p.W)
perrow = int(np.sqrt(num_plots))
percol = perrow
for j,offset in enumerate(indx_off):
fig = plt.figure()
fig.subplots_adjust(wspace=0.1, hspace=0.1)
for k in range(num_plots):
W = weights[j][k]
N = W.shape[0]
ax = fig.add_subplot(perrow, percol, k)
if contour:
plot_pseudospectra_contour(W, bounds=[-3, 3, -3, 3], npts=50, ax=ax, colorbar=False, invert=False, log=False, levels=levels)
else:
plot_pseudospectra(W, bounds=[-3, 3, -3, 3], npts=50, ax=ax, colorbar=True, log=True, invert=invert)
plt.axhline(0.0, color='k', axes=ax)
plt.axvline(0.0, color='k', axes=ax)
cir = pylab.Circle((0.0, 0.0), radius=1.00, fc='gray', fill=False)
pylab.gca().add_patch(cir)
plt.xticks([], [])
plt.yticks([], [])
if not contour:
p = pdata[offset + k]
for m in range(N):
ev = p.eigen_values[m]
ax.plot(ev.real, ev.imag, 'ko', markerfacecolor='w')
if offset == 0:
plt.suptitle('Pseudospectra of Top %d Networks' % num_plots)
else:
plt.suptitle('Pseudospectra of Bottom %d Networks' % num_plots)
plt.show()