m = 4 # number of sources to estimate
h = 66 # number of lags for whiteness test
i = 0
for p in [22, 33]:
i += 1
print("Model order:", p)
print(" Performing CSPVARICA")
var = scot.backend["var"](p)
result = cspvarica(data, var, classes, m)
if result.a.is_stable():
s = ""
else:
s = "*NOT* "
print(" VAR model is {}stable.".format(s))
# discard the first p residuals
# r = result.var_residuals[p:, :, :]
print(" Testing VAR residuals for whiteness up to lag", h)
pr = splot.plot_whiteness(result.a, h, repeats=100, axis=plt.subplot(2, 1, i))
if pr < 0.05:
plt.gca().set_title("model order {}: residuals significantly " "non-white with p={:f}".format(p, pr))
else:
plt.gca().set_title("model order {}: residuals white " "with p={:f}".format(p, pr))
splot.show_plots()
i += 1
print('Model order:', p)
print(' Performing CSPVARICA')
var = scot.backend['var'](p)
result = cspvarica(data, var, classes, m)
if result.a.is_stable():
s = ''
else:
s = '*NOT* '
print(' VAR model is {}stable.'.format(s))
# discard the first p residuals
# r = result.var_residuals[p:, :, :]
print(' Testing VAR residuals for whiteness up to lag', h)
pr = splot.plot_whiteness(result.a,
h,
repeats=100,
axis=plt.subplot(2, 1, i))
if pr < 0.05:
plt.gca().set_title('model order {}: residuals significantly '
'non-white with p={:f}'.format(p, pr))
else:
plt.gca().set_title('model order {}: residuals white '
'with p={:f}'.format(p, pr))
splot.show_plots()
np.fill_diagonal(alpha['foot'], 0)
order = None
for cls in ['hand', 'foot']:
np.fill_diagonal(alpha[cls], 0)
w = alpha[cls]
m = alpha[cls] > 4
# use same ordering of components for each class
if not order:
order = cuthill_mckee(m)
# fixed color, but alpha varies with connectivity strength
r = np.ones(w.shape)
g = np.zeros(w.shape)
b = np.zeros(w.shape)
a = (alpha[cls] - 4) / max(np.max(alpha['hand'] - 4),
np.max(alpha['foot'] - 4))
c = np.dstack([r, g, b, a])
plotting.plot_circular(colors=c,
widths=w,
mask=m,
topo=topo,
topomaps=mixmaps,
order=order)
plt.title(cls)
plotting.show_plots()
topo.set_locations(locs)
mixmaps = plotting.prepare_topoplots(topo, ws.mixing_)
# Force diagonal (self-connectivity) to 0
np.fill_diagonal(alpha['hand'], 0)
np.fill_diagonal(alpha['foot'], 0)
order = None
for cls in ['hand', 'foot']:
np.fill_diagonal(alpha[cls], 0)
w = alpha[cls]
m = alpha[cls] > 4
# use same ordering of components for each class
if not order:
order = cuthill_mckee(m)
# fixed color, but alpha varies with connectivity strength
r = np.ones(w.shape)
g = np.zeros(w.shape)
b = np.zeros(w.shape)
a = (alpha[cls]-4) / max(np.max(alpha['hand']-4), np.max(alpha['foot']-4))
c = np.dstack([r, g, b, a])
plotting.plot_circular(colors=c, widths=w, mask=m, topo=topo,
topomaps=mixmaps, order=order)
plt.title(cls)
plotting.show_plots()
