compute recurrence in networks

fmri_recurrence.py

import numpy as np
from pyunicorn.timeseries import RecurrenceNetwork
from matplotlib import pyplot as pl
from mpl_toolkits.mplot3d import Axes3D

# Loading the data for 1 subject
data = np.load('all_fmri_data.npy')[0]
pl.plot(data)
pl.show()

# Find a good time lag for embedding
possible_lags = np.arange(1, data.size)
for lag in possible_lags:
    ac = np.corrcoef(data[lag:], data[:-lag])[0, 1]
    # Stop as soon as AC(tau) drops below 1/e
    if ac <= 1. / np.exp(1):
        break
print "Using lag:", lag

# Construct a recurrence network
r_net = RecurrenceNetwork(data, dim=3, tau=lag, recurrence_rate=0.05)
r_net.save("ice_core_recurrence_network.gml")

# Visualize the embedded time series
embedded_ts = r_net.embedding
x, y, z = embedded_ts.T
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.plot(x, y, z)
plt.show()

# Visualize the recurrence matrix
adjacency_matrix = r_net.adjacency
plt.imshow(adjacency_matrix, cmap="binary")
plt.show()

# Show some measures
print "Transitivity:", r_net.transitivity()
print "Determinism", r_net.determinism()