DELAY_MIN = 1
DELAY_MAX = 6
corrMat, corrDelMat = crossCorr(data, DELAY_MIN, DELAY_MAX, est='corr')
sprMat, sprDelMat = crossCorr(data, DELAY_MIN, DELAY_MAX, est='spr')
rez_file_h5.close()
#######################
# Plot
#######################
plotImshowMat([corrMat, corrDelMat, sprMat, sprDelMat],
['Corr', 'Spr', 'CorrDel', 'SprDel'],
"Cross-correlation for " + filename,
shape=(2, 2),
lims=[[-1, 1], [0, DELAY_MAX], [-1, 1], [0, DELAY_MAX]],
draw=True,
savename=filename.split('.')[0] + '.png')
########################
## Cross-correlation with velocities
########################
#data2 = np.vstack((DS1.data[:, :-1], DS1.data[:, 1:] - DS1.data[:, :-1]))
#print(data2.shape)
#corrMat, corrDelMat = crossCorr(data2, DELAY_MIN, DELAY_MAX, est='corr')
#sprMat, sprDelMat = crossCorr(data2, DELAY_MIN, DELAY_MAX, est='spr')
# b) Initialise analysis object and define settings
network_analysis = BivariateTE()
settings = {
'cmi_estimator': 'JidtGaussianCMI',
'max_lag_sources': DELAY_MAX,
'min_lag_sources': DELAY_MIN
}
# c) Run analysis
results = network_analysis.analyse_network(settings=settings,
data=dataIDTxl)
# d) Convert results into comfortable form
results_mat = idtxlResults2matrix(results, N_NODE, method='BivariateTE')
rez_file_h5.close()
#######################
# Plot Results
#######################
plotImshowMat(results_mat, ['BivariateTE', "Delays", "p-value"],
'IDTxl analysis of ' + filename, (1, 3),
lims=[[0, 1], [0, DELAY_MAX], [0, 1]],
draw=True,
savename=filename.split('.')[0] + '.png')
plt.show()
DELAY_MAX = 2
#data = np.random.uniform(0, 1, N_NODE*N_DATA).reshape((N_NODE, N_DATA))
# Generate progressively more random data
data = np.zeros((N_NODE, N_DATA))
data[0] = np.random.normal(0, 1, N_DATA)
for i in range(1, N_NODE):
data[i] = np.hstack((data[i-1][1:], data[i-1][0]))
corrMat, corrDelMat = crossCorr(data, DELAY_MIN, DELAY_MAX, est='corr')
sprMat, sprDelMat = crossCorr(data, DELAY_MIN, DELAY_MAX, est='spr')
plotImshowMat(
[corrMat, corrDelMat, sprMat, sprDelMat],
np.array(["Correlation", "Corr. Delays", "Spearmann", "Spr. Delays"]),
"Test 1: Channels are shifts of the same data",
(2, 2),
lims = [[-1, 1], [0, DELAY_MAX], [-1, 1], [0, DELAY_MAX]],
draw = True
)
'''
Test 2:
Generate random data, all copies of each other, each following one a bit more noisy than prev
Expected outcomes:
* Correlation decreases with distance between nodes, as they are separated by more noise
* Correlation should be approx the same for any two nodes given fixed distance between them
'''
N_NODE = 5
N_DATA = 1000
# 4) For each pair, select connection with strongest TE, out of those that are not NAN
TE_mat_2D = np.zeros((N_NODE, N_NODE)) + np.nan
P_mat_2D = np.zeros((N_NODE, N_NODE)) + np.nan
for i in range(N_NODE):
for j in range(N_NODE):
maxThis = np.nanmax(TE_mat_3D[i, j])
if not np.isnan(maxThis):
TE_mat_2D[i, j] = maxThis
P_mat_2D[i, j] = P_mat_3D[i,j,np.nanargmax(TE_mat_3D[i,j])]
else:
Lag_mat_2D[i, j] = np.nan
#######################
# Plot Results
#######################
plotImshowMat(
[TE_mat_2D, Lag_mat_2D, P_mat_2D, numNotNAN],
['NiftyTE', "Delays", "p-value", "count"],
'NIfTy analysis of ' + filename,
(1, 4),
lims = [[0, 1], [0, DELAY_MAX], [0, 1], [0, N_NODE]],
draw = True,
savename = filename.split('.')[0] + '.png'
)
plt.show()
'cmi_estimator': 'JidtGaussianCMI',
'max_lag_sources': DELAY_MAX,
'min_lag_sources': DELAY_MIN
}
# c) Run analysis
results_lst = [
net_analysis.analyse_network(settings=settings, data=dataIDTxl)
for net_analysis in network_analysis_lst
]
# d) Convert results into comfortable form
results_mat_lst = [
idtxlResults2matrix(results, N_NODE, method=method)
for results, method in zip(results_lst, method_lst)
]
#######################
# Plot Results
#######################
for results_matrices, method, title in zip(results_mat_lst, method_lst,
title_lst):
plotImshowMat(results_matrices,
np.array([method, "Delays", "p-value"]),
title, (1, 3),
lims=[[0, 1], [0, DELAY_MAX], [0, 1]],
draw=True)
plt.show()