Real Time Network Estimation

This is a "real-time" implementation of SINGLE algorithm. Allows for the estimation of networks on-the-fly

Example:

We provide simulated data in the Sample Data folder.

import numpy, pandas
from ThetaEstimation import onlineSINGLE

# read in data:
data = numpy.array(pandas.read_csv('Sample Data/biggerdata.csv'))
# This dataset contains 300 observations over 10 nodes.

# initialise - note that here the first 15 observations are provided as a burn-in. 
# Alternatively only 1 observation can be provided (i.e. no burn-in)

Theta = onlineSINGLE(data=data[0:15,:], l1=1, l2=.5, ff=.9)

# For each of the remaining 285 observations we can update our 
# estimate of the precision as follows:
for i in range(15, 300):
    Theta.updateTheta(newX = data[i,:])
    
# results for precision matrices are stored in Theta.Z (a list)    

The resulting estimated partial correlations for a given edge are shown below. In green we have the partial correlation where temporal homogeneity is not enforced and in blue (dashed) we have the estimated partial correlation using the proposed method which is noticably smoother.