Browse files

Edited readme

  • Loading branch information...
davidhallac committed Jun 15, 2017
1 parent a151916 commit 259b61ab33f9172989e8caaa7332d3f042b7612d
Showing with 45 additions and 2 deletions.
  1. +45 −2
  2. 0 →
@@ -1,5 +1,48 @@
TVGL is a python solver for inferring dynamic networks from raw time series data. For implementation details, refer to the paper, available at:
Code to run experiments for ``Network Inference via the Time-Varying Graphical Lasso''.
Stay tuned for a full solver (coming soon!)
Download & Setup
Download the source code by running the following code in the terminal:
git clone
TVGL can be called through the following file:
data : a T-by-n numpy array with the raw data (each row is a new timestamp)
lengthOfSlice : Number of samples in each ``slice'', or timestamp
lamb : the lambda regularization parameter controlling the network sparsity (as described in the paper)
beta : the beta parameter controlling the temporal consistency (as described in the paper)
indexOfPenalty : The regularization penalty to use (1 = L1, 2 = L2, 3 = Laplacian, 4 = L_inf, 5 = perturbed node)
verbose = False : Whether or not to run ADMM in ``verbose'' mode (to print intermediate steps)
eps = 3e-3 : Threshold at which we treat output network weight as zero
epsAbs = 1e-3 : ADMM absolute tolerance threshold (see full details in
epsRel = 1e-3 : ADMM relative tolerance threshold (see
Running the following script provides an example of how the TVGL solver can be used:
File renamed without changes.

0 comments on commit 259b61a

Please sign in to comment.