This repository contains code for joint estimation of multiple Granger graphical models using non-convex penalty functions.
This project is for jointly estimating multiple Granger graphical models from a set of time-series.
The contributors are: Parinthorn Manomaisaowapak and Jitkomut Songsiri
Department of Electrical Engineering, Faculty of Engineering Chulalongkorn University, Bangkok, Thailand e-mail: parinthorn@gmail.com, jitkomut.s@chula.ac.th
The detail of mathematical formulation is described in our manuscript:
Joint learning of multiple Granger causal networks via non-convex regularizations: Inference of group-level brain connectivity arXiv:2105.07196 [cs.LG]
Description demo.m: Matlab script for computing joint GC networks from generated data
data_generation_code/: This folder contains data generation process used in simulation experiments.
experiment/: This folder contains the benchmarking experiments and the real data experiment.
function/: This folder contains the functions used in the experiments.
results2plot: This folder contains performance data of the simulation experiments and used to plot the figures in the manuscript.