Developed software on dynamic network identification:
- Matlab tool for Bayesian topology identification of linear dynamic networks.
Given a set of time series, the algorithm estimates the dynamic interconnection structure of the signals, which is equivalent to the estimation of Granger causality. The interaction of the signals is modeled by a network of transfer functions in the algorithm, which then employs a Bayesian criterion and a search algorithm to obtain an estimate of the network topology.
Matlab software is available through CodeOcean. The method is documented in an ECC-2019 paper.
- A GUI-supported Matlab Toolbox SYSDYNET for structural network operations, network identifiability analysis and synthesis, and predictor model construction for estimating single modules in a dynamic network.
The Toolbox operates on the topology (adjacency matrix) of a network, that includes the location and correlation structure of external disturbances and excitation signals. It provides the user with tools to select measurable nodes, to assign a priori fixed/known modules, to analyse and synthesize full network or single module identifiability, and to construct and analyze predictor models for single module identification.A public Beta-version has been released in the early spring van 2023 and is available through the SYSDYNET Landing Page.
