Identification of so-called dynamic networks is one of the most challenging problems appeared recently in control literature . Bayesian approach introduces a new approach for linear dynamic networksidentification . It implements aggressive shrinkage of small impulse responses while larger impulse responses are conveniently subject to stable spline regularization . Inference is performed by a Markov Chain MonteCarlo scheme, tailored to the dynamic context and able to efficiently returnt the posterior of the modules in sampled form. We include numerical studies that show how the new approach can accurately reconstruct sparse networks dynamics also when thousands of unknown impulse response coefficients must be inferred from data sets of relatively small size . The resulting model is called stablespline horseshoe (SSH) prior, which implements aggressive shrinking of smallimpulse responses. It is convenient subject tostable spline normalization. Inference was tailored to a dynamic context. It includes a new model of the dynamic network dynamics. The new approach has been used to accurately reconstructs dense networks dynamics. It can also be used to reconstruct sparse Networks dynamics. We have published a number of numerical studies to help us with our new approach.

Author(s) : Gianluigi Pillonetto

Links : PDF - Abstract

Code :
Coursera

Keywords : dynamic - approach - networks - dynamics - responses -

Leave a Reply

Your email address will not be published. Required fields are marked *