Deutsch Intern
    Machine Learning for Complex Networks

    New preprint on inference of higher-order interactions


    Given the observation of a dynamical process in an an unobserved networked system, what can we say about the underlying interaction topology? Our latest study led by Dr. Unai Alvarez-Rodriguez addresses this important question.

    A key difference to prior research on network inference is that we propose a statistical framework that unifies different types of higher-order interactions in networks, i.e. interactions that involve more than just a single interaction between a pair of nodes. In our work, we instead consider time-ordered sequences of multi-body interactions, i.e. particular sequences of two or more nodes that interact with each other. 

    With our work, we introduce a general methodology that allows to decompose observed sequences of system states into a parsimonious interaction model. The study was led by Dr. Unai Alvarez-Rodriguez. The preprint is available on arXiv.