Abstract: It is always tempting, while analyzing evolution of a multicomponent system, to reduce its dimension. One need though to keep key properties of the original system while making such reduction. I'll talk about recently emerged theory of isospectral transformations of matrices and networks which already allowed to advance some traditional areas of mathematics as estimation of matrices'/polynomials' spectra, global stability of dynamical networks, etc, and looks quite promising for analysis of real world networks. The talk will be accessible to undergraduates with knowledge of basics of linear algebra.