“…A vast literature [9], [10], [11] has focused on reducing the dimensionality of complex dynamical systems. The fundamental approach of model reduction is to generate a lower order model that preserves key features of the original system, see e.g., [12], [13], [14], [15], [16], [17]. Typically, this implies that information about the dominant modes/components of the system is preserved, while information about the nondominant modes/components is removed.…”