{"id":"arxiv/2607.16035","url":"https://arxiv.org/abs/2607.16035","tags":["stat.ME","math.DS","math.PR","nlin.CD","physics.data-an"],"score":0,"title":"Dynamic models with $p$ parameters are identified by $2p+1$ random features","domain":"stat.ME","source":"arxiv","authors":["Michael Wieck-Sosa","Cosma Rohilla Shalizi"],"reviews":0,"subject":"stat.ME","summary":"A foundational principle in nonlinear dynamics is that the structure of a dynamical system can be recovered from a small number of generic measurements or coordinates. We develop an analogous principle for the identification of dynamic models for time series \\textit{with noise}, which builds on previous identification results for noiseless dynamical systems. The noise is allowed to be non-iid, non-Gaussian, and dependent on the state. Our results cover noisily observed differential equations and discrete-time dynamical systems, as well as stochastic models with process noise. We illustrate…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-07-17","reactions":{}}