Researchers have developed a new data-driven method for calibrating stochastic reduced-order models, which are used to represent complex system dynamics. This approach leverages a novel relationship between model coefficients and the conditional score of the finite-time transition density. By expressing derivatives of correlation functions as stationary expectations, the method constrains model parameters directly from finite-lag statistics without requiring trajectory differentiation or state-space partitioning. The validated framework offers a scalable way to learn stochastic models from data that accurately capture statistical and dynamical properties. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Introduces a novel method for learning stochastic models from data, potentially improving simulations of complex systems.
RANK_REASON This is a research paper published on arXiv detailing a new modeling technique.