New method improves SDE surrogate model accuracy for path-dependent observables

This paper introduces a novel variational loss function for learning surrogate models of stochastic differential equations (SDEs). The proposed goal-oriented learning approach uses an error bound for path-space observables, which standard loss functions often fail to provide. This method is demonstrated to improve accuracy in predicting statistics like first hitting times and exhibits robustness to shifts in data distribution, particularly for overdamped Langevin systems. AI

IMPACT Introduces a novel loss function for SDE surrogate models, potentially improving efficiency and accuracy in simulations for various applications.

RANK_REASON This is a research paper detailing a new methodology for learning surrogate models of SDEs. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• Joanna Zou
• stochastic differential equations

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →