Language models as master equation solvers

arXiv:2308.02514v2 Announce Type: replace
Abstract: Master equations are of fundamental importance in modeling stochastic dynamical systems.However, solving master equations is challenging due to the exponential increase in the number of possible states or trajectories with the dimension of the state space. In this study, we propose repurposing language models as a machine learning approach to solve master equations. We design a prompt-based neural network to map rate parameters, initial conditions, and time values directly to the state joint probability distribution …

abstract arxiv cs.ai cs.lg equation importance language language models machine machine learning master math.ds modeling space state stochastic study systems type