［ 2017年02月27日 ］
RIKEN International Symposium on Data Assimilation 2017
"Principled Data Assimilation in Nonlinear Complex Systems"
We formulate a framework for transferring information from an observed complex nonlinear system to a physically based model of the processes underlying the observations. To evaluate the statistical path integral involved, we discuss in some detail the variational method of Laplace for evaluating the required integrals. We discuss how to determine how many observations are required to allow for accurate state estimations and predictions. We discuss how to proceed when too few measurements are available. We introduce a method to find the global minimum of the nonlinear function, the action, identified in the path integral. Examples are drawn from neurobiology and from numerical weather prediction.
|名前:Henry Abarbanel||所属:UC San Diego|