tfp.experimental.sequential.EnsembleKalmanFilterState

State for Ensemble Kalman Filter algorithms.

Contents:

  • step: Scalar Integer tensor. The timestep associated with this state.
  • particles: Structure of Floating-point Tensors of shape [N, B1, ... Bn, Ek] where N is the number of particles in the ensemble, Bi are the batch dimensions and Ek is the size of each state.
  • extra: Structure containing any additional information. Can be used for keeping track of diagnostics or propagating side information to the Ensemble Kalman Filter.

step

particles

extra