IKF_CG_particle_cell: Inverse Kalman Filter with Conjugate Gradient for Particle...

View source: R/RcppExports.R

IKF_CG_particle_cellR Documentation

Inverse Kalman Filter with Conjugate Gradient for Particle Systems

Description

Implements the IKF-CG (Inverse Kalman Filter - Conjugate Gradient) algorithm for the one-interaction physical model with 1D output, different particle numbers for different time, and non-identity diagonal noise.

Usage

IKF_CG_particle_cell(param, kernel_type, delta_x_all, output, A_all_v, 
                     sort_d_all_ix, sigma_2_vec, num_neighbors_vec, 
                     tilde_nu, D, n_t_record, tol = 1e-6, maxIte = 1000)

Arguments

param

A numeric vector containing model parameters. The first element is the log of beta (inverse range parameter), and the second element is the log of tau (variance ratio parameter).

kernel_type

A string specifying the covariance kernel type: "matern_5_2" or "exp".

delta_x_all

A numeric vector of successive differences of sorted inputs.

output

A numeric vector of observations.

A_all_v

A numeric vector containing the non-zero entries in matrix A.

sort_d_all_ix

An integer vector of indices for sorting distances.

sigma_2_vec

A numeric vector of variances for each time point.

num_neighbors_vec

An integer vector specifying number of neighbors for each observation.

tilde_nu

A numeric value for numerical stabilization.

D

An integer specifying the dimension of observations.

n_t_record

An integer vector containing number of particles at each time point.

tol

A numeric value specifying convergence tolerance (default: 1e-6).

maxIte

An integer specifying maximum number of iterations (default: 1000).

Value

Returns a list containing:

x

A numeric vector of solved coefficients.

resid

A numeric vector of residuals at each iteration.

ite

An integer indicating the number of iterations performed.

References

Fang, X., & Gu, M. (2024). The inverse Kalman filter. arXiv:2407.10089.


FastGaSP documentation built on April 4, 2025, 5:16 a.m.