Parameter docs shared by GPModel, gpb.cv, and gpboost
A string specifying the likelihood function (distribution) of the response variable
Default = "gaussian"
A vector or matrix with labels of group levels for grouped random effects
A vector or matrix with covariate data for grouped random coefficients
A vector with indices that relate every random coefficients
to a "base" intercept grouped random effect. Counting starts at 1.
A matrix with coordinates (features) for Gaussian process
A vector or matrix with covariate data for Gaussian process random coefficients
A string specifying the covariance function for the Gaussian process.
The following covariance functions are available:
"exponential", "gaussian", "matern", "powered_exponential", "wendland", and "exponential_tapered".
For "exponential", "gaussian", and "powered_exponential", we follow the notation and parametrization of Diggle and Ribeiro (2007).
For "matern", we follow the notation of Rassmusen and Williams (2006).
For "wendland", we follow the notation of Bevilacqua et al. (2019).
A covariance function with the suffix "_tapered" refers to a covariance function that is multiplied by
a compactly supported Wendland covariance function (= tapering)
A numeric specifying the shape parameter of the covariance function
(=smoothness parameter for Matern and Wendland covariance). For the Wendland covariance function,
we follow the notation of Bevilacqua et al. (2019)).
This parameter is irrelevant for some covariance functions such as the exponential or Gaussian.
A numeric specifying the range parameter of the Wendland covariance function / taper. We follow the notation of Bevilacqua et al. (2019)
A boolean. If true, the Vecchia approximation is used
An integer specifying the number of neighbors for the Vecchia approximation
A string specifying the ordering used in the Vecchia approximation.
"none" means the default ordering is used, "random" uses a random ordering
A string specifying the type of Vecchia approximation used for making predictions.
"order_obs_first_cond_obs_only" = observed data is ordered first and the neighbors are only observed points,
"order_obs_first_cond_all" = observed data is ordered first and the neighbors are selected among all points
(observed + predicted), "order_pred_first" = predicted data is ordered first for making predictions,
"latent_order_obs_first_cond_obs_only" = Vecchia approximation for the latent process and observed data is
ordered first and neighbors are only observed points, "latent_order_obs_first_cond_all" = Vecchia approximation
for the latent process and observed data is ordered first and neighbors are selected among all points
an integer specifying the number of neighbors for the Vecchia approximation
for making predictions
A vector with IDs / labels indicating independent realizations of
random effects / Gaussian processes (same values = same process realization)
If TRUE, the data (groups, coordinates, covariate data for random coefficients) is freed in R after initialization
A vector with response variable data
A matrix with covariate data for fixed effects ( = linear regression term)
A list with parameters for the model fitting / optimization
optimizer_cov Optimizer used for estimating covariance parameters. Options: "gradient_descent", "fisher_scoring", and "nelder_mead". Default = "gradient_descent"
optimizer_coef Optimizer used for estimating linear regression coefficients, if there are any (for the GPBoost algorithm there are usually none). Options: "gradient_descent", "wls", and "nelder_mead". Gradient descent steps are done simultaneously with gradient descent steps for the covariance parameters. "wls" refers to doing coordinate descent for the regression coefficients using weighted least squares. Default="wls" for Gaussian data and "gradient_descent" for other likelihoods.
maxit Maximal number of iterations for optimization algorithm. Default=1000.
delta_rel_conv Convergence criterion: stop optimization if relative change in parameters is below this value. Default=1E-6.
init_coef Initial values for the regression coefficients (if there are any, can be NULL). Default=NULL.
init_cov_pars Initial values for covariance parameters of Gaussian process and random effects (can be NULL). Default=NULL.
lr_coef Learning rate for fixed effect regression coefficients if gradient descent is used. Default=0.1.
lr_cov Learning rate for covariance parameters. If <= 0, internal default values are used. Default value = 0.1 for "gradient_descent" and 1. for "fisher_scoring"
use_nesterov_acc If TRUE Nesterov acceleration is used. This is used only for gradient descent. Default=TRUE
acc_rate_coef Acceleration rate for regression coefficients (if there are any) for Nesterov acceleration. Default=0.5.
acc_rate_cov Acceleration rate for covariance parameters for Nesterov acceleration. Default=0.5.
momentum_offset Number of iterations for which no mometum is applied in the beginning. Default=2.
trace If TRUE, information on the progress of the parameter optimization is printed. Default=FALSE.
convergence_criterion The convergence criterion used for terminating the optimization algorithm. Options: "relative_change_in_log_likelihood" (default) or "relative_change_in_parameters".
std_dev If TRUE, (asymptotic) standard deviations are calculated for the covariance parameters
A vector of optional external fixed effects which are held fixed during training.