Trains a Semi-Parametric Quantile Regression with eXtreme modeling (SPQRX) neural network using training and validation datasets. The function standardizes covariates, normalizes the response variable to [0,1], constructs spline basis representations, and fits the SPQRX architecture.
Usage
fit.spqrx(
input_dim,
hidden_dim,
n.knots,
x_training,
x_validation,
y_training,
y_validation,
hyperparameter = NULL,
package.it = TRUE,
pre_normalize = FALSE,
pre_train = TRUE
)Arguments
- input_dim
Integer. Number of input covariates.
Integer or vector. Number of hidden units in the network.
- n.knots
Integer. Number of spline knots.
- x_training
Matrix or data frame of training covariates.
- x_validation
Matrix or data frame of validation covariates.
- y_training
Numeric vector of training responses.
- y_validation
Numeric vector of validation responses.
- hyperparameter
List. Optional hyperparameters including tail parameters (e.g., p_a, p_b, c1, c2).
- package.it
Logical. If TRUE, returns a packaged model object with normalization and knot information.
- pre_normalize
Logical. Indicates whether prediction functions should assume pre-normalized inputs.
- pre_train
Logical. If TRUE, performs pre-training of the heavy-tail component.