Smooth approximation to the tilted absolute value cost function used to fit a QRNN model. Optional left censoring, monotone constraints, and additive constraints are supported.
qrnn.cost(weights, x, y, n.hidden, w, tau, lower, monotone,
additive, eps, Th, Th.prime, penalty, unpenalized)weight vector of length returned by qrnn.initialize.
covariate matrix with number of rows equal to the number of samples and number of columns equal to the number of variables.
response column matrix with number of rows equal to the number of samples.
number of hidden nodes in the QRNN model.
vector of weights with length equal to the number of samples;
NULL gives equal weight to each sample.
desired tau-quantile.
left censoring point.
column indices of covariates for which the monotonicity constraint should hold.
force additive relationships.
epsilon value used in the approximation functions.
derivative of the hidden layer transfer function Th.
weight penalty for weight decay regularization.
column indices of covariates for which the weight penalty should not be applied to input-hidden layer weights.
numeric value indicating tilted absolute value cost function, along with attribute containing vector with gradient information.
Cannon, A.J., 2011. Quantile regression neural networks: implementation in R and application to precipitation downscaling. Computers & Geosciences, 37: 1277-1284. doi:10.1016/j.cageo.2010.07.005
Cannon, A.J., 2017. Non-crossing nonlinear regression quantiles by monotone composite quantile regression neural network, with application to rainfall extremes. EarthArXiv <https://eartharxiv.org/wg7sn>. doi:10.17605/OSF.IO/WG7SN