nprophet_params: Tuning Parameters for Neural Prophet Models

In AlbertoAlmuinha/neuralprophet: 'Neural Prophet' Algorithm

Description

Tuning Parameters for Neural Prophet Models

Usage

trend_reg(range = c(0, 100), trans = NULL)

trend_reg_threshold(range = c(0, 10), trans = NULL)

num_hidden_layers(range = c(0L, 10L), trans = NULL)

d_hidden(range = c(0L, 500L), trans = NULL)

ar_sparsity(range = c(0, 1), trans = NULL)

Arguments

range	A two-element vector holding the defaults for the smallest and largest possible values, respectively.
trans	A trans object from the scales package, such as scales::log10_trans() or scales::reciprocal_trans(). If not provided, the default is used which matches the units used in range. If no transformation, NULL.

Details

The main parameters for Neural Prophet models are:

• trend_reg: the trend rate changes can be regularized by setting trend_reg to a value greater zero. This is a useful feature that can be used to automatically detect relevant changepoints.

• trend_reg_threshold: Threshold for the trend regularization

• num_hidden_layers: num_hidden_layers defines the number of hidden layers of the FFNNs used in the overall model.

• d_hidden: d_hidden is the number of units in the hidden layers.

• ar_sparsity: For ar_sparsity values in the range 0-1 are expected with 0 inducing complete sparsity and 1 imposing no regularization at all

Examples

trend_reg()

num_hidden_layers()

ar_sparsity()

AlbertoAlmuinha/neuralprophet documentation built on Dec. 17, 2021, 7:47 a.m.