n: Neural Networks for BAMLSS
In bamlss: Bayesian Additive Models for Location, Scale, and Shape (and Beyond)

n	R Documentation

Neural Networks for BAMLSS

Description

This smooth constructor implements single hidden layer neural networks.

Usage

## The neural network smooth constructor.
n(..., k = 10, type = 2)

## Initialize weights.
n.weights(nodes, k, r = NULL, s = NULL,
  type = c("sigmoid", "gauss", "softplus", "cos", "sin"),
  x = NULL, ...)

## Second weights initializer, internally calls n.weights.
make_weights(object, data, dropout = 0.2)

## Boosted neural net predictions.
predictn(object, newdata, model = NULL,
  mstop = NULL, type = c("link", "parameter"))

Arguments

`...`	For function `n()` a formula of the type `~x1+x2+x3` that specifies the covariates that should be modeled by the neural network. For function `predictn()`, arguments to be passed to `predict.bamlss`.
`k`	For function `n()`, the number of hidden nodes of the network. Note that one can set an argument `split = TRUE` to split up the neural network into, e.g., `nsplit = 5` parts with `k` nodes each. For function `n.weights()`, argument `k` is the number of input variables of the network (number of covariates).
`type`	Integer. Type `1` fits a complete network in each boosting iteration, `type = 2` selects the best fitting node in each boosting iteration. for function `n.weights()`, the type of activation function that should be used. For function `predictn()`, the type of prediction that should be computed.
`nodes`	Number of nodes for each layer, i.e., can also be a vector.
`r`, `s`	Parameters controlling the shape of the activation functions.
`x`	A scaled covariate matrix, the data will be used to identify the range of the weights.
`object`, `data`	See `smooth.construct`. For function `predictn()`, a boosted `"bamlss"` object.
`dropout`	The fraction of inner weights that should be set to zero.
`newdata`	The data frame that should be used for prediction.
`model`	For which parameter of the distribution predictions should be computed.
`mstop`	The stopping iteration for which predictions should be computed. The default is to return a matrix of predictions, each column represents the prediction of one boosting iteration.