deeptransformation_init: Initializing Deep Transformation Models
In davidruegamer/deepregression: Fitting Deep Distributional Regression

deeptransformation_init

R Documentation

Initializing Deep Transformation Models

Description

Initializing Deep Transformation Models

Usage

deeptransformation_init(
  n_obs,
  ncol_structured,
  ncol_deep,
  list_structured,
  list_deep,
  lambda_lasso = NULL,
  lambda_ridge = NULL,
  weights = NULL,
  learning_rate = 0.01,
  optimizer = optimizer_adam(lr = learning_rate),
  monitor_metric = list(),
  orthog_fun = orthog,
  orthogX = NULL,
  split_fun = split_model,
  order_bsp,
  use_bias_in_structured = FALSE,
  train_together = NULL,
  split_between_shift_and_theta = NULL,
  interact_pred_trafo = NULL,
  addconst_interaction = NULL,
  penalize_bsp = 0,
  order_bsp_penalty = 2,
  base_distribution = "normal",
  batch_shape = NULL,
  atm_lags = 0,
  atm_toplayer = NULL
)

Arguments

`n_obs`	number of observations
`ncol_structured`	a vector of length #parameters defining the number of variables used for each of the parameters. If any of the parameters is not modelled using a structured part the corresponding entry must be zero.
`ncol_deep`	a vector of length #parameters defining the number of variables used for each of the parameters. If any of the parameters is not modelled using a deep part the corresponding entry must be zero. If all parameters are estimated by the same deep model, the first entry must be non-zero while the others must be zero.
`list_structured`	list of (non-linear) structured layers (list length between 0 and number of parameters)
`list_deep`	list of deep models to be used (list length between 0 and number of parameters)
`lambda_lasso`	penalty parameter for l1 penalty of structured layers
`lambda_ridge`	penalty parameter for l2 penalty of structured layers
`weights`	observation weights used in the likelihood
`learning_rate`	learning rate for optimizer
`optimizer`	optimizer used (defaults to adam)
`monitor_metric`	see `?deepregression`
`orthog_fun`	function defining the orthogonalization
`orthogX`	vector of columns defining the orthgonalization layer
`split_fun`	see `?deepregression`
`order_bsp`	NULL or integer; order of Bernstein polynomials; if not NULL, a conditional transformation model (CTM) is fitted.
`use_bias_in_structured`	whether or not to use a bias in structured layers
`train_together`	see `?deepregression`
`split_between_shift_and_theta`	see `?deepregression`
`interact_pred_trafo`	specifies a transformation function applied to the interaction predictor using a layer lambda (e.g. to ensure positivity)
`addconst_interaction`	additive constant added to predictor matrix to ensure positivity
`penalize_bsp`	scalar value > 0; amount of penalization of Bernstein polynomials
`order_bsp_penalty`	integer; order of Bernstein polynomial penalty. 0 results in a penalty based on integrated squared second order derivatives, values >= 1 in difference penalties
`base_distribution`	a string ("normal", "logistic") or TFP distribution; the base distribution for the transformation model. Per default `tfd_normal(loc = 0, scale = 1)` but any other distribution is possible (e.g., `tfd_logistic(loc = 0, scale = 1)`).