tuneHyperparams: Hyperparameter tuning used by NIMIWAE() function
In DavidKLim/NIMIWAE: Non-Ignorably Missing Importance Weighted Autoencoder

View source: R/tuneHyperparams.R

tuneHyperparams

R Documentation

Hyperparameter tuning used by NIMIWAE() function

Description

Hyperparameter tuning used by NIMIWAE() function

Usage

tuneHyperparams(
  FUN = NULL,
  method = "NIMIWAE",
  dataset,
  data,
  data_types,
  data_types_0,
  Missing,
  g,
  rdeponz = F,
  learn_r = T,
  phi0 = NULL,
  phi = NULL,
  Cs,
  ignorable = F,
  covars_r = rep(1, ncol(data)),
  arch = "IWAE",
  draw_xmiss = T,
  sigma = "elu",
  h = c(128L, 64L),
  n_hidden_layers = c(1L, 2L),
  n_hidden_layers_r0 = NULL,
  bs = 1000L,
  lr = c(0.001, 0.01),
  dim_z = as.integer(c(floor(ncol(data)/2), floor(ncol(data)/4))),
  niws = 5L,
  n_imputations = 5L,
  n_epochs = 2002L,
  data_types_HIVAE = NULL,
  one_hot_max_sizes = NULL,
  ohms = NULL,
  MissingDatas = NULL,
  save_imps = F,
  dir_name = ".",
  normalize = T
)

Arguments

`FUN`	run_<> function for other methods (not NIMIWAE)
`method`	String specifying the method to tune hyperparameters. Can be "NIMIWAE" (default), "MIWAE", "VAEAC", "HIVAE", "MEAN", or "MF". Hyperparameters are not tuned for "MEAN" or "MF"
`data`	Data matrix (N x P)
`data_types`	vector of length=ncol(data). Valid values: "real", "count", "cat" or "pos"
`Missing`	Missingness mask matrix (N x P)
`g`	Training-validation-test split partitioning
`rdeponz`	TRUE/FALSE: Whether to allow missingness (r) to depend on the latent variable (z). Default is FALSE
`learn_r`	TRUE/FALSE: Whether to learn missingness model via appended NN (TRUE, default), or fit a known logistic regression model (FALSE). If FALSE, 'phi0' and 'phi' must be specified
`phi0`	(optional) Intercept of logistic regression model, if learn_r = FALSE.
`phi`	(optional) Vector of coefficients of logistic regression model for each input covariates 'covars_r', if learn_r = FALSE. 'phi' must be the same length as the number of input covariates, or 'sum(covars_r)'.
`Cs`	(optional) # factors for categorical variables. Must be of length = # of categorical variables.
`ignorable`	TRUE/FALSE: Whether missingness is ignorable (MCAR/MAR) or nonignorable (MNAR, default). If missingness is known to be ignorable, "ignorable=T" omits missingness model.
`covars_r`	Vector of 1's and 0's of whether each feature is included as covariates in the missingness model. Need not be specified if 'ignorable = T'. Default is using all features as covariates in missingness model. Must be length P (or 'ncol(data)')
`arch`	Architecture of NIMIWAE. Can be "IWAE" or "VAE". "VAE" is specific case of the "IWAE" where only one sample is drawn from the joint posterior of (z, xm).
`sigma`	activation function ("relu" or "elu")
`h`	integer, number of nodes per hidden layer
`n_hidden_layers`	integer, #hidden layers (except missingness model Decoder_r)
`bs`	integer, batch size (training)
`lr`	float, learning rate
`dim_z`	integer, dimensionality of latent z. Default: 5
`data_types_HIVAE`	Specify for HIVAE only.
`one_hot_max_sizes`	Specify for VAEAC only.
`ohms`	Specify for VAEAC only.
`MissingDatas`	Specify for VAEAC only.
`n_hidden_layers_r`	integer, #hidden layers for Decoder_r (default: 0)
`niw`	integer, number of importance weights (samples drawn from each latent space). Default: 5
`n_epoch`	integer, maximum number of epochs (without early stop). Default: 2002