make_generator_from_dataframe: Make a DataGenerator from a data.frame or data.table
In mlr-org/mlr3keras: mlr3 Keras extension

make_generator_from_dataframe

R Documentation

Make a DataGenerator from a data.frame or data.table

Description

Creates a Python Class that internally iterates over the data.

Usage

make_generator_from_dataframe(
  dt,
  x_cols = NULL,
  y_cols,
  x_transform = NULL,
  y_transform = NULL,
  generator = keras::image_data_generator(),
  batch_size = 32L,
  shuffle = TRUE,
  seed = 1L,
  y_cols_to_categorical = TRUE,
  subset = NULL,
  ignore_class_split = FALSE
)

Arguments

`dt`	`data.frame`\|`data.table` Data container to iterate over.
`x_cols`	`character` Names of features to be used. Defaults to all but y_cols.
`y_cols`	`character` Target variable. Automatically converted to one-hot if "y_cols_to_categorical" is TRUE.
`x_transform`	`function` Function used to transform data to a keras input format for features.
`y_transform`	`function` Function used to transform data to a keras input format for the response.
`generator`	`Python Object` A generator as e.g. obtained from `keras::image_data_generator`. Used for consistent train-test splits.
`batch_size`	`integer` Batch size.
`shuffle`	`logical` Should data be shuffled?
`seed`	`integer` Set a seed for shuffling data.
`y_cols_to_categorical`	`logical` Should target be converted to one-hot representation?
`subset`	`character` Should samples be generated from 'training' or 'validation' set? Only applicable together with a 'generator'.
`ignore_class_split`	`logical` Test whether all class labels appear in all splits.