AutoKeras Image Regressor Model

Source: R/model_image_regressor.R

AutoKeras image regression class.
It is used for image regression. It searches convolutional neural network architectures for the best configuration for the image dataset. To `fit`, `evaluate` or `predict`, format inputs as:

  • x : array. The shape of the data should be 3 or 4 dimensional, the last dimension of which should be channel dimension.

  • y : array. The targets passing to the head would have to be array or data.frame. It can be single-column or multi-column. The values should all be numerical.

model_image_regressor(
  output_dim = NULL,
  loss = "mean_squared_error",
  metrics = NULL,
  name = "image_regressor",
  max_trials = 100,
  directory = tempdir(),
  objective = "val_loss",
  overwrite = TRUE,
  seed = runif(1, 0, 1e+07)
)

Arguments

output_dim

: numeric. The number of output dimensions. Defaults to `NULL`. If `NULL`, it will infer from the data.
loss

: A Keras loss function. Defaults to use "mean_squared_error".
metrics

: A list of Keras metrics. Defaults to use "mean_squared_error".
name

: character. The name of the AutoModel. Defaults to "image_regressor".
max_trials

: numeric. The maximum number of different Keras Models to try. The search may finish before reaching the `max_trials`. Defaults to `100`.
directory

: character. The path to a directory for storing the search outputs. Defaults to `tempdir()`, which would create a folder with the name of the AutoModel in the current directory.
objective

: character. Name of model metric to minimize or maximize, e.g. "val_accuracy". Defaults to "val_loss".
overwrite

: logical. Defaults to `TRUE`. If `FALSE`, reloads an existing project of the same name if one is found. Otherwise, overwrites the project.
seed

: numeric. Random seed. Defaults to `runif(1, 0, 10e6)`.

Value

A non-trained image regressor AutokerasModel.

Details

Important: The object returned by this function behaves like an R6 object, i.e., within function calls with this object as parameter, it is most likely that the object will be modified. Therefore it is not necessary to assign the result of the functions to the same object.

Examples

if (FALSE) {
library("keras")

# use the MNIST dataset as an example
mnist <- dataset_mnist()
c(x_train, y_train) %<-% mnist$train
c(x_test, y_test) %<-% mnist$test

library("autokeras")

# Initialize the image regressor
reg <- model_image_regressor(max_trials = 10) %>% # It tries 10 different models
  fit(x_train, y_train) # Feed the image regressor with training data

# If you want to use own valitadion data do:
reg <- model_image_regressor(max_trials = 10) %>%
  fit(
    x_train,
    y_train,
    validation_data = list(x_test, y_test)
  )

# Predict with the best model
(predicted_y <- reg %>% predict(x_test))

# Evaluate the best model with testing data
reg %>% evaluate(x_test, y_test)

# Get the best trained Keras model, to work with the keras R library
export_model(reg)
}