In AlfonsoRReyes/rDeepThought:

Source: https://cran.r-project.org/web/packages/kerasR/vignettes/introduction.html

# Building a model in Keras starts by constructing an empty Sequential model.
library(kerasR)
mod <- Sequential()

# add a dense layer to our model 
# set the number of input variables equal to 13
mod$add(Dense(units = 50, input_shape = 13))

#  add an activation defined by a rectified linear unit to the model
mod$add(Activation("relu"))

# add a dense layer with just a single neuron to serve as the output layer
mod$add(Dense(units = 1))

#  compile it before fitting its parameters or using it for prediction
keras_compile(mod,  loss = 'mse', optimizer = RMSprop())

boston <- load_boston_housing()

#  scale the data matrices
X_train <- scale(boston$X_train)
Y_train <- boston$Y_train
X_test <- scale(boston$X_test)
Y_test <- boston$Y_test

#  fit the model from this data
keras_fit(mod, X_train, Y_train,
          batch_size = 32, epochs = 200,
          verbose = 1, validation_split = 0.1)

# the model does not do particularly well probably due to over-fitting on such as small set.
pred <- keras_predict(mod, normalize(X_test))
sd(as.numeric(pred) - Y_test) / sd(Y_test)

AlfonsoRReyes/rDeepThought documentation built on May 3, 2019, 6:42 p.m.