Experiment.R File

# Flags
FLAGS <- flags(flag_integer('dense_units', 32))

# Model
model <- keras_model_sequential() %>% 
         layer_dense(units = FLAGS$dense_units, 
                     activation = 'relu',
                     input_shape = c(21)) %>% 
         layer_dropout(rate = 0.1) %>% 
         layer_dense(units = 16,
                     activation = 'relu') %>% 
         layer_dropout(rate = 0.1) %>% 
         layer_dense(units = 3, activation = 'softmax')
# Compile
model %>% compile(loss = 'categorical_crossentropy',
                  optimizer = 'adam',
                  metrics = 'accuracy')
# Fit
history <- model %>% 
         fit(training,
             trainLabels,
             epochs = 50,
             batch_size = 32,
             validation_split = 0.2)