karas_experiment.R

devtools::install_github("rstudio/keras")

library(keras)
install_keras()

# load MNIST data
mnist <- dataset_mnist()
x_train <- mnist$train$x
y_train <- mnist$train$y
x_test <- mnist$test$x
y_test <- mnist$test$y

# reshape
dim(x_train) <- c(nrow(x_train), 784)
dim(x_test) <- c(nrow(x_test), 784)
# rescale
x_train <- x_train / 255
x_test <- x_test / 255
# encode
y_train <- to_categorical(y_train, 10)
y_test <- to_categorical(y_test, 10)

# make a model
model <- keras_model_sequential()
model %>%
  layer_dense(units = 256,
              activation = "relu",
              input_shape = c(784)) %>%
  layer_dropout(rate = 0.4) %>%
  layer_dense(units = 128, activation = "relu") %>%
  layer_dropout(rate = 0.3) %>%
  layer_dense(units = 10, activation = "softmax")

model %>% compile(
  loss = 'categorical_crossentropy',
  optimizer = optimizer_rmsprop(),
  metrics = c('accuracy')
)

history <- model %>% fit(
  x_train,
  y_train,
  epochs = 30,
  batch_size = 128,
  validation_split = 0.2
)

plot(history)

model %>% evaluate(x_test, y_test)

y_predicted <- model %>% predict_classes(x_test)

# deal with mismatches

wrong_actual_y <- mnist$test$y[mnist$test$y != y_predicted] 
wrong_actual_x <- mnist$test$x[mnist$test$y != y_predicted,,] 
wrong_pred_y <- y_predicted[mnist$test$y != y_predicted]

i <- 6
digit <- wrong_actual_x[i,,] # 
plot(as.raster(digit, max=255))
print(wrong_actual_y[i])
print(wrong_pred_y[i])