R语言的MNIST预测

# Install and load the required package  
# install.packages("keras")  
library(keras)  

# Function to get data  
get_data <- function() {  
  mnist <- dataset_mnist()  
  x_test <- mnist$test$x  
  y_test <- mnist$test$y  
  
  # Reshape and normalize  
  x_test <- array_reshape(x_test, c(nrow(x_test), 28 * 28)) / 255  
  return(list(x_test = x_test, y_test = y_test))  
}  

# Sigmoid activation function  
sigmoid <- function(x) {  
  return(1 / (1 + exp(-x)))  
}  

# Softmax activation function  
softmax <- function(x) {  
  exp_x <- exp(x - max(x))  # Prevent overflow  
  return(exp_x / rowSums(exp_x))  
}  

# Predict function  
predict <- function(network, x) {  
  W1 <- network$W1  
  W2 <- network$W2  
  W3 <- network$W3  
  b1 <- network$b1  
  b2 <- network$b2  
  b3 <- network$b3  
  
  a1 <- x %*% W1 + b1  
  z1 <- sigmoid(a1)  
  a2 <- z1 %*% W2 + b2  
  z2 <- sigmoid(a2)  
  a3 <- z2 %*% W3 + b3  
  y <- softmax(a3)  
  
  return(y)  
}  

# Initialize network  
initialize_network <- function(input_size, hidden_size1, hidden_size2, output_size) {  
  network <- list()  
  network$W1 <- matrix(rnorm(input_size * hidden_size1), nrow = input_size) * 0.01  
  network$b1 <- matrix(0, nrow = 1, ncol = hidden_size1)  
  network$W2 <- matrix(rnorm(hidden_size1 * hidden_size2), nrow = hidden_size1) * 0.01  
  network$b2 <- matrix(0, nrow = 1, ncol = hidden_size2)  
  network$W3 <- matrix(rnorm(hidden_size2 * output_size), nrow = hidden_size2) * 0.01  
  network$b3 <- matrix(0, nrow = 1, ncol = output_size)  
  return(network)  
}  

# Main program  
data <- get_data()  # Get test data and labels  
x <- data$x_test  
y_true <- data$y_test  
network <- initialize_network(28 * 28, 128, 64, 10)  # Initialize network  

accuracy_cnt <- 0  
for (i in 1:nrow(x)) {  
  y <- predict(network, x[i, , drop = FALSE])  # Keep input shape  
  p <- which.max(y) - 1  # R indices start at 1  
  if (p == y_true[i]) {  
    accuracy_cnt <- accuracy_cnt + 1  
  }  
}  

accuracy <- accuracy_cnt / nrow(x)  
cat(sprintf("Accuracy: %.2f%%\n", accuracy * 100))