What should be the most appropriate parameter values for the MNIST dataset?

[ykocaturk]

data preparation

x_train <- as.matrix(train[,2:785] /255)

x_test <- as.matrix(test[,2:785] /255)
y_test <- test$y

y_train <- nnet::class.ind(train$y)
colnames(y_train) <- paste("val", colnames(y_train), sep = "")
f <- as.formula(paste(paste(colnames(y_train), collapse = "+"), " ~ ",
paste(colnames(x_train), collapse = "+")))