model $=$ keras.models.Sequential$()$

model.add$($keras$.$layers.Flatten$($input$\_$shape$=[32,32,3]))$ #for input layer

for $\_$ in range$(20)$: #for hidden layer

model.add$($keras$.$layers.Dense$($

$100,$

activation$=$'elu',

kernel$\_$initializer$=\text{'}$he$\_$normal'

$))$

model.add$($keras$.$layers.Dense$(10,$ activation$=$'softmax'$))$

optimizer $=$ keras.optimizers.Nadam$($learning$\_$rate$=5$e$-5)$

model.compile$($

loss$=$'sparse$\_$categorical$\_$crossentropy',

optimizer$=$optimizer,

metrics$=[\text{'}$accuracy$\text{'}])$

early$\_$stopping$\_$cb $=$ keras.callbacks.EarlyStopping$($patience$=4)$

history$=$ model.fit$($X$\_$train, y$\_$train, epochs$=30,$

validation$\_$data$=($X$\_$valid, y$\_$valid$),$

callbacks$=[$early$\_$stopping$\_$cb$])$

Please on the above code can you add Batch normalization layer after or before each hidden layer $\text{'}$s activation function $?$