I have a neural network that I constructed in
keras that goes from a LSTM recurrent layer > dropout > flattened > dense layer of 1 unit.
Does this make sense to have dropout regularization at this stage? Would this be creating sparsity in the penultimate layer or would it be evening out the connections to the final prediction?
bidirectional_1 (Bidirection (None, 60, 512) 1181696 _________________________________________________________________ dropout_2 (Dropout) (None, 60, 512) 0 _________________________________________________________________ flatten_1 (Flatten) (None, 30720) 0 _________________________________________________________________ dense_1 (Dense) (None, 1) 30721 =================================================================