Should the dropout be removed after the neural network has trained?

Question

Neural networks

dropout is set in order to randomly turn off a certain number of neurons, thereby avoiding retraining. Do I need to remove the dropout after the neural network has trained?

Answer 1

The question is not entirely clear. What do you mean by "remove dropout"?
Dropout does not need to be applied to an already trained neural network, if you are talking about it.

Answer 2

Usually, the network is put into inference mode, for a dropout, this means multiplying the values ​​of all outputs by (1-p) instead of zeroing p outputs.