I have been reading about dropout and understand what happens when training the network but I don't understand how it would work in calculating the network. Can someone please explain how the process differs from calculating the output of a neural network without dropout in?
As a final step to the training you just multiply all the weights by the probability that they were NOT dropped at each training step. Inference then proceeds as normal i.e. the same as for a network without dropout.
In other words, aside from this multiplication adjustment to the all the weights, dropout does not apply to the forward pass of the network.