I'm studying the Fully Convolutional Networks right now and when it's clear that the receptive field is not dependent on the input size (the whole network in a way is independent from the input size), I'm not sure how the offset coming from the receptive field is handled.


In the above model from the WaveNet paper, for one output node, four input nodes have to be present. What happens if only one input node is passed to the network? Should the data be zero padded to compensate the receptive field? or does this architecture have a build-in mechanism to cope with that?

The goal is a situation where the number of inputs is the same as the number of outputs



You must log in to answer this question.