We wanna build a DNN model to predict unrolling factor though our features represent variable length of inputs. Knowing that we have to give our features at once "0 padding" look like the only solution that may solve our problem.
Many people suggest to use a masking layer so that padded values are ignored.
I have find some posts talking about using 0 padding in RNN, though I do not know if it's applied same way in case of MLPs.
How does it work masking layer with 0 padding in case of MLPs? f What are advantages of disadvantages of this solution ? In which cases this solution is useful and when it is not ?