The class implements a layer that subtracts its second input from the first.
This layer has no settings.
This layer has no trainable parameters.
The layer should have at least two inputs with float or int data. The blobs of all inputs should have the same dimensions.
The layer has one output of the same size as each of the inputs. The output blob contains the element-wise sub of the input blobs.