How to apply the moving-average operation on the last layer of the neural network in Keras?
For instance, let\'s say, the expected output of the regression problem should