Calibrating a Deep Neural Network with Its Predecessors | Read Paper on Bytez