GradientTape It records intermediate processes of functions, operations on Tape one by one. In other words, computational operation is saved in Tape. Reverse mode automatic differentiation TensorFlow calculates the differentiation of x for the loss by using the Reverse mode automatic differentiation with dx = tape.gradient(loss, x) Backpropagate the differential of x for the loss, repeat the tas..
