Keras Weight Delta in training batch

Question

I am pretty new to Machine Learning with Tensorflow and Keras.

I am trying to implement the Amnesiac Machine Unlearning approach as described in http://hdl.handle.net