The Nature paper is open access, looking into it. It seems just adding another layer of activation function. Input to hidden layer 1 is called synaptic and Hidd

The Nature paper is open access, looking into it. It seems just adding another layer of activation function. Input to hidden layer 1 is called synaptic and Hidden layer 1 to output is called dendritic. Can someone explain what how the update rules, are different from backprop?