In common ANN, nodes are based on a activation function, like a sigmoid or tanh. The update of synaptic weight is based on the derivative of this function. when node are represented with spiking neurons,  what does activation function become ? how update of synaptic weight are done ?