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