Fig. 4

Percentage of synaptic weights of the three inputs in the neural network approach (model 1): f _struct/f _soil, velocity contrast and PGA for the five typology classes as defined by Lagomarsino and Giovinazzi (2006)