a, b and c represents the parameters of GULP that you want to estimate, and yeah PyMC3 will sample those parameters. Let suppose you were interesting in computing the Young’s modulus and all that you have is a set of experimental measurement of the Young’s modulus and a parameterized model that allows you to compute the Young’s modulus, SMC-ABC will let you find the parameters that make the theoretical values of the Young’s modulus come closer to the experimental ones. In the previous example results is the theoretical Young’s modulus and observed array is the experimental values. Notice that the values in observed array should not necessary be from an real experiment they can be from a simulation or some other source, by they are basically the data you want mimic with your simulator.