When there are several outputs of custom model, should i need to rescale the scales of outputs to put them in the same regions? And how to set the hard boundary of ϵ for each parameter. (The criterion of sample of parameters from a distribution is π(θ|d(y0,y∗))⩽ϵ.)
For instance:
def model_0D(theta, args, **kwds):
para = theta
Y1 = 2.8405e-4para[3]**2para[0]**1.6para[1]*para[2]
Y2 = …
Y3 = …
return Y1, Y2, Y3
‘’’’’’’’’
Sim = pm.Simulator(‘sim’, model_0D, params=([Ti, Vhs, tau, P]), epsilon=np.array([2.7e3,0.12,1.1]), observed=np.array([98.4e3,4.6,10.1]))
trace = pm.sample_smc(kernel=“ABC”, parallel=False)