Getting the current sample variable value so it can be used to compute the observed inputs in a secondary model likelihood

[Vital]

Hello @junpenglao ! Thank you very much for your reply!

I have not been able to make that work because I need a function which can integrate under a theano vector.

However, in this post you shared there are people doing the same :

That approach works: I can can integrate the specially if I apply an interpolation on the propopsed

Using quad from scipy.integrate with pymc RVs as parameter

I’m trying to write a custom random variable, in which I need to do numerical integration with pymc RVs as parameter. An example is stated as below: #define some pymc RVs a1=pm.Normal(‘a1’,mu=0,sd=1) a2=pm.Nromal(‘a2’,mu=0,sd=1) #the function to be integrated def f(x,p): retrun tt.power(x,p[0]*p[1]) scipy.integrate.quad(f,0,1,args=[a1,a2]) However, quad would return an error since a1,a2 are not real number, and I can’t find something like a1.value() to get the current value of the random…

This seems like and old post: Maybe you know if someone has shared a working pymc3 code with a integration op for a y = f(x) function?