I’m trying to create a custom likelihood function similar to the one described here : P(d|m) \propto exp \frac{-||d-f(m)||^2}{\sigma^2} where again d is observed data and m are parameters. In my case however, f involves a numerical integration, something like f(\lambda, m) \propto \int{\eta(\lambda)…

Implementing something like tt.trapz() for numerical integration over a set domain?

junpenglao August 27, 2019, 6:37am 4

Oh right. I think you might be able to change the quad integrator in the post with scipy.integrate.trapz - maybe you can give this a try?

Topic		Replies	Views
Custom theano Op to do numerical integration Questions theano	28	9073	January 18, 2023
PyMC3 : Using a parameter as a limit of integral Questions	6	2859	July 1, 2018
Using quad from scipy.integrate with pymc RVs as parameter Questions	7	2192	November 30, 2017
Implementing a more sophisticated pymc3 model v3 theano , modeling	8	805	June 11, 2022
Creating Theano Op to work on vectors element-wise within Model()? Questions	5	2468	August 30, 2017