Hi all, I’d like to use Turing.jl to sample from a custom likelihood function. I’ve seen this post here and the DynamicHMCexamples.jl package, which gives very nice tutorials for doing this using DynamicHMC, but I would also like to try this using the Turing.jl framework with the AdvancedHMC.jl bac…

Parts of the documentation are unfortunately not up to date. I’m working on improving it in the next weeks. You can use a custom multivariate distribution as you pointed out. Turing uses the Bijectors.jl package for transformations of constrained distribution. So you would need to use an existing t…

What you need to do is specify the lower and upper bounds of your parameters with your priors. I think you can simply use Uniform(lb, ub) in your use case. Here is a fairly straightforward example of specifying a custom likelihood distribution in Turing. Here is example code for running the linked …

[image] depasquale: Speaking of which, what is the difference between AdvancedHMC.jl and DynamicHCM.jl? I know there are differences in terms of the NUTS configuration. AdvancedHMC uses Stan’s default configuration, whereas DynaicHMC uses something else. From personal experience, I found Dyna…

AFACT DynamicHMC and AdvancedHMC use basically the same almost-current NUTS algorithm after the initial point and stepsize have been established. This algorithm is mostly what one would find in Stan, except for the recent (and somewhat controversial, at least from a procedural sense) updates about a…

I’m not sure I understand @depasquale properly. I want to sample from a custom multidimensional likelihood function, with no prior on the parameters When you say you want to “sample from a custom multidimensional likelihood”, do you mean you want to sample the “observations” given some parameter…

Custom likelihood with Turing.jl

Specific Domains Probabilistic Programming

trappmartin December 7, 2019, 4:34am 7

Here is a feature overview of different NUTS implementations and AdvancedHMC. There are more details in Kai’s AABI paper as well as various empirical evaluations.

Topic		Replies	Views
Automatic transformation of constrained distributions in Turing.jl General Usage question , turing	0	155	June 14, 2023
Custom Models For Bayesian Inference Probabilistic Programming	4	438	December 5, 2023
Custom likelihoods in Turing.jl General Usage	15	3835	October 26, 2018
Constrained Priors and Bijectors Probabilistic Programming	7	461	November 12, 2022
Question on Bayesian Inference with Julia, using Turing and/or AdvancedHMC and/or Distributions Statistics question	11	2006	March 11, 2020

Custom likelihood with Turing.jl

Related topics