Hi everyone i made a very self-contained package to efficiently perform Approximate Bayesian Computation, this package implements a hybrid MCMC sampler with 3 different affine invariant proposals, solving many of the shortcomings of other MCMC algorithms (adaptive and non adaptive) this package com…

i just implemented a WIP version of ABC SMC - Differential Evolution, it is scary fast, give it a spin :wink: ABCDE is the method name

[image] francesco.alemanno: i just implemented a WIP version of ABC SMC - Differential Evolution, it is scary fast, give it a spin :wink: Nice! It is indeed very fast for me. With the same settings and the mixture model, it’s 10x faster than ABCSMCPR and 3x faster than ApproxBayes. If it w…

instead of a square accept/reject kernel (ie. distance stat within epsilon) can it handle smooth kernels?

[image] robsmith11: Are there any downsides relative to SMC of which to be careful? not that i know of :slight_smile: [image] dlakelan: instead of a square accept/reject kernel (ie. distance stat within epsilon) can it handle smooth kernels? at the moment no, as smooth kernels tend to…

I started a similar effort some time ago here: [image] GitHub - JuliaApproxInference/LikelihoodfreeInference.jl: Likelihood-Free... Likelihood-Free Inference for Julia. Contribute to JuliaApproxInference/LikelihoodfreeInference.jl development by creating an account on GitHub. …

i just implemented a WIP version of ABC SMC - Differential Evolution, it is scary fast, give it a spin :wink: ABCDE is the method name Wow, it is really fast. That’s awesome, thank you!

The package is now registered, have fun! and remember to let me know if you find issues or obvious limitations that might have gone unnoticed :smiley:

Thanks for the very interesting package :+1:. I have been playing around a bit with the mixture model in the documentation. I find that the confidence intervals seem to be too broad, the true parameters are inside more than 90% of the times it is repeated, in my limited experimentation. For example,…

Thank you for your feedback! :slight_smile: glad you find it interesting. ABC methods most of the time suffer from having wider CI’s than exact methods, since both the imperfect distance metric, and the non-zero tolerance tend to affect CI’s. i think the example can be improved by considering a Kol…

[ANN-RFC] KissABC.jl- Approximate Bayesian Computation

Package Announcements

francesco.alemanno June 4, 2020, 9:20am 16

@jbrea that would be nice! a full fledged powerful ABC package, i would also like to collaborate

Topic		Replies	Views
Extracting each iteration result in KissABC.jl New to Julia question , inference , modelling , kissabc	4	178	September 5, 2024
Whats going on in kissABC.jl package smc() function Machine Learning question , bayesian-inference , smc , approximate-bayesian , kissabc	5	277	September 6, 2024
Status of ABC-SMC samplers in Julia Statistics turing , bayesian-inference , monte-carlo , abc , smc	10	1193	October 6, 2023
Differential evolution MCMC in Julia? Probabilistic Programming question , turing , mcmc	18	1532	May 26, 2024
A simple MCMC code for beginners for learning Offtopic tutorials , mcmc	14	6692	March 31, 2023

[ANN-RFC] KissABC.jl- Approximate Bayesian Computation

Related topics