Notice
Recent Posts
Recent Comments
Link
관리 메뉴

뛰는 놈 위에 나는 공대생

[Python] MCMC Sampling library 본문

프로그래밍 Programming/파이썬 Python

[Python] MCMC Sampling library

보통의공대생 2023. 8. 24. 10:54

MCMC sampling을 위한 라이브러리

 




marathon : c++ 라이브러리

c++ 라이브러리는 고려하지 않았으므로 패스

 




Pyro : Pyro is a universal probabilistic programming language (PPL) written in Python and supported by PyTorch on the backend. Pyro enables flexible and expressive deep probabilistic modeling, unifying the best of modern deep learning and Bayesian modeling. It was designed with these key principles

Universal: Pyro can represent any computable probability distribution.
Scalable: Pyro scales to large data sets with little overhead.
Minimal: Pyro is implemented with a small core of powerful, composable abstractions.
Flexible: Pyro aims for automation when you want it, control when you need it.

 

- 살펴보면 Variational inference 기반임



PyMC : PyMC (formerly PyMC3) is a Python package for Bayesian statistical modeling focusing on advanced Markov chain Monte Carlo (MCMC) and variational inference (VI) algorithms. Its flexibility and extensibility make it applicable to a large suite of problems. 

 

https://github.com/pymc-devs/pymc

 

GitHub - pymc-devs/pymc: Bayesian Modeling in Python

Bayesian Modeling in Python. Contribute to pymc-devs/pymc development by creating an account on GitHub.

github.com

 

 


Sampyl : Sampyl is a Python library implementing Markov Chain Monte Carlo (MCMC) samplers in Python. It’s designed for use in Bayesian parameter estimation and provides a collection of distribution log-likelihoods for use in constructing models. 

 

코드 자체가 이해하기 쉬운 편

 

https://github.com/mcleonard/sampyl

 

GitHub - mcleonard/sampyl: MCMC samplers for Bayesian estimation in Python, including Metropolis-Hastings, NUTS, and Slice

MCMC samplers for Bayesian estimation in Python, including Metropolis-Hastings, NUTS, and Slice - GitHub - mcleonard/sampyl: MCMC samplers for Bayesian estimation in Python, including Metropolis-Ha...

github.com

 

Comments