The lectures, after an introduction to the problem, discuss two fundamental sampling techniques, rejection and importance sampling, and then turn their focus to MCMC, introducing the required background, the Metropolis Hastings algorithms, Gibbs sampling, and finally discussing some diagnostics and an extension of MCMC, Hamiltonian Monte Carlo, heavily used nowadays in inference engines, e.g. Pyro. 

Introduction

https://youtu.be/KwscSt19VIA

Rejection Sampling

https://youtu.be/NmMbBqc78MQ

Importance Sampling 

https://youtu.be/oLi07pGwmvU

Markov Chains

https://youtu.be/B-oLwP8oVHs

Detailed Balance

https://youtu.be/oa5r55j031U

Metropolis-Hastings MCMC

https://youtu.be/CrGVfgby9ec

Gibbs sampling

https://youtu.be/kbP7ifl5S8Q

Convergence diagnostics

https://youtu.be/WajcxG54qRg

Hamiltonian Monte Carlo

https://youtu.be/GtwCPnmkzfI

Ultime modifiche: venerdì, 28 febbraio 2025, 16:46