🚊 Metropolis-Hastings

Metropolis-Hastings is a 🎯 Markov Chain Monte Carlo algorithm for approximate sampling from distribution if we have access to the unnormalized distribution . Our transition function works in two steps.

1. Propose from using some distribution .
2. Choose to accept this change with probability

Intuitively, our transition is picking some sample that’s near . If it moves in the direction of greater probability, we have a greater chance of accepting it. Thus, our samples gradually get closer to high probability, low energy areas, and begin to model the target distribution.