Hands-On Machine Learning for Algorithmic Trading

Gibbs sampling

Gibbs sampling simplifies multivariate sampling to a sequence of one-dimensional draws. From a starting point, it iteratively holds n-1 variables constant while sampling the n^th variable. It incorporates this sample and repeats.

The algorithm is very simple and easy to implement but produces highly correlated samples that slow down convergence. Its sequential nature also prevents parallelization.