Jethro's Braindump

Slice Sampling

Slice sampling has similarities to Gibbs sampling, Rejection Sampling and the Metropolis-Hastings method. The advantage over simple Metropolis-Hastings methods is that it is robust to the choice of parameters. Like rejection sampling, it asymptotically draws samples from volume under the curve \(p^\star(x)\).

A good treatment of slice sampling can be found in Mackay’s Book Information Theory, Inference and Learning Algorithms.