Packages

package algorithms

Ordering
  1. Alphabetic
Visibility
  1. Public
  2. Protected

Type Members

  1. class ForwardChain[A] extends MarkovChain[A]

    Markov chain which draws the next sample from a proposal distribution

  2. class Metropolis[A] extends MarkovChain[A]

    Metropolis algorithm (MCMC), provides samples from the evaluator distribution by drawing from generator and stochastic accept/reject decisions generator needs to be symmetric

  3. class MetropolisHastings[A] extends MarkovChain[A]

    Metropolis-Hastings algorithm - generates random samples from a target distribution using only samples from a proposal distribution

  4. class MetropolisHastingsWithPrefetching[A] extends MarkovChain[A]

    Metropolis-Hastings algorithm with speculative executing - generates random samples from a target distribution using only samples from a proposal distribution

    Metropolis-Hastings algorithm with speculative executing - generates random samples from a target distribution using only samples from a proposal distribution

    The prefetching strategy precomputes (part of) the tree of possible future states and evaluates the states in parallel. If proposals are cheap to compute but evaluation expensive (and single threaded), this strategy might lead to a speed improvement, especially when the number of acceptance is quite low.

    See for example Accelerating Metropolis–Hastings algorithms: Delayed acceptance with prefetching M. Banterle, C. Grazian, C .Robert https://arxiv.org/pdf/1406.2660.pdf for a good discussion of the idea and possibilities for further improvements.

Ungrouped