Improved MCMC Algorithm

Amos 7.0 implements Hamiltonian MCMC as an alternative to the random walk Metropolis algorithm. There are some limits on when Hamiltonian MCMC can be used. (Search for "Hamiltonian" in the online help for more information and for some references.) Hamiltonian MCMC is often substantially faster than random walk Metropolis for numerical data.

Automatic MCMC Adaptation

Amos 7.0 provides a new Adapt button that adjusts the parameters of the MCMC algorithm automatically. The Adapt button analyzes the history of MCMC observations and updates the parameters of the MCMC algorithm (either random walk Metropolis or Hamiltonian) to improve its performance.