HarlMCMC shake




Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Mar 7, 2013

Two short animations illustrate the differences between a Metropolis-Hastings (MH) sampler and a Hamiltonian Monte Carlo (HMC) sampler.

We test the samplers on nine mixtures of Gaussians, ranging from a mixture of four elongated Gaussians in the upper right-hand corner, to a single spherical Gaussian in the middle, to a mixture of 100 spherical Gaussians in the lower lefthand corner. We plot the contours of these distributions in white and the last 10 samples in red.

In the MH case, the samples are connected in sequential order by yellow lines. In the HMC case, the samples are connected (in yellow) by the Hamiltonian evolution sequence generated during the proposal of the next (accepted) sample. The MH proposals are set to spherical Gaussians with relatively small variance in the outer eight target Gaussian mixtures and relatively large variance for the central spherical Gaussian target distribution.

Tamara Broderick (UC Berkeley)

David Duvenaud (U Cambridge)

Code available at:

Happy International Year of Statistics!

  • Category

  • Song

  • Artist

  • Licensed to YouTube by

    • BBTV_MadDecent (on behalf of Jeffree/ Mad Decent); AMRA, CMRRA, UMPG Publishing, LatinAutor - UMPG, UMPI, LatinAutor, UBEM, and 17 Music Rights Societies


When autoplay is enabled, a suggested video will automatically play next.

Up next

to add this to Watch Later

Add to

Loading playlists...