no code implementations • NeurIPS 2013 • Matthew Johnson, James Saunderson, Alan Willsky
Sampling inference methods are computationally difficult to scale for many models in part because global dependencies can reduce opportunities for parallel computation.
