Exponential Stochastic Cellular Automata for Massively Parallel Inference
Manzil Zaheer, Michael Wick, Jean-Baptiste Tristan, Alexander Johannes Smola, Guy Lewis Steele · International Conference on Artificial Intelligence and Statistics · 2016
We propose an embarrassingly parallel, memory ecient inference algorithm for latent variable models in which the complete data likelihood is in the exponential family. The algorithm is a stochastic cellular automaton and converges to a valid maximum a posteriori fixed point. Applied to latent Dirichlet allocation we find that our algorithm is over an order or magnitude faster than the fastest current approaches. A simple C++/MPI implementation on a 20-node Amazon EC2 cluster samples at more than 1 billion tokens per second. We process 3 billion documents and achieve predictive power competitive with collapsed Gibbs sampling and variational inference.