Sciweavers

CACM
2011

The sequence memoizer

12 years 10 months ago
The sequence memoizer
We propose an unbounded-depth, hierarchical, Bayesian nonparametric model for discrete sequence data. This model can be estimated from a single training sequence, yet shares statistical strength between subsequent symbol predictive distributions in such a way that predictive performance generalizes well. The model builds on a specific parameterization of an unbounded-depth hierarchical Pitman-Yor process. We introduce analytic marginalization steps (using coagulation operators) to reduce this model to one that can be represented in time and space linear in the length of the training sequence. We show how to perform inference in such a model without truncation approximation and introduce fragmentation operators necessary to do predictive inference. We demonstrate the sequence memoizer by using it as a language model, achieving state-of-the-art results.
Frank Wood, Jan Gasthaus, Cédric Archambeau
Added 12 May 2011
Updated 12 May 2011
Type Journal
Year 2011
Where CACM
Authors Frank Wood, Jan Gasthaus, Cédric Archambeau, Lancelot James, Yee Whye Teh
Comments (0)