Lost Relatives of the Gumbel Trick

Conference Object
Change log
Tripuraneni, N 

The Gumbel trick is a method to sample from a discrete probability distribution, or to estimate its normalizing partition function. The method re- lies on repeatedly applying a random perturbation to the distribution in a particular way, each time solving for the most likely configuration. We derive an entire family of related methods, of which the Gumbel trick is one member, and show that the new methods have superior properties in several settings with minimal additional computational cost. In particular, for the Gumbel trick to yield computational benefits for discrete graphical models, Gumbel perturbations on all configurations are typically replaced with so- called low-rank perturbations. We show how a subfamily of our new methods adapts to this set- ting, proving new upper and lower bounds on the log partition function and deriving a family of sequential samplers for the Gibbs distribution. Finally, we balance the discussion by showing how the simpler analytical form of the Gumbel trick enables additional theoretical results.

machine learning, statistics, partition function, Gumbel trick, ICML
Journal Title
ICML'17 Proceedings of the 34th International Conference on Machine Learning
Conference Name
ICML 2017
Journal ISSN
Volume Title
Alan Turing Institute under EPSRC grant EP/N510129/1, and by the Leverhulme Trust via the CFI.