qpaths_uai_2021
None
view repo
Many common machine learning methods involve the geometric annealing path, a sequence of intermediate densities between two distributions of interest constructed using the geometric average. While alternatives such as the moment-averaging path have demonstrated performance gains in some settings, their practical applicability remains limited by exponential family endpoint assumptions and a lack of closed form energy function. In this work, we introduce q-paths, a family of paths which is derived from a generalized notion of the mean, includes the geometric and arithmetic mixtures as special cases, and admits a simple closed form involving the deformed logarithm function from nonextensive thermodynamics. Following previous analysis of the geometric path, we interpret our q-paths as corresponding to a q-exponential family of distributions, and provide a variational representation of intermediate densities as minimizing a mixture of α-divergences to the endpoints. We show that small deviations away from the geometric path yield empirical gains for Bayesian inference using Sequential Monte Carlo and generative model evaluation using Annealed Importance Sampling.
READ FULL TEXT
Annealed importance sampling (AIS) is the gold standard for estimating
p...
read it
Parallel tempering (PT) is a class of Markov chain Monte Carlo algorithm...
read it
The exponential family is well known in machine learning and statistical...
read it
It is well-known that the Bhattacharyya, Hellinger, Kullback-Leibler,
α-...
read it
The informational energy of Onicescu is a positive quantity that measure...
read it
The recently proposed Thermodynamic Variational Objective (TVO) leverage...
read it
Collaborative uploading describes a type of crowdsourcing scenario in
ne...
read it
None
Comments
There are no comments yet.