MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning
Machine Learning
2019-06-05 v1 Machine Learning
Abstract
Efficient approximation lies at the heart of large-scale machine learning problems. In this paper, we propose a novel, robust maximum entropy algorithm, which is capable of dealing with hundreds of moments and allows for computationally efficient approximations. We showcase the usefulness of the proposed method, its equivalence to constrained Bayesian variational inference and demonstrate its superiority over existing approaches in two applications, namely, fast log determinant estimation and information-theoretic Bayesian optimisation.
Cite
@article{arxiv.1906.01101,
title = {MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning},
author = {Diego Granziol and Binxin Ru and Stefan Zohren and Xiaowen Doing and Michael Osborne and Stephen Roberts},
journal= {arXiv preprint arXiv:1906.01101},
year = {2019}
}
Comments
18 pages, 3 figures, Published at Entropy 2019: Special Issue Entropy Based Inference and Optimization in Machine Learning