Frederik Eaton's Ph.D. Research

Frederick Eaton's Ph.D. Research

I conducted my Ph.D. research in the Computational and Biological Learning Lab in the Department of Engineering at the University of Cambridge in England.

I graduated on January 21, 2012. This web page is an archive of my old departmental web page, and hosts my research papers and related files.

Research Interests

I studied approximate inference algorithms and frameworks, applied to discrete statistical models.

Publications

Summary

2013

FH Eaton, Z Ghahramani. Model reductions for inference: generality of pairwise, binary, and planar factor graphs. Neural Computation May 2013. Vol 25, No. 5, pp. 1213-1260

This paper proves results concerning the representability of statistical models using factor graphs with constraints on topology, factor size, or variable domain. It characterises the expressive power of planar binary pairwise graphs, and introduces a new notion of model reduction in machine learning.

2012

FH Eaton. Combining Approximations for Inference (PhD thesis)

My PhD thesis, which explores the approximate inference problem from a functional perspective, asking what are the ways of "combining" two approximations and how these could be used to build new algorithms.

2011

FH Eaton. A conditional game for comparing approximations. In Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics (AISTATS 2011). Notable Paper Award

Defines a game which can be used to compare the accuracy of two approximations to the marginal probabilities of a statistical model. Apparently this is the first method that anyone has proposed for making such comparisons.

2009

FH Eaton, Z Ghahramani. Choosing a Variable to Clamp: Approximate Inference Using Conditioned Belief Propagation. In Proceedings of the Twelfth International Conference on Artificial Intelligence and Statistics (AISTATS 2009)

Proposes an algorithm for applying Belief Propagation to a model using divide-and-conquer, by recursively conditioning on specific variables. This algorithm can be seen as an approximate version of cutset conditioning. The paper also introduces "BBP" or "Back Belief Propagation", an application of back-propagation to belief propagation. BBP is used here for choosing condition variables, but has many other applications, for instance to parameter learning tasks in computer vision (Domke, "Parameter Learning with Truncated Message-Passing". CVPR 2011) and to empirical risk minimisation in general Machine Learning (Stoyanov et al, "Empirical Risk Minimization of Graphical Model Parameters Given Approximate Inference, Decoding, and Model Structure". AISTATS 2011).