Optimal Belief Approximation

Abstract

In Bayesian statistics probability distributions express beliefs. However, for many problems the beliefs cannot be computed analytically and approximations of beliefs are needed. We seek a loss function that quantifies how "embarrassing" it is to communicate a given approximation. We reproduce and discuss an old proof showing that there is only one ranking under the requirements that (1) the best ranked approximation is the non-approximated belief and (2) that the ranking judges approximations only by their predictions for actual outcomes. The loss function that is obtained in the derivation is equal to the Kullback-Leibler divergence when normalized. This loss function is frequently used in the literature. However, there seems to be confusion about the correct order in which its functional arguments, the approximated and non-approximated beliefs, should be used. The correct order ensures that the recipient of a communication is only deprived of the minimal amount of information. We hope that the elementary derivation settles the apparent confusion. For example when approximating beliefs with Gaussian distributions the optimal approximation is given by moment matching. This is in contrast to many suggested computational schemes.

made improvements on the proof and the language

References in corpus (3)

• Using Relative Entropy to Find Optimal Approximations: an Application to Simple Fluids
• Divergence and Sufficiency for Convex Optimization
• Maximum Entropy and Sufficiency

Cited by in corpus (13)

• Information theory for fields
• A variational approach to moment-closure approximations for the kinetics of biomolecular reaction networks
• A Bayesian Framework for Cosmic String Searches in CMB Maps
• Marginal process framework: A model reduction tool for Markov jump processes
• Maximizing Information Gain for the Characterization of Biomolecular Circuits
• A Reputation Game Simulation: Emergent Social Phenomena from Information Theory
• Information Field Theory and Artificial Intelligence
• Towards information optimal simulation of partial differential equations
• Bayesian decomposition of the Galactic multi-frequency sky using probabilistic autoencoders
• Consistency and convergence of simulation schemes in Information field dynamics
• Towards Bayesian Data Compression
• GLASS: A General Likelihood Approximate Solution Scheme
• Quantifying imperfect cognition via achieved information gain