The Hard Truth
Bayesian methodology provides the means to combine prior knowledge about competing models of reality and available data to draw inferences about the validity of those models. The posterior quantifies the degree of certainty one has about those models. We propose a method to determine the uncertainty in a specific feature of a Bayesian solution. Our approach is based on an analogy between the negative logarithm of the posterior and a physical potential. This analogy leads to the interpretation of gradient of this potential as a force that acts on the model. As model parameters are perturbed from their maximum a posteriori (MAP) values, the strength of the restoring force that drives them back to the MAP solution is directly related to the uncertainty in those parameter estimates. The correlations between the uncertainties of parameter estimates can be elucidated.
KeywordsMaximum Entropy Negative Logarithm Posterior Probability Distribution Probabilistic Display Physical Potential
Unable to display preview. Download preview PDF.
- J. Skilling. Clouds, presented at the Workshop on Maximum Entropy and Bayesian Methods, July 19–24, 1992, Paris.Google Scholar
- P. R. Bevington. Data Reduction and Error Analysis for the Physical Sciences. McGraw-Hill, New York, 1969.Google Scholar
- K. M. Hanson. Flexible prior models in Bayesian image analysis. In A. Mohammad-Djafari and G. Demoment, editors, Maximum Entropy and Bayesian Methods, pages 399–406. Kluwer Academic, 1993.Google Scholar
- G. S. Cunningham, K. M. Hanson, G. R. Jennings, Jr., and D. R. Wolf. An object-oriented optimization system. Proc. IEEE Int. Conf Image Processing, 111: 826–830, 1994.Google Scholar
- K. M. Hanson. Making binary decisions based on the posterior probability distribution associated with tomographic reconstrcutions. In C. R. Smith, G. J. Erickson, and P. O. Neudorfer, editors, Maximum Entropy and Bayesian Methods, pages 313–326. Kluwer Academic, 1991.Google Scholar