Ashley F. Emery
University of Washington, Seattle, WA, 98195-2600
Sang Ien
University of Washington, Seattle, WA, 98195-2600
Estimating thermal properties using Bayesian inference can be computationally demanding. The usual maximum likelihood (non-linear least squares) gives point values and an rough idea of the uncertainty in the parameters if their distributions are approximately normal. More accurate results require a Monte Carlo approach which for complex models is unrealistic. The Variational Bayesian approximation gives results comparable to Markov Chain Monte Carlo with only a few evaluation of the simulation model.