Adaptive Model-Switch Thermodynamic Integration: A Bayes Factor Estimator

Bayes factors are a widely used model comparison criterion, but are often quite challenging and computationally expensive to estimate well. We propose a variation of the model-switch thermodynamic integration method for estimating Bayes factors that decreases the variance of the estimator and essentially eliminates integration error. We choose samples whose density in $\beta$ is proportional to the standard deviation of previously drawn samples. Rather than averaging many samples at each $\beta$, we use a very dense set of $\beta$ values, collecting one sample at each in order of increasing $\beta$ and one sample at each in order of decreasing $\beta$ to minimize “thermic lag”. We study our procedure with a toy model (based loosely on some signal recovery problems). In this model we find that adaptive sampling gives us a 2.7-fold decrease in the spread of the estimated Bayes factor. We test our procedure both drawing directly from the posterior distribution and using a Markov Chain Monte Carlo sampler.