Epidemiological models have tremendous potential to forecast disease burdenand quantify the impact of interventions. Detailed models are increasinglypopular, however these models tend to be stochastic and very costly toevaluate. Fortunately, readily available high-performance cloud computing nowmeans that these models can be evaluated many times in parallel. Here, webriefly describe PSPO, an extension to Spall's second-order stochasticoptimization algorithm, Simultaneous Perturbation Stochastic Approximation(SPSA), that takes full advantage of parallel computing environments. The mainfocus of this work is on the use of PSPO to maximize the pseudo-likelihood of astochastic epidemiological model to data from a 1861 measles outbreak inHagelloch, Germany. Results indicate that PSPO far outperforms gradient ascentand SPSA on this challenging likelihood maximization problem.
展开▼
