Abstract:

As Bayesian inference gaining momentum in pharmaceutical research, more and more toolkits become available for data-based study of complex dynamical systems. In the talk we will look at one such tool, Stan, to review its inference capability and algorithm underneath. We will use the problem of ODE parameter estimation in pharmacokinetics as a running example, to look deeper on how numerical solution of ODEs are connected to Hamiltonian Monte Carlo sampling, parallel computing, sensitivity analysis, as well as automatic differentiation algorithms.

Bio:

Yi Zhang is an Associate Director at Sage Therapeutics, Inc. He received BS & MS degrees in Civil Engineering at Chongqing University, and MS in Mathematics and PhD in Ocean Engineering as Oregon State University. Before joining Sage, he was a mathematical software engineer working on Bayesian inference engine on pharmacometrics at Metrum Research Group, and previously a scientific researcher on finite element and optimization solver at Altair Engineering, Inc. His current research focuses on applying Bayesian inference, numerical analysis, and high performance computing to drug development research. He is a member dev team of open-source Bayesian probabilistic programming language Stan, as well as the maintainer of pharmacokinetics & pharmacodynamics software library Torsten.