Complex numerical models are ubiquitous in physical, atmospheric, biological, and engineering sciences. These models, often called simulators, are used to describe complicated interactions among many variables and processes in the systems and are usually accompanied by massive data. These simulators, acting as the forward model, are often required in forward propagation of uncertainty in simulation-based experiments and sensitivity analysis. In addition, we often need to extract information about parameters or unknown processes based on the simulators and data, called inverse problems. Surrogate models are often needed to reduce computational cost of forward model simulation and inverse problems. However, many surrogate modeling methods rely on the assumption of a particular parametric functional form or the assumption of Gaussian distribution, and often don’t scale well with the dimensionality or adapt to the geometry of input and output spaces. Presentations in this invited session will highlight some latest statistical methodological developments to address these challenges on surrogate modeling in uncertainty quantification and novel applications of surrogate models in forward simulation, calibration and inverse problems in biosciences, remote sensing, engineering, and climate sciences. 

Session Makeup: 4 speakers (listed alphabetically) 

Robert Gramacy, Professor, Department of Statistics, Virginia Tech, rbg@vt.edu 

Jon Hobbs, Data Scientist, NASA Jet Propulsion Laboratory, Caltech, jhobbs@jpl.nasa.gov 

Emily L. Kang, Professor, Division of Statistics and Data Science, Department of Mathematical Sciences, University of Cincinnati, kangel@ucmail.uc.edu  

Pulong Ma, Assistant Professor, School of Mathematical and Statistical Sciences, Clemson University, plma@clemson.edu

Organiser: Dr Emily Lei Kang 

Chair: Alex Konomi 

Speaker: Robert Gramacy 

Speaker: Jon Hobbs 

Speaker: Dr Emily Lei Kang 

Speaker: Dr Pulong Ma