Description

The goals and structure of the workshop are summarized as follows.

Participants. The workshop will bring together a diverse group of university and laboratory researchers to consider foundational mathematical, statistical, and computational methods for verification and validation in complex application areas for predictive computational science.
Invited Lectures. The workshop is organized around invited technical talks describing both key methods and representative applications of predictive computational science, interspersed with structured group discussions. Significant time is scheduled for informal technical interactions. An overview of the presentations and discussions will be published.

Purpose. Talks will be designed to survey important themes and open research problems in the topics discussed. The underlying foundation of the workshop is concentration around the issues of verification and validation, but individual speakers are not expected to specifically address verification and validation per se. Rather, the V&V emphasis will be most manifest in the planned discussion sessions. The purpose of the meeting is to emphasize the technical content that is required for successful and consequential V&V.
Discussions. An important aspect of the workshop is the planned discussion session on the training of scientists in V&V, both for those that want to specialize in the development of methods specifically aimed at V&V and for the general computational science community that need to be aware of and perhaps incorporate V&V methodologies in their codes. This session will address issues such as what courses students need to take to be V&V savvy, what new courses need to be created to meet the need to train students in V&V, and what mechanisms can be used to give students experience in V&V in realistic settings.

Questions for the Discussion Sessions

A key feature of the workshop will be the structured discussion sessions that will be held every day. These sessions are meant to facilitate discussions among methods experts and between methods experts and applications scientists. The goal is to come to a consensus about which methods are most promising for V&V, what directions methods development should take in the future if they are to be useful for V&V, how students should be trained so that they can participate in method development for V&V or so that they can incorporate V&V methods in their applications research. We will allow maximal flexibility so that the discussions can lead to where they will be most fruitful. At the same time, the discussions will remain focused in the V&V goals of the workshop.

The following is a list of the questions around which the discussion sessions could be framed. We would also like invited lecturers to address these issues where possible.

What is the relative importance of quantifying uncertainty, or some fraction of it, in computational science evidenced by current practice?
What methods, or components of methods – mathematical, statistical, computational – for V&V have already proven to be useful or have the potential to be useful and are generally applicable for a broad set of computational science predictions? What are the application problems, or areas, with the biggest need for new and better V&V methods?
What are realistic goals over the next few years and over the longer term for useful V&V based on sound technical methods? What kind and what quality of information can be extracted via V&V methods and how do we communicate the accuracy, fidelity, credibility of computational models after we have performed V&V to non-technical people?
What are the elements, e.g., courses, internships, etc., of V&V education? Why is this “in addition to” traditional computational science education?

Organizers:

Clayton Webster (Chair), Sandia National Laboratories
Scott Collis, Sandia National Laboratories
Max Gunzburger, Florida State University
Timothy Trucano, Sandia National Laboratories
David Womble, Sandia National Laboratories