It is time to stop quivering in our boots in pointless fear of the future and just roll up our sleeves and build it.
- Ray Pierrehumbert

Tuesday, July 26, 2011

Computation and Prediction

I'll be in the right place at the right time for an interesting event for a change and will report.
"The Emerging Age of Predictive Computational Science"

with Dr. Tinsley Oden, Institute for Computational Engineering and Sciences (ICES), The University of Texas at Austin

When: Tuesday, August 2, 2011
Where: AT&T Conference Center / Amphitheater (Room 204) / 1900 University Ave.
Cost: Free and open to the public

Summary:

What exactly is the purpose of scientific discovery if it is not to inform us sufficiently to make predictions of the outcome of physical events and processes? Science is the enterprise of acquiring knowledge, and knowledge enables us, so we thought, to anticipate how things will behave-to forecast the way things will happen in the future.

A modern look at this idea suggests that scientific predictions aren't as straightforward as some may think, particularly with the enhanced power of scientific discovery made possible by computer models and computers. The anatomy of computer predictions has recently become the subject of intense research, because we are now relying on computer models to predict events of enormous importance in making critical decisions that affect our welfare and security, such as climate change, the performance of energy and defense systems, the biology of diseases, and the outcome of medical procedures. Just how good are predictions of such complex phenomena, and how can we quantify the inevitable uncertainty in such predictions?

This presentation traces the development of scientific thinking and philosophy that underlie predictivity. It is argued that the fundamental issues of affecting the quantitative prediction of physical events using computer models are code and solution verification, model calibration and validation, and uncertainty quantification. These are the components of Predictive Science. It is also argued that the subjective probability inherent in Bayesian statistics provides a general framework for understanding and implementing predictive computational methods. Some examples of progress in this area at ICES will be presented.

4 comments:

Victor said...

Lecture at 6:30 pm, networking reception 5:45.

http://austinforum.org/speakers/oden.html

Dan Olner said...

"... model calibration and validation, and uncertainty quantification. These are the components of Predictive Science."

Really? Would I wrong in thinking the term `validation' comes from software development, and has since leaked back into model justification, and via there to claiming it's a vital part of 'predictive science?'

jstults said...

Would I wrong in thinking the term `validation' comes from software development, and has since leaked back into model justification, and via there to claiming it's a vital part of 'predictive science?'

Yes, you would wrong. Two different usages in two different communities.

The folks from PECOS have done some interesting work on fluid model validation (mostly all DoE funded).

Dan Olner said...

Thanks for the link jstults. Actually, having just checked, 'validation' is used waaaay back to the early twentieth century at least...