SAMSI Program on Development Assessment and Utilization of Complex Computer Models

NPCDS Programme on Design and Analysis of Computer Experiments for Complex Systems

Summer School on the Design and Analysis of Computer Experiments

General Information
Course Contents
Application
Accommodations
Detailed Schedule
Additional Information

General Information

The design and analysis of experiments continues to make important and far-reaching contributions to scientific investigation. Historically, experimenters have relied on physical experiments to help understand processes. The rapid growth in computing power has made the computational simulation of complex systems feasible and helped avoid physical experimentation that might otherwise be too time consuming, costly, or even impossible to observe.

With the availability of computer models come new and important challenges. For the person building a computer simulator, it is tempting to model more and more precisely the system under study as more computer power becomes available. Consequently in many situations, the dimensionality of the input to the computer code can be very large. In others, simulation of the complex phenomena becomes computationally expensive. Still in other applications, the output of the simulator may be univariate, multivariate, or functional data that is a very complex function of the inputs. The statistical challenges lie in the design and analysis of experiments (computer experiments) on complex computer codes.

The Summer School brings together graduate students, post-doctoral fellows and industry researchers with leading academic researchers from around the world.

All Summer School activities will take place at Simon Fraser University's IRMACS (Interdisciplinary Research in the Mathematical and Computational Sciences (www.irmacs.sfu.ca) facility. Specifically, we will utilize the IRMACS Presentation/Visualization Theatre for all presentations. Furthermore, the Summer School will have a dedicated IRMACS lab with 20-30 Apple Power Macintosh G5 computers. Participants will also have accounts on the Beowulf style cluster super-computer if required. Working groups will also have dedicated access to smaller meeting rooms with large touch-screen plasma monitors for visualization.

Organizing Committee: Derek Bingham (Committee Chair and NPCDS Liaison, SFU), David Higdon (LANL), Jerry Sacks (NISS), Will Welch (UBC); Susie Bayarri (SAMSI Program Liaison, U. Valencia), James Berger (SAMSI Directorate Liaison)

Course Contents

We have attempted to create a novel format for the endeavor to make it appealing to all participants. There are three components to the program.

Jerry Sacks and Will Welch (UBC) will teach a two-day short course on the design and analysis of computer experiments and models. The aim of the short course is to provide the participants with an introduction to cutting edge statistical methods used in computer experiments and model validation and also a common language for the next stages of the Summer School.

Topics to be covered:

• Random Function Models
• Designs
• Data Adaptive Random Function Model/Assessment of Accuracy
• Detailed Example
• Visualization

Leading researchers have been invited for a symposium on software for experimentation on, and validation of, complex computer codes. The motivation for the symposium is threefold:

• To familiarize the participants with the computational challenges and available software for the design and analysis of computer models;


• To provide an environment for collaborative research;


• To provide the tools for the third component of the summer school - industrial problem solving.

Topics to be covered:

• Review of Day 1: Case Study Example
• Optimization
• Model Calibration
• Validation
• Follow-up discussion with lecturers

The final component of the summer school is group problem solving of a real scientific problem on cosmology. The scientists at Los Alamos National Laboratory are among the world leaders in development of new methodology for computer experiments. They have agreed to lead part of the program where the goal is to analyze a real computer experiment. Participants will become familiar with the scientific problem of interest and the running of the computer simulator (though we expect to have a suite of trials performed before the Summer School). Participants will form working groups and use the knowledge from the first two stages of the School to attempt to address the scientific problem.

Cosmology Problem for Problem Solving Sessions: During the past two decades, due primarily to rapid advances in several observational technologies, a new "concordance model" of cosmology has now been cemented, with its parameters known to the remarkable level of roughly 10%. In the concordance model, the Universe consists of approximately 70% dark (vacuum) energy, 25% dark matter that interacts only gravitationally, and 5% normal, baryonic matter. Depending on the detailed level at which the model is described, the number of parameters associated with it ranges from approximately a half-dozen to twenty. These phenomenological parameters must be estimated by detailed comparison with observational and experimental data (where available).

Although the details of galaxy formation are still murky, much information may be gleaned from simulations that track only the behavior of dark matter. The code is fast and has well-characterized errors, hence it is well suited to generate a large number of experimental designs for statistical analysis.

Statistical problems facing the scientists include developing:

• statistical methodology to combine physical and computer observations to form a cohesive predictive model


• experimental design methodology to determine input settings at which to run the simulator


• basis representations for reducing the dimension of the simulator output


• uncertainty to give coherent inference.

Application

Interested individuals should apply, using the ON-LINE APPLICATION FORM. This form also includes the application for financial support. You will be notified as soon as possible after your application if your participation will be possible. Note that the application/registration deadline is July 12, 2006

In order to ensure your application/registration is correct, we ask that you:

• refresh/reload the application/registration page to ensure you have all updates
• type in your information (cutting and pasting will distort the information we receive)
• make any clarifications/corrections, in the Special Requests section
• click the submit button only once

Accommodations

The Summer School lodging is located on the Simon Fraser University campus. There are a variety of accommodations available. There are a number of private residence rooms (http://www.sfu.ca/conference-accommodation/residence.html) in the usual university dormitory style. These are new and quite attractive, and cost $40.00 (Canadian) per day. In addition, there are also a few standard hotel rooms (http://www.sfu.ca/conference-accommodation/simon.html), costing $79.00 (queen) per day and $99.00 (Queen/Sofa) per day. Please make reservations at the indicated websites. Other hotel accommodations are available off campus upon request.

Schedule (Detailed Schedule)

DAY 2 (Short Course)

DAY 3 AND 4 (Software Symposium)
Details to be added later

DAY 5 (Problem Solving)

• Presentation of scientific application


• Presentation of statistical problem


• Introduction to the simulator code


• Introduction to IRMACS computing facility


• Formation of working groups


• Team problem solving

DAY 6 (Problem Solving)

• Team problem solving


• Presentation of solution outlines

For Additional Information, Contact:

Entire site © 2001-2010, Statistical and Applied Mathematical Sciences Institute. All Rights Reserved.