Maximizing Performance and Cost Effectiveness
Optimization is the process of varying input parameters to a model with the ultimate goal of achieving improved performance for one or more objectives.
In most problems of interest, there are multiple objectives that must be simultaneously optimized and multiple constraints that must be respected. This is the job of any optimization algorithm.
In addition to having multiple objectives and constraints, many problems involve discrete decision spaces. That is, there is a limited selection of possible choices that can be made and the task is to choose from amongst them. Problems like this present particular difficulties for optimization algorithms.
Sandia has developed a number of applications that have been successfully demonstrated on these large-scale and difficult problems.
Why is optimization important?
- While analysis answers the question "What will be the outcome if I do …" optimization answers the question, "What should I do to get the best outcome?"
- It often identifies high-performing, non-intuitive solutions
- It can be used to quantify the tradeoffs that typically exist between multiple objectives (for example, cost and performance)
- It aids designers and decision makers in understanding the nature of their problem and provides a mathematical basis to support the decisions they make
- For large-scale systems, the performance gains achieved from optimization can be significant, as can the cost savings
What are the objectives for developing this capability?
- Provide customers with a rich set of highly functional optimization capabilities
- Provide analysts with the means of supplying decision makers with high-quality, defensible results
What are the research areas?
- Selection of optimization techniques for best problem-specific performance
- Optimization under uncertainty (OUU)
- Visualization methods for effective communication of complex results
- Distributed optimization techniques for computationally-expensive simulation models