What is it about?
SOCEMO is an efficient and effective optimization algorithm for problems that have computationally expensive black box objective functions (function evaluations can take minutes to hours). We employ surrogate models to cheaply approximate the expensive functions. We show SOCEMO's efficiency for a range of test and application problems. The MATLAB codes are available from the author.
Featured Image
Why is it important?
Computer simulations are used in many science areas to study physical phenomena. As we are able to understand and model the underlying physics better, our simulation models become increasingly more expensive to run. Thus, when optimizing simulation model parameters, we need methods that require only few of these expensive simulations to find the global optimum. In contrast to other multi-objective algorithms, SOCEMO is able to find Pareto-optimal solutions much more efficiently.
Perspectives
Optimization problems can have a range of characteristics. Understanding these characteristics helps us to decide what solution methods should be used to find the optimal solution efficiently. We focus in this article on computationally expensive black box problems with multiple conflicting criteria. The algorithm SOCEMO is applicable to a wide range of application problems that have these characteristics. Thus, if your objective functions are expensive to obtain (computing time or monetary), SOCEMO is an efficient algorithm. We also believe that algorithm implementations should be made available to the readers, and thus we make the MATLAB implementation available (email the author).
Juliane Mueller
Lawrence Berkeley National Laboratory
Read the Original
This page is a summary of: SOCEMO: Surrogate Optimization of Computationally Expensive Multiobjective Problems, INFORMS Journal on Computing, November 2017, INFORMS,
DOI: 10.1287/ijoc.2017.0749.
You can read the full text:
Contributors
The following have contributed to this page
