The OPTMILP Option Tuner

The OPTMILP procedure provides many solver techniques and algorithms, including branch-and-bound, cutting planes, and heuristics. It also provides control options that you can adjust to improve the performance of these techniques. Although the default values of the control options have been tuned to work well for most instances, you might need to adjust one or more option values for a specific problem. The OPTMILP option tuner is a tool that enables you to explore alternative (and potentially better) option configurations for your optimization problems.

To use the tuner, you specify a single problem or set of problems to be solved and a list of options to be tuned. You can specify initial values for the options to be tuned. The tuner then uses a heuristic local search technique to generate a sequence of configurations. A configuration is a set of the specified options to be tuned along with a fixed value for each option. The tuner attempts to locate configurations that enable the OPTMILP procedure to process problems more quickly than the default option values or specified initial values.

The tuner option can be run in either single-machine mode or distributed mode. In single-machine mode, you can specify the number of threads to use on a single computer. In distributed mode, you can specify the number of computer nodes and the number of threads per node to use on a distributed computing environment. The tuner option only supports nondeterministic mode, so the tuning results may vary among different runs.

Note: Distributed mode requires SAS High-Performance Optimization software.