next up previous contents
Next: Optimization from Matlab Up: External Interfaces Previous: External Interfaces   Contents


Optimizing Through an External Command

Figure 3.1.1: Screenshot of an external runtime problem configuration.
Image screenshot-ExtRun

For easy optimization using a runnable program, we provide the ExternalRuntimeProblem class. So if you have a runnable function implementation called ``testEval'' (or ``testEval.exe''), that takes a series of double valued arguments and produces a result value as output (on stdout), you can select ExternalRuntimeProblem in the GUI, enter the path and program name into the command field, adapt boundaries and the problem dimension and just start optimization. For a programm called testEval with a two-dimensional solution space, which you want to optimize within $ [-5,10]^{2}$ , the configuration can be seen in Fig. 3.1.1. In the example, testEval may be called and, in the single objective case, should produce output as in:

$ /home/username/testEval 7.1642923 -4.2346211

28.1371802

$ _

EVA2 starts a system process to run the external program and converts input/output using String objects, which is by itself rather slow. Still, it may outweigh the costs of reimplementing it anew and can even be negligible if the runtime of the calculation is of higher order of magnitude. It may also just come in handy if you want to try what EVA2 might do for you.


next up previous contents
Next: Optimization from Matlab Up: External Interfaces Previous: External Interfaces   Contents
Marcel Kronfeld 2011-05-05