Genetic optimization with Dakota 2/3: A sandbox for experimentation


by Olivier Cleynen

Part 1: The optimization loop
Part 2: A sandbox for experimentation
Part 3: Practice

In order to play with the configuration settings of Dakota’s moga optimizer, you may use a “sandbox” demonstration optimization case in which the results are quickly obtained and easily visualized.

Evolution of a population in a Dakota moga dual-objective sandbox optimization

Attached is a demonstration optimization case. A two-objective, three-parameter optimization with external evaluation (in Python) is configured and extensively commented. Bash and Python scripts post-process the output of the simulation and display what is happening (the performance of individuals in each generation, and the size of the corresponding populations).

Evolution of population size during optimization in a Dakota moga sandbox optimization

Playing with the main settings of the mutation and crossover operators, and watching the results on the performance and population size graphs, is a great way to learn how they work.

Part 1: The optimization loop
Part 2: A sandbox for experimentation
Part 3: Practice

← Genetic optimization with Dakota 1/3: The optimization loop

Genetic optimization with Dakota 3/3: Practice →