SELF-ADAPTATION OF THE REPRESENTATION OF A REAL-WORLD DESIGN IN AN ITERATIVE OPTIMIZATION

摘要

A computer- implemented method for iteratively optimizing a real -world structure, the method comprising the steps of : a.) encoding the structure in a parameter set comprising object parameters encoding the real -world structure and strategy parameters, b.) modifying the parameter set wherein the strategy parameters control the modification of the value of the object parameters between two steps of the iterative optimization, c.) repeating step b.) until a termination criterion is met, wherein step b.) comprises the step of modifying the topology and not only the values of the parameter set, the topology of the parameter set being modified by adapting the number of object parameters as well as adapting the associated strategy parameters, and when modifying the topology of the parameter set, the object parameters and the strategy parameters are set such that after carrying out a modification step b.) the expectation value (E) and the variance (VAR) of the parameter set with modified topology are equal to the expectation value and the variance of the parameter set with unmodified topology.