EDUCATING INITIAL SOLUTIONS FOR GENETIC ALGORITHMS: A CHIP PLANNING OPTIMIZATION EXAMPLE

This paper aims at addressing a particular aspect of multiobjective optimization problems in Electronic Design Automation (EDA). More specifically, we will present our study of the impact on quality and performance of evolutionary algorithms initial solutions. A new scheme for initial population generation is presented, which improves the quality and efficiency of chip planning optimization using genetic algorithm. The underlying concept is to educate the initial solutions by artificially introducing high quality genes that will only show up in the natural evolutionary process. Since these genes are initially hidden, they do not impose their strength in the early phase of evolution, therefore preventing any premature convergence. 1