GENOM: circuit-level optimizer based on a modified genetic algorithm kernel

This paper presents a new circuit/system-level optimizer based on a modified genetic algorithm kernel, including search space decomposition and a premature convergence prevention procedure. The tool allows the circuit performance analysis by either an equation-based approach or through the use of an electrical simulator, such as HSPICE. The achieved increase on optimization efficiency, compared to a standard genetic algorithm implementation, as well as the general purpose of the described approach are illustrated by a multi-objective, multi-constraint optimization of some well known circuits.