In this article, the dc filter design and optimization problem is studied for dc electrical power distribution systems onboard more-electric aircraft. Component sizing models are built to serve as the basis of the optimization whose objectives are mass and power loss of this filter. A categorization strategy of search and surrogate algorithms is proposed and used for the target multiobjective optimization problem (MOOP). A genetic algorithm is utilized as a search algorithm to identify potential best solutions based on a set of filter sizing functions (subject to constraints). In addition, two machine learning (ML) algorithms are considered as surrogate algorithms to address the same optimization problem. In the ML training process, a constraint violation model is applied since there are various constraints in optimization, and this kind of classification model is relatively difficult to train. A support vector machine is applied for the constraint violation model; after that, two artificial neural networks are trained as the final surrogate model for mapping design variables to filter performance. To address these issues, a novel category of search and surrogate algorithms is proposed. Both algorithms are explored to solve the filter MOOP, and their optimization results are compared at the end.