View article

Knowledge extraction aims at detecting similarities and patterns hidden in the Pareto-optimal solutions arising from the outcome of a multi-objective optimization problem. The patterns may emerge from generic relationships of several variables or objective functions. Knowledge extraction is expected to bring out valuable information about a problem and is termed as a task of “innovization” elsewhere. While certain automated innovization methods have been proposed, in this report, we attempt to formalize the overall computational task from a machine learning and data analytics point of view. The results can be used to improve modeling and understand interdependencies among different objectives.