Evolutionary Computation: A Framework for Single & Multi-Criterion Optimization and Decision-Making

Full title: Evolutionary Computation: An Emerging Framework for Practical Single and Multi-Criterion Optimization and Decision-Making Many optimization problems from engineering, science and business involve complex objective and constraint functions and other practicalities which violate the assumptions typically required for provable optimization algorithms. Differentiability, convexity and regularities of problems cannot be expected to be present in most practical problems. While classical gradient-based and convex programming methods are the best approaches when the problems satisfy the assumptions, there is a growing need for alternate methods which can be generically applied to any problem to achieve an optimal or a near-optimal solution. In this chapter, we introduce an emerging search and optimization method—evolutionary computation (EC)—which uses a population of solutions in every iteration and employs a series of operators that mimic natural evolutionary principles in arriving at better populations through generations. The population approach, flexibility of their operators for customization to different problem classes, and their direct search approach make EC methods applicable to a wide variety of optimization problems. This chapter discusses their working principles, presents case studies involving single and multi-criterion optimization problems, and discusses a few current research directions in the context of multi-criterion optimization and decision-making. Author: Kalyanmoy Deb