Angol
Many machine learning tasks involve solving complicated optimization problems, such as working on non-differentiable, non-continuous, non-unique objective functions, and sometimes it is even hard to define an explicit objective function. Evolutionary learning applies evolutionary algorithms to address the optimization problems for machine learning, and delivers encouraging performance in applications. However, due to the heuristic nature of evolutionary optimization, most of the promising results were empirical and lacking theoretical supports. This inadequacy suppresses evolutionary learning from being well received in the machine learning community which favors solid theoretical approaches. Recently there are great research efforts devoted to address the aforementioned embarrassing issue. This book presents a series of efforts in this direction. It is composed of four parts. Part I briefly introduces evolutionary learning and some preliminaries. Part II presents general theoretical tools for the analysis of running time and approximation performance of evolutionary algorithms. Based on these general tools, Part III presents a series of theoretical results about major factors of evolutionary optimization, such as recombination, representation, inaccurate evaluation, population, noise. Part IV presents development of evolutionary learning algorithms with provable theoretical guarantees for some representative tasks, where evolutionary learning offers excellent performance.
