This book, part of Springer’s “Genetic and Evolutionary Computation” series, contains the proceedings of the “14th meeting of the annual Genetic Programming Theory and Practice Workshop, [held] by the Center for the Study of Complex Systems at the University of Michigan.” The book comprises 14 chapters. The contributors are experts in genetic programming (GP) from various countries. The book’s chapters are on diverse topics; nevertheless, the common thread connecting them is GP.

The first chapter looks at how structural and semantic similarity measures may be used for studying “population diversity ... in the evolutionary dynamics of GP.” This is done by assuming three variations: standard GP, offspring selection GP, and age-layered population structure GP. The next chapter studies hybrid genetic and behavioral diversity methods in GP for a group of similar problems. The third chapter illustrates an interesting application of GP with grammars for the problem of predicting tax avoidance. The next chapter discusses a game controller that attempts to simulate human learning behavior, with an aim to learn from experience and play new games without a steep learning curve. The fifth chapter looks at runs in GP; the authors use tools to visualize GP runs and study the evolution of solutions over 20 generations. The next chapter describes the use of “linear genomes that are translated into hierarchical programs for execution”; for this purpose, the authors utilize the Push programming language. Chapter 7 discusses the use of “random walkers and hill climbers to study how robustness and evolvability are related to the structure of genotypic, phenotypic, and fitness networks.” The effect of these networks on the evolutionary search process is also studied.

Chapter 8 conveys a strong message: local search is underutilized in GP. The authors demonstrate this by providing experimental evidence that local search can help in overcoming two major imperfections of GP. The first imperfection of GP is needlessly large programs, that is, “the size (number of nodes) of the best solution and/or the average size of all the individuals increases even when the quality of the solutions [does not improve].” The second imperfection of GP is that it uses ineffective search operators. The next chapter expands the EC-Star rule-set representation to permit probabilistic classifiers. EC-Star is a “massively distributed evolutionary platform that uses age-varying fitness as the basis for distribution.” Chapter 10 indicates “how proportional analogy problems can be solved with GP,” and also “how analogical reasoning can be [employed] in GP ... for problem decomposition.”

The next chapter infers that algorithms featuring “extreme accuracy in [symbolic regression] do not translate directly into symbolic multi-class classification.” The author presents “an evolutionary algorithm for optimizing a single symbolic multi-class classification candidate.” Chapter 12 presents a general model to construct “geometric dispersion operators for geometric semantic genetic programming in the context of symbolic regression.” The authors demonstrate that dispersion operators can improve search. The next chapter researches how GP can help the expert-driven procedure of building up data-driven models. The focus is on helping asset model growth with evolutionary increments. The last chapter showcases “a GP-based automated machine learning system” that makes optimal use of “a series of feature preprocessors and machine learning models.” The objective is to “maximiz[e] classification accuracy on a supervised classification problem.”

This highly technical book is meant for a very specialized audience: researchers in GP. The topics discussed offer interesting insight into how research in GP is evolving. There are many references for further exploration. I strongly recommend this book for researchers in evolutionary computing and GP.