Recently, a new variant of Geometric Semantic Genetic ProgrammingGeometric Semantic Genetic Programming (GSGP)(GSGP) was introduced that, while maintaining the property of inducing a unimodal error surface for all supervised learning problems, is able to generate models that are compact enough to be interpretable by humans. This variant is called the Semantic Learning algorithm based on Inflate and deflateSemantic Learning algorithm based on Inflate and deflate Mutations MutationSLIM (SLIM_GSGP) and, as the name suggests, it is based on two types of mutation: one (inflate) that generates offspring larger than their parents, similar to traditional geometric semantic mutation, and the other (deflate) that generates offspring smaller than their parents. This chapter reviews and extends the initial work on SLIMSLIM_GSGPGeometric Semantic Genetic Programming (GSGP) by introducing two novel variants, thoroughly studying the geometric characteristics of the SLIMSLIM_GSGPGeometric Semantic Genetic Programming (GSGP) operators and discussing the experimental results and their interpretation in greater depth. The main conclusion is that SLIMSLIM_GSGPGeometric Semantic Genetic Programming (GSGP) is a very promising method, warranting significant investment inBloat future research.Semantic learning