This chapter discusses a last class of evolutionary algorithms, the genetic programming (GP) framework. This method aims to evolve computer programs so as to perform a given task or to propose analytic expressions describing a set of data. This is a particular class of machine learning technique for which the search space is that of computer instructions. But, obviously, such an ambitious goal requires to restrict the type of computer programs that can be considered. They must be resilient to genetic evolution such as crossover and mutations. The text first presents the standard tree-based, functional representation of a GP and the associated genetic operators. An example of using this approach to derive a trading model in finance is discussed. Secondly, a linear and procedural representation of a GP is proposed, with simple examples of deriving Boolean or algebraic expressions explaining a given set of data. The strength and weakness of both representations are discussed.