Recursion, lambda abstraction and genetic programming
Tina Yu, Christopher D. Clack · UCL Discovery (University College London) · 1998
Module creation and reuse are essential for Genetic Programming (GP) to be effective with larger and more complex problems. This paper presents a particular kind of program structure to serve these purposes: modules are represented as λ abstractions and their reuse is achieved through an implicit recursion. A type system is used to preserve this structure. The structure of λ abstraction and implicit recursion also provides structure abstraction in the program. Since the GP paradigm evolves program structure and contents simultaneously, structure abstraction can reduce the search effort for good program structure. Most evolutionary effort is then focused on the search for correct program contents rather than the structure. Experiments on the Even-N-Parity problem show that, with the structure of λ abstractions and implicit recursion, GP is able to find a general solution which works for any value of N very efficiently.