Efficient tree traversal to reduce code growth in tree-based genetic programming

Abstract

Genetic programming is an evolutionary optimization method following the principle of program induction. Genetic programming often uses variable-length tree structures for representing candidate solutions. A serious problem with variable-length representations is code growth: during evolution these tree structures tend to grow in size without a corresponding increase in fitness. Many anti-bloat methods focus solely on size reduction and forget about fitness improvement, which is rather strange when using an "optimization" method. This paper reports the application of a semantically driven local search operator to control code growth and improve best fitness. Five examples, two theoretical benchmark applications and three real-life test problems are used to illustrate the obtained size reduction and fitness improvement. Performance of the local search operator is also compared with various other anti-bloat methods such as size and depth delimiters, an expression simplifier, linear and adaptive parsimony pressure, automatically defined functions and Tarpeian bloat control

Evolution of an Effective Brain-Computer Interface Mouse via Genetic Programming with Adaptive Tarpeian Bloat Control

“…The Tarpeian method has been used in a variety of studies and applications. For example, in (Mahler et al, 2005) its performance and generalisation capabilities were studied, while it was compared with other bloat-control techniques in (Luke and Panait, 2006;Wyns and Boullart, 2009;Alfaro-Cid et al, 2010). The method has been used with success in the evolution of bin packing heuristics (Burke et al, 2007;Allen et al, 2009), in the evolution of image analysis operators (Roberts and Claridge, 2004), in artificial financial markets based on GP (MartinezJaramillo and Tsang, 2009), just to mention a few.…”

Covariant Tarpeian Method for Bloat Control in Genetic Programming

“…For example, Mahler et al (2005) studied its performance and generalisation capabilities, while it was compared with other bloat-control techniques in (Luke and Panait, 2006;Wyns and Boullart, 2009;Alfaro-Cid et al, 2010). The method has been used with success in the evolution of bin packing heuristics (Burke et al, 2007;Allen et al, 2009), in the evolution of image analysis operators (Roberts and Claridge, 2004), in artificial financial markets based on GP (Martinez-Jaramillo and Tsang, 2009), in predicting protein networks (Geffner et al, 2008), in the design of passive analog filters using GP (Chouza et al, 2009), in the prediction of proteinprotein functional associations (Garcia et al, 2008) and in the simplification of decision trees via GP (Garcia-Almanza and Tsang, 2006).…”