Developments in Cartesian Genetic Programming: self-modifying CGP

Abstract

Self-modifying Cartesian Genetic Programming (SMCGP) is a general purpose, graph-based, developmental form of Genetic Programming founded on Cartesian Genetic Programming. In addition to the usual computational functions, it includes functions that can modify the program encoded in the genotype. This means that programs can be iterated to produce an infinite sequence of programs (phenotypes) from a single evolved genotype. It also allows programs to acquire more inputs and produce more outputs during this iteration. We discuss how SMCGP can be used and the results obtained in several different problem domains, including digital circuits, generation of patterns and sequences, and mathematical problems. We find that SMCGP can efficiently solve all the problems studied. In addition, we prove mathematically that evolved programs can provide general solutions to a number of problems: n-input even-parity, n-input adder, and sequence approximation to p.

Tag-based modularity in tree-based genetic programming

“…The code that we used for the experiments presented here is freely available for others to use for their own experiments. 4 Aside from the investigation of the hypotheses and alternative approaches proposed above, several other lines of 4 http://hampshire.edu/lspector/tags-gecco-2012 future work seem to us to be potentially worthwhile. First, it should be informative to conduct experiments like those reported here on a much wider range of problems.…”

“…The core idea of tagging and of tag-based (closest-match) retrieval of modules should be applicable in many other forms of genetic programming-for example, in machine code genetic programming [13], Cartesian genetic programming [4], and grammatical evolution [14]-and the problems that we have encountered here may not occur, or may not be as severe, in those contexts. It might turn out that the representational constraints of traditional, tree-based genetic programming are a poor match to the requirements for tag-based modularity.…”

Cartesian Genetic Programming for Image Processing

“…Specialised 'Input' functions are also provided (INP, INPP, SKIP), which manipulate a pointer that indexes the available inputs and return the currently indexed input. A full description can be found in (Harding et al, 2010b) and (Harding et al, 2010a). The output is taken from the last node in the genome.…”

An Investigation of the Importance of Mechanisms and Parameters in a Multicellular Developmental System

“…Although the ability of artificial development to be scalable has been demonstrated by simple experiments [13], [19], [30], [31], [37], successful use of development in the design of systems at desired complexities that tackle real world problems is yet to be achieved. A recent achievement of an ADS has been presented by Harding et al who evolved a self-modifying Cartesian genetic programming that can find general solutions to parity and binary addition [38].…”

“…An example of a non-grammatical macro-model developmental system is the self-modifying CGP, which models a CGP system that could alter its own structure over time after the evolution phase is complete [14], [38]. A macromodel developmental system should be computationally more efficient when compared to a micro-model developmental system in modeling developmental behavior.…”