Chemical Genetic Programming.pptVIP

Chemical Genetic Programming – Evolutionary Optimization of the Translationfrom Genotypic Strings to Phenotypic Trees Outline Background and motivation Chemical Genetic Algorithm Genetic Programming Grammatical Evolution Chemical Genetic Programming Experiment Results Future direction Genetic Algorithms GA Population based, evolutionary algorithm Useful for optimizing multi-dimensional problems, for example: Functional optimization Synthesis of neural networks Antenna optimization Static mapping from genotypic to phenotypic data Chemical Genetic Algorithm CGA Suzuki, Sawai. 2002. Introduced biological translation processes into conventional GAs Dynamic genotypic-phenotypic relationship Shown to improve performance over conventional GAs More diversity and higher evolvability than conventional Gas Able to escape from local optima Genetic Programming GP Koza. 1992. Evolution of tree based programs Useful for solving problems, for example: Synthesis of electric circuit boards Symbolic regression “Evolution of side winding locomotion” No distinction between genotype and phenotype Grammatical Evolution GE Ryan, O’Neill, Collins. 1998. Distinguished between genotype and phenotype 1-Dimensional genotypic data translated to phenotypic tree Using predetermined grammar Maintained static translation Chemical Genetic Programming CGP Introduction of chemical processes to genetic programming Evolutionary optimization of translation from genotypic strings to phenotypic trees Translation Translation in GE Translation Using Rule Numbers The root symbol is always ‘I’ Rules are applied from left to right Rules are applied to the leftmost non-terminal symbol A Single Cell A population is a collection of cells There are four parts to each cell DNA sequence tRNA pool Amino acid pool Translation table Cell – DNA Three part to each DNA sequence Cell – tRNA Pool Used in chemical reaction Each tRNA relates a single rule number to a single non-terminal symbol Cell – Amino Acid Po