Parallel Functional Genetic Algorithms in Standard ML from RPL2

TL;DRAbstract

RPL2 is an imperative genetic algorithm language, whose data-parallel constructs may be readily expressed as standard higher-order functions. Here we consider the conversion of RPL2 programs to Standard ML, to test our skeleton based parallelising compiler. Results from running a simple GA to solve a Travelling Salesperson Problem on a Fujitsu AP3000 are presented. 1 Introduction Genetic algorithms(GAs) [1] are based on concepts from evolutionary biology and are increasingly used to solve problems which require the exploration of large search spaces. A genome encodes a potential solution to a problem as a linear sequence of symbols. The relative merits of such solutions are characterised by a fitness function which effectively interprets each genome and returns a fitness value. In a typical GA, a genome pool is initialised with random genomes. In each round of evolution, genomes are selected randomly for cross over, where symbols from one genome are exchanged with those from another...

Chat with Paper

AI Agents for this Paper