FlexGP.py: Prototyping Flexibly-Scaled, Flexibly-Factored Genetic Programming for the Cloud

Abstract

Running genetic programming on the cloud presentsresearchers with great opportunities and challenges. Weargue that standard island algorithms do not have theproperties of elasticity and robustness required to runwell on the cloud. We present a prototyped design for adecentralised, heterogeneous, robust, self-scaling,self-factoring, self-aggregating genetic programmingalgorithm. We investigate its properties using asoftware 'sandbox'.