Evolving efficient solutions to complex problems using the artificial epigenetic network

Abstract

The artificial epigenetic network (AEN) is a computational model which is able to topologically modify its structure according to environmental stimulus. This approach is inspired by the functionality of epigenetics in nature, specifically, processes such as chromatin modifications which are able to dynamically modify the topology of gene regulatory networks. The AEN has previously been shown to perform well when applied to tasks which require a range of dynamical behaviors to be solved optimally. In addition, it has been shown that pruning of the AEN to remove non-functional elements can result in highly compact solutions to complex dynamical tasks. In this work, a method has been developed which provides the AEN with the ability to self prune throughout the optimisation process, whilst maintaining functionality. To test this hypothesis, the AEN is applied to a range of dynamical tasks and the most optimal solutions are analysed in terms of function and structure.