Evolution of an Effective Brain-Computer Interface Mouse via Genetic Programming with Adaptive Tarpeian Bloat Control

Abstract

The Tarpeian method for bloat control has been shownto be a robust technique to control bloat. Thecovariant Tarpeian method introduced last year, solvesthe problem of optimally setting the parameters of themethod so as to achieve full control over the dynamicsof mean program size. However, the theory supportingsuch a technique is applicable only in the case offitness proportional selection and for a generationalsystem with crossover only. we propose an adaptivevariant of the Tarpeian method, which does not sufferfrom this limitation. The method automatically adjuststhe rate of application of Tarpeian bloat control so asto achieve a desired program size dynamics. We test themethod in a variety of standard benchmark problems aswell as in a real-world application in the field ofBrain Computer Interfaces, obtaining excellentresults.