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 shown to be a robust technique to control bloat. The covariant Tarpeian method introduced last year, solves the problem of optimally setting the parameters of the method so as to achieve full control over the dynamics of mean program size. However, the theory supporting such a technique is applicable only in the case of fitness proportional selection and for a generational system with crossover only. In this paper, we propose an adaptive variant of the Tarpeian method, which does not suffer from this limitation. The method automatically adjusts the rate of application of Tarpeian bloat control so as to achieve a desired mean program size. We test the method in a variety of standard benchmark problems as well as in a realworld application in the field of Brain Computer Interfaces, obtaining excellent results.

A genetic programming approach to the evolution of brain–computer interfaces for 2-D mouse–pointer control

“…As a first attempt to address these issues, we have started to explore hybrid systems where one SVM or an ensemble of SVMs is trained on a per-subject basis, but where a trajectory integrator which is applicable across subjects is evolved by GP [22,23]. In these systems, once evolved, the GP integrator can be used over and over again.…”

“…The most successful BCI approaches for 2-D pointer control, to date, are those based on frequency analysis and the detection of l (8-13 Hz) and b (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) Hz) rhythms in EEG [5] and those using cortical electrode arrays (e.g., [12]), i.e., electrodes directly placed on the surface of the brain. Both, however, have serious drawbacks.…”

The Minimization of Public Facilities With Enhanced Genetic Algorithms Using War Elimination

“…There are several alternative formulations of the GA that aim at the problem of homogeneous population and avoiding local minima such as Tarpeian method [26], [27], which randomly kills individuals not adhering to given standards or Island model [28], [29] which partitions the population into sub-populations, where only local interactions are allowed (migrants are moved between sub-populations periodically). The disadvantage of these modifications over our solution is mainly their implementation difficulty as in the case of the island model (working with parallel populations is burdensome to implement) or a general lack of robustness and stability (Tarpeian method) [26]- [29]. To address these problems we propose a modification based on a novel operator called war inspired by the patterns in the population evolution observed during wartime periods.…”