Managing team-based problem solving with symbiotic bid-based genetic programming

Abstract

Bid-based Genetic Programming (GP) provides an elegant mechanism for facilitating cooperative problem decomposition without an a priori specification of the number of team members. This is in contrast to existing teaming approaches where individuals learn a direct input-output map (e.g., from exemplars to class labels), allowing the approach to scale to problems with multiple outcomes (classes), while at the same time providing a mechanism for choosing an outcome from those suggested by team members. This paper proposes a symbiotic relationship that continues to support the cooperative bid-based process for problem decomposition while making the credit assignment process much clearer. Specifically, team membership is defined by a team population indexing combinations of GP individuals in a separate team member population. A Pareto-based competitive coevolutionary component enables the approach to scale to large problems by evolving informative test points in a third population. The ensuing Symbiotic Bid-Based (SBB) model is evaluated on three large classification problems and compared to the XCS learning classifier system (LCS) formulation and to the support vector machine (SVM) implementation LIBSVM. On two of the three problems investigated the overall accuracy of the SBB classifiers was found to be competitive with the XCS and SVM results. At the same time, on all problems, the SBB classifiers were able to detect instances of all classes whereas the XCS and SVM models often ignored exemplars of minor classes. Moreover, this was achieved with a level of model complexity significantly lower than that identified by the SVM and XCS solutions.

Rethinking multilevel selection in genetic programming

“…In addition, if the notches of two boxes do not overlap, the median of the two datasets differ at the 0.95 confidence interval. The detection accuracy mentioned in [9,15,18], and the multi-class detection rate or the score in [9,10] are equivalent to ODRs and ADRs in our experiments, respectively.…”

“…The results of the control algorithms, some presented in the format of box plots/violin plots, were gathered from [9,10,15,18]. Violin plots are a combination of a box plot and a kernel density plot which shows the probability density of the data at different values.…”

“…Bupa and Pima are two datasets known for poor class separability; the rate of error has been observed in the region of 30% across a wide range of machine learning algorithms [10]. Other reasons to select those datasets are that they have established performance levels [9,10,15,18], and they all represent real-world data, rather than artificial data. Therefore, the results obtained on those datasets make us sufficiently confident to judge if the new computational multilevel selection framework is applicable to solve real-world problems.…”

Evolution of division of labor in genetically homogenous groups

“…Numerous techniques have been proposed to automate the evolution of problem decomposition [8,13,18,24]. These techniques evolve specialists that produce solutions to subproblems.…”

“…These techniques evolve specialists that produce solutions to subproblems. The overall solution is a group of specialists comprising either one representative member of each species [8,18,24] or a group selected by the evolutionary algorithm [13]. The primary approach used for problem decomposition is the cooperative coevolution architecture [18], which evolves two or more species in isolated populations.…”