Coevolutionary Computation

Abstract

This article proposes a general framework for the use of coevolution to boost the performance of genetic search. It combines coevolution with yet another biologically inspired technique, called lifetime fitness evaluation (LTFE). Two unrelated problems--neural net learning and constraint satisfaction--are used to illustrate the approach. Both problems use predator-prey interactions to boost the search. In contrast with traditional "single population" genetic algorithms (GAs), two populations constantly interact and co-evolve. However, the same algorithm can also be used with different types of co-evolutionary interactions. As an example, the symbiotic coevolution of solutions and genetic representations is shown to provide an elegant solution to the problem of finding a suitable genetic representation. The approach presented here greatly profits from the partial and continuous nature of LTFE. Noise tolerance is one advantage. Even more important, LTFE is ideally suited to deal with coupled fitness landscapes typical for coevolution.

Finding numerical solutions of diophantine equations using ant colony optimization

“…Though the procedure could find solutions for smaller equations, it was not fully random in nature and seemed more like a steepest ascent hill climbing and hits on local optimum points. Abraham et al [2] and Abraham and Sanglikar [3] proposed a novel reciprocally induced co-evolution method [30] [32] based on "host parasite co-evolution" [21,29,39] to tide over the repeated occurrence of local hilltops in a typical GA. Joya et al [24] applied higher order Hopfield neural network and Abraham et al [4] used feed forward back propagation network to find numerical solutions of Diophantine equation. Abraham and Sanglikar [5] proposed simulated annealing as a viable search strategy for finding numerical solutions of a Diophantine equation.…”

Development and application of a hybrid genetic algorithm for resource optimization and management

“…-The problem is epistatic -solutions over the feasible region are closely coupled and small alterations in the nodal weightings trigger cumulative effects (perturbations) in the solution space. Consequently, this affects the fitness of each traversal path, and also has some impact on the efficiency of a GA system, often leading to termination at local optima (Paredis, 1993(Paredis, , 1995. • Each ellipse has multiple cost values, resulting from its interactions with the circles.…”

“…Therefore, the fitness network serves a dual purpose by integrating two general paradigms: genetic search and state-space search. Some authors have advocated the adoption of this GSSS approach in solving COPs (Paredis, 1993(Paredis, , 1995. This is the approach we adopted in solving the problem.…”

Coevolution of lags and RBFNs for time series forecasting: L-Co-R algorithm

“…This kind of learning processes in which two complementary and dependent objectives exist is suited to be tackled by coevolutionary algorithms (Paredis 1995). Therefore, we propose Lags COevolving with Rbfns (L-Co-R), a coevolutionary algorithm able to jointly solve the two problems.…”