Coevolutionary Algorithms and Evolutionary Simulation

When a system exhibits co-dependencies - such as inter-species dependencies, dependencies between species and environment - it is often more effective to coevolve models for each component separately, rather than evolve one model for the whole system (see Coevolution for biological aspects of coevolution).

The simplest case is two interacting species, where we might have time series data on the species' abundances, and on important environmental variables. A coevolu-tionary approach evolves equations for the two species semi-independently, bringing them together only in fitness evaluation (an equation's fitness might be measured as the mean error it generates in combination with individuals from the other population, or in some other way).

Coevolutionary systems may represent the modeled system in greater or lesser detail. In the extreme, individuals in the population directly represent individuals in the real world - we have moved beyond coevolutionary algorithms to coevolutionary simulation. However, it is difficult to determine the exact boundary.

Note that many other processes, for example, diffusion of economic strategies, resemble genetic processes. People copy strategies from other people, adapting them and combining them with other strategies (cf.reproduc-tion, mutation, crossover), and prefer strategies from successful people (fitness-based selection). Thus, coevo-lutionary systems may effectively model socioeconomic systems.

Typical applications

Coevolutionary algorithms have been used to model (human) socioeconomic dynamics in rangeland ecosystem models, the role of extinction in evolutionary ecosystem dynamics, the coevolutionary behavior of food webs, and the process of community assembly in ecosystems, to name but a few.

Solar Power

Solar Power

Start Saving On Your Electricity Bills Using The Power of the Sun And Other Natural Resources!

Get My Free Ebook

Post a comment