We propose a multi-objective method Age-Fitness Pareto Optimization (AFPO) for avoiding premature convergence in evolutionary algorithms, and demonstrate a 3× performance improvement over comparable methods.

Previous research has shown that partitioning an evolving population into age groups can greatly improve the ability to identify global optima and avoid converging to local optima. Here, we propose that treating age as an explicit optimization criterion can increase performance even further, with fewer algorithm implementation parameters.

The proposed method evolves a population on the two-dimensional Pareto front comprising (1) how long the genotype has been in the population (age); and (2) its performance (fitness).

We compare this approach with previous approaches on the Symbolic Regression problem, sweeping the problem difficulty over a range of solution complexities and number of variables.

Our results indicate that the multi-objective approach identifies the exact target solution more often that the age-layered population and standard population methods. The multi-objective method also performs better on higher complexity problems and higher dimensional datasets—finding global optima with less computational effort.

…Here, we consider using the ALPS concept of age as a fundamental property in the evolutionary optimization. Rather than using age to partition the population into layers, we use age as an independent dimension in a multi-objective Pareto front optimization. In this context, a solution is selected for if it has both higher fitness and lower genotypic age than other solutions.

…2. Algorithm: The age of a solution is measured in generations. All randomly initialized individuals start with age of one. With each generation an individual exists in the population, its age is incremented by one. During crossover and mutation events, the age is inherited as the maximum age of the parents⁶.

The Age-Fitness Pareto Population method uses a single population, in contrast to the population layers in the ALPS algorithm. The algorithm tracks the fitness of each individual as in a normal evolutionary algorithm, and also the genotypic age. The individuals in the population can be thought of lying on a two-dimensional plane of age and fitness, as in Figure 1. The multi-objective optimization task is to identify the non-dominated Pareto front of the problem domain [Multi-Objective Optimization Using Evolutionary Algorithms]; here, the objectives are to maximize the fitness with minimum age.

See Also: