Applying Factored Evolutionary Algorithms to the B-Spline Knot Selection Problem

Stochastic search algorithms are an effective method for solving the B-spline knot selection problem. Unfortunately, they often require a large number of (in this case, expensive) fitness function evaluations to find good solutions. In an attempt to mitigate this issue, we apply a modern cooperative coevolutionary algorithm – Factored Evolutionary Algorithms (FEAs) – to the B-spline knot selection problem. We demonstrate FEA’s performance on a variety of benchmark functions, and we compare FEA to traditional stochastic search methods. We also propose an FEA-specific method to evaluate Mean Squared Error (MSE) more efficiently, thus reducing the theoretical runtime.