A Novel Version of the Bacterial Memetic Algorithm with Modified Operator Execution Order

The Three Step Bacterial Memetic Algorithm is proposed. This new version of the Bacterial Memetic Algorithm with Modified Operator Execution Order (BMAM) is applied in a practical problem, namely is proposed as the Fuzzy Neural Networks (FNN) training algorithm. This paper strove after the improvement of the function approximation capability of the FNNs by applying a combination of evolutionary and gradient based (global and local search) algorithms. The method interleaves the bacterial mutation (optimizes the rules in one bacterium) and a local seach method applied for each clone with the Levenberg Marquardt method to reach the local optimum. In our novel algorithm various kinds of fast algorithm with less complexity, like Quasi-Newton, Conjugate Gradient, Gradient Descent furthemore Gradient Descent with Adaptive Learning Rate and Momentum are nested in the bacterial mutation. The benefits arising from the combination between various fast local search methods and memetic algorithm have been investigated in this paper.