Optimized feature selection using NeuroEvolution of Augmenting Topologies (NEAT)

In most real-world problems, we are dealing with large size datasets. Reducing the number of irrelevant/redundant features dramatically reduces the running time of a learning algorithm and leads to a more general concept. In this paper, realization of feature selection through NeuroEvolution of Augmenting Topologies (NEAT) [1] is investigated which aims to pick a subset of features that are relevant to the target concept. Two major goals in machine learning are discovery and improvement of solutions to complex problems. Complexification, the incremental elaboration of solutions through adding new structure, achieves both these goals. Hence, in this work, the power of complexification through the NEAT method is demonstrated which evolves increasingly complex neural network architectures. When compared to the evolution of networks with fixed structure, NEAT discovers significantly more sophisticated strategies. The results show NEAT can provide better accuracy result than conventional MLP and leads to improve feature selection accuracy.