Augmented Neat Algorithm For Enhanced Cognitive Interaction (NEAT-X)

Abstract

Artificial neural networks (ANNs) are utilized in a
variety of practical applications, from pattern recognition to
controlling robots. Neuroevolution (NE), which involves the
artificial evolution of neural networks through the use of genetic
algorithms, has demonstrated significant potential in tackling
complicated reinforcement learning tasks. This paper provides a
comprehensive overview of the leading methods for evolving
artificial neural networks (ANNs), called NeuroEvolution of
Augmenting Topologies(NEAT). NEAT excels in evolving neural
networks with diverse structures but faces scalability challenges,
especially with extensive networks or high-dimensional input
spaces. As the complexity of the problem increases, the search
space expands exponentially, hindering NEAT's exploration
effectiveness. After performing mutation, we identify the best
mutations, and similar substructures are discovered and added to
the mutation list. The improved version of NEAT algorithm
requires less computational resources and will give optimized
solution. After adding it to the mutation list with some minor
modifications, it is demonstrated that the performance of NEAT
can be improved