NEAT, or NeuroEvolution of Augmenting Topologies, is a genetic algorithm developed by Kenneth Stanley and Risto Miikkulainen in 2002 that generates evolving artificial neural networks. It aims to find a balance between fitness and diversity by altering network structures and weights. NEAT stands out for its ability to arrive at effective networks quickly and its direct encoding scheme. The competing conventions problem is solved through tracking gene history. There are different variations of NEAT like rtNEAT and cgNEAT used in various applications, such as designing video game content. odNEAT is an online, decentralized version designed for multi-robot systems to optimize neural network-based controllers continuously.
https://en.wikipedia.org/wiki/Neuroevolution_of_augmenting_topologies