Neat algorithm for simple games

Evolutionary algorithms are a subfield of artificial intelligence, which, even if deep learning is the leading factor nowadays, gained more attention in recent years, due to the increase of processing power. One of the most popular topics in this field is Neuroevolution, since is combines the advantages of evolutionary algorithms with the advantages of neural and deep neural networks. In this paper, we demonstrate how can NeuroEvolution of Augmenting Topologies (NEAT) be used for game playing. The agent that we trained was able to rapidly learn how to play 2 very popular games, namely Dinosaur Game and Flappy Bird, without the need of any specialized computational device (no GPUs were needed). Multiple experiments were conducted, for different scenarios, showing how changing the game environment can change the learning performance of the algorithm.