Neuroevolution game.
Fall 2021 - Computational Intelligence.
This game has been developed as an assignment for students at Amirkabir University of Technology to apply neuroevolution using a simple game.
I've used FCNN(Fully Connected Neural Network) for the neural network part of this project in nn.py with architecture [4, 10, 1], which means in this NN I've used 4 neurons as input, 10 neurons as hidden layer and 1 neuron for the last layer.
In evolution.py, I've implemented several selection methods such as top-k, roulette wheel and SUS , and I also have implemented Q tournament and roulette wheel for generating new population(in this project new population means new weights for foresaid neural network). In the vectorization part (create_input_vector function in player.py) I've used cosine distance between the player and the first and second obstacles, also the normalized position of the player.