Implementing a Genetic Neural Network

This paper presents my work on an implementation of an Artificial Neural Network trained with a Genetic Algorithm. The project involved initially randomly generated weights to a neural network which were optimized via the combination of three methods, selection, crossover and mutation, in the goal of imitating genetic evolution. Experiments were done comparing random selection and fitness level probability selection in their efficiency for pairing networks through the generations. Four methods of crossover were designed and tested against each other in transferring weights efficiently. Multiple methods of mutation were created using random number generation, and these methods were tested against each other. The results of the design and trials are a close look at the effects of each method in genetic algorithms