An Evolved Neural Network/HC Hybrid for Tablature Creation in GA-based Guitar Arranging

In this paper we describe a technique for creating guitar tablature using a neural network. Training data was parsed from an online repository of human-created tablatures. The contents of both the input layer and the set of training data have been optimized through genetic search in order to maximize the accuracy of the network. The output of the network is improved upon with a local heuristic hill-climber (HC). We implement this model in an existing system for generating guitar arrangements via genetic algorithm (GA). When compared to the original system for generating tablature, we note modest improvement in tablature quality and drastic improvements in