Composing Music with BPTT and LSTM Networks: Comparing Learning and Generalization Aspects
Débora Corrêa, José Hiroki Saito, Sandra Abib · 2008
Many researchers have used neural networks on music composition. The architecture of the network and the representation of the training data have influence on the results. We propose to compare BPTT and LSTM networks in musical composition task with two different pitch representations. We present the learning algorithms of both networks and the results obtained in the composition of new melodies.