A Neural Probabilistic Model for Predicting Melodic Sequences

We present an approach for modelling melodic sequences us- ing Restricted Boltzmann Machines, with an application to folk melody classification. 1Results show that this model's predictive performance is slightly better in our experiment than that of previously evaluated n- gram models (7). The model has a simple structure and in our evaluation it scaled linearly in the number of free parameters with length of the modelled context. A set of these models is used to classify 7 different styles of folk melodies with an accuracy of 61.74%.