The first application of neural networks for music information retrieval we looked at was for categorizing songs into different genres.
Based on "Automatic Musical Pattern Feature Extraction Using Convolutional Neural Network" (Li, T., Chan, A., Chun, A..), we wanted to split music into at most four categories.
Following some brief attempts at coming up with our own categorizations, we opted to use Philip Tagg's axiomatic triangle, a system used by musicologists since the 1980s. This gives us three categories into which all music can be sorted: