Deep Music Analogy Via Latent Representation Disentanglement

• DOI: 10.48550/arxiv.1906.03626
• Publication Date: 2019-01-01
• Authors: Ruihan Yang; Dingsu Wang; Ziyu Wang; Tianyao Chen; Junyan Jiang; Gus Xia
• Keywords: Representation

Analogy-making is a key method for computer algorithms to generate both natural and creative music pieces. In general, an analogy made by partially transferring the abstractions, i.e., high-level representations their relationships, from one piece another; however, this procedure requires disentangling representations, which usually takes little effort musicians but non-trivial computers. Three sub-problems arise: extracting latent observation, so that each part has unique semantic interpretation, mapping back actual music. paper, we contribute explicitly-constrained variational autoencoder (EC$^2$-VAE) as unified solution all three sub-problems. We focus on pitch rhythm of 8-beat clips conditioned chords. producing analogies, model helps us realize imaginary situation "what if" composed using different contour, pattern, or chord progression borrowing other Finally, validate proposed disentanglement objective measurements evaluate examples subjective study.